We present a novel approach to modeling fictional narratives. The proposed model creates embeddings that represent a story such that similar narratives, that is, reformulations of the same story, will result in similar embeddings. We showcase the prowess of our narrative-focused embeddings on various datasets, exhibiting state-of-the-art performance on multiple retrieval tasks. The embeddings also show promising results on a narrative understanding task. Additionally, we perform an annotation-based evaluation to validate that our introduced computational notion of narrative similarity aligns with human perception. The approach can help to explore vast datasets of stories, with potential applications in recommender systems and in the computational analysis of literature.

Current top-performing coreference resolution approaches are limited with regard to the maximum length of texts they can accept. We explore a recursive merging technique of entities that allows us to apply coreference models to texts of arbitrary length, as found in many narrative genres. In experiments on established datasets, we quantify the drop in resolution quality caused by this approach. Finally, we use an under-explored resource in the form of a fully coreference-annotated novel to illustrate our model’s performance for long documents in practice. Here, we achieve state-of-the-art performance, outperforming previous systems capable of handling long documents.

Large Language Models (LLMs) have emerged as dominant foundational models in modern NLP. However, the understanding of their prediction processes and internal mechanisms, such as feed-forward networks (FFN) and multi-head self-attention (MHSA), remains largely unexplored. In this work, we probe LLMs from a human behavioral perspective, correlating values from LLMs with eye-tracking measures, which are widely recognized as meaningful indicators of human reading patterns. Our findings reveal that LLMs exhibit a similar prediction pattern with humans but distinct from that of Shallow Language Models (SLMs). Moreover, with the escalation of LLM layers from the middle layers, the correlation coefficients also increase in FFN and MHSA, indicating that the logits within FFN increasingly encapsulate word semantics suitable for predicting tokens from the vocabulary.

In contemporary society, the proliferation of hate speech is increasingly prevalent across various social media platforms, with a notable trend of incorporating memes to amplify its visual impact and reach. The conventional text-based detection approaches frequently fail to address the complexities introduced by memes, thereby aggravating the challenges, particularly in low-resource languages such as Amharic. We develop Amharic meme hate speech detection models using 2,000 memes collected from Facebook, Twitter, and Telegram over four months. We employ native Amharic speakers to annotate each meme using a web-based tool, yielding a Fleiss’ kappa score of 0.50. We utilize different feature extraction techniques, namely VGG16 for images and word2Vec for textual content, and build unimodal and multimodal models such as LSTM, BiLSTM, and CNN. The BiLSTM model shows the best performance, achieving 63% accuracy for text and 75% for multimodal features. In image-only experiments, the CNN model achieves 69% in accuracy. Multimodal models demonstrate superior performance in detecting Amharic hate speech in memes, showcasing their potential to address the unique challenges posed by meme-based hate speech on social media.

The prevalence of digital media and evolving sociopolitical dynamics have significantly amplified the dissemination of hateful content. Existing studies mainly focus on classifying texts into binary categories, often overlooking the continuous spectrum of offensiveness and hatefulness inherent in the text. In this research, we present an extensive benchmark dataset for Amharic, comprising 8,258 tweets annotated for three distinct tasks: category classification, identification of hate targets, and rating offensiveness and hatefulness intensities. Our study highlights that a considerable majority of tweets belong to the less offensive and less hate intensity levels, underscoring the need for early interventions by stakeholders. The prevalence of ethnic and political hatred targets, with significant overlaps in our dataset, emphasizes the complex relationships within Ethiopia’s sociopolitical landscape. We build classification and regression models and investigate the efficacy of models in handling these tasks. Our results reveal that hate and offensive speech can not be addressed by a simplistic binary classification, instead manifesting as variables across a continuous range of values. The afro-XLMR-large model exhibits the best performances achieving F1-scores of 75.30%, 70.59%, and 29.42% for the category, target, and regression tasks, respectively. The 80.22% correlation coefficient of the Afro-XLMR-large model indicates strong alignments.

This paper presents WISMIR3, a multi-modal dataset comprising roughly 300K text-image pairs from Wikipedia. With a sophisticated automatic ETL pipeline, we scraped, filtered, and transformed the data so that WISMIR3 intrinsically differs from other popular text-image datasets like COCO and Flickr30k. We prove this difference by comparing various linguistic statistics between the three datasets computed using the pipeline. The primary purpose of WISMIR3 is to use it as a benchmark to challenge state-of-the-art text-image retrieval approaches, which already reach around 90% Recall@5 scores on the mentioned popular datasets. Therefore, we ran several text-image retrieval experiments on our dataset using current models, which show that the models, in fact, perform significantly worse compared to evaluation results on COCO and Flickr30k. In addition, for each text-image pair, we release features computed by Faster-R-CNN and CLIP models. With this, we want to ease and motivate the use of the dataset for other researchers.

Exploring and quantifying semantic relatedness is central to representing language and holds significant implications across various NLP tasks. While earlier NLP research primarily focused on semantic similarity, often within the English language context, we instead investigate the broader phenomenon of semantic relatedness. In this paper, we present SemRel, a new semantic relatedness dataset collection annotated by native speakers across 13 languages: Afrikaans, Algerian Arabic, Amharic, English, Hausa, Hindi, Indonesian, Kinyarwanda, Marathi, Moroccan Arabic, Modern Standard Arabic, Spanish, and Telugu. These languages originate from five distinct language families and are predominantly spoken in Africa and Asia – regions characterised by a relatively limited availability of NLP resources. Each instance in the SemRel datasets is a sentence pair associated with a score that represents the degree of semantic textual relatedness between the two sentences. The scores are obtained using a comparative annotation framework. We describe the data collection and annotation processes, challenges when building the datasets, baseline experiments, and their impact and utility in NLP.

Large Vision-Language Models (LVLMs) are increasingly adept at generating contextually detailed and coherent responses from visual inputs. However, their application in multimodal decision-making and open-ended generation is hindered by a notable rate of hallucinations, where generated text inaccurately represents the visual contents. To address this issue, this paper introduces the Instruction Contrastive Decoding (ICD) method, a novel approach designed to reduce hallucinations during LVLM inference. Our method is inspired by our observation that what we call disturbance instructions significantly exacerbate hallucinations in multimodal fusion modules. ICD contrasts distributions from standard and instruction disturbance, thereby increasing alignment uncertainty and effectively subtracting hallucinated concepts from the original distribution. Through comprehensive experiments on discriminative benchmarks (POPE and MME) and a generative benchmark (LLaVa-Bench), we demonstrate that ICD significantly mitigates both object-level and attribute-level hallucinations. Moreover, our method not only addresses hallucinations but also significantly enhances the general perception and recognition capabilities of LVLMs.

We present a new approach called MeritOpt based on the Personalized Federated Learning algorithm MeritFed that can be applied to Natural Language Tasks with heterogeneous data. We evaluate it on the Low-Resource Machine Translation task, using the datasets of South East Asian and Finno-Ugric languages. In addition to its effectiveness, MeritOpt is also highly interpretable, as it can be applied to track the impact of each language used for training. Our analysis reveals that target dataset size affects weight distribution across auxiliary languages, that unrelated languages do not interfere with the training, and auxiliary optimizer parameters have minimal impact. Our approach is easy to apply with a few lines of code, and we provide scripts for reproducing the experiments (https://github.com/VityaVitalich/MeritOpt).

Images increasingly constitute a larger portion of internet content, encoding even more complex meanings. Recent studies have highlight the pivotal role of visual communication in the spread of extremist content, particularly that associated with right-wing political ideologies. However, the capability of machine learning systems to recognize such meanings, sometimes implicit, remains limited. To enable future research in this area, we introduce and release VIDA, the Visual Incel Data Archive, a multimodal dataset comprising visual material and internet memes collected from two main Incel communities (Italian and Anglophone) known for their extremist misogynistic content. Following the analytical framework of Shifman (2014), we propose a new taxonomy for annotation across three main levels of analysis: content, form, and stance (hate). This allows for the association of images with fine-grained contextual information that help to identify the presence of offensiveness and a broader set of cultural references, enhancing the understanding of more nuanced aspects in visual communication. In this work we present a statistical analysis of the annotated dataset as well as discuss annotation examples and future line of research.

In this system demonstration paper, we present the Concept Over Time Analysis extension for the Discourse Analysis Tool Suite.The proposed tool empowers users to define, refine, and visualize their concepts of interest within an interactive interface. Adhering to the Human-in-the-loop paradigm, users can give feedback through sentence annotations. Utilizing few-shot sentence classification, the system employs Sentence Transformers to compute representations of sentences and concepts. Through an iterative process involving semantic similarity searches, sentence annotation, and fine-tuning with contrastive data, the model continuously refines, providing users with enhanced analysis outcomes. The final output is a timeline visualization of sentences classified to concepts. Especially suited for the Digital Humanities, Concept Over Time Analysis serves as a valuable tool for qualitative data analysis within extensive datasets. The chronological overview of concepts enables researchers to uncover patterns, trends, and shifts in discourse over time.

This paper explores the potential of Large Language Models (LLMs) to enhance qualitative data analysis (QDA) workflows within the open-source QDA platform developed at our university. We identify several opportunities within a typical QDA workflow where AI assistance can boost researcher productivity and translate these opportunities into corresponding NLP tasks: document classification, information extraction, span classification, and text generation. A benchmark tailored to these QDA activities is constructed, utilizing English and German datasets that align with relevant use cases. Focusing on efficiency and accessibility, we evaluate the performance of three prominent open-source LLMs - Llama 3.1, Gemma 2, and Mistral NeMo - on this benchmark. Our findings reveal the promise of LLM integration for streamlining QDA workflows, particularly for English-language projects. Consequently, we have implemented the LLM Assistant as an opt-in feature within our platform and report the implementation details. With this, we hope to further democratize access to AI capabilities for qualitative data analysis.

This paper presents UHH’s approach developed for the AVeriTeC shared task. The goal of the challenge is to verify given real-world claims with evidences from the Web. In this shared task, we investigate a Retrieval-Augmented Generation (RAG) model, which mainly contains retrieval, generation, and augmentation components. We start with the selection of the top 10k evidences via BM25 scores, and continue with two approaches to retrieve the most similar evidences: (1) to retrieve top 10 evidences through vector similarity, generate questions for them, and rerank them or (2) to generate questions for the claim and retrieve the most similar evidence, again, through vector similarity. After retrieving the top evidences, a Large Language Model (LLM) is prompted using the claim along with either all evidences or individual evidence to predict the label. Our system submission, UHH, using the first approach and individual evidence prompts, ranks 6th out of 23 systems.

Large Language Models (LLMs) such as ChatGPT, GitHub Copilot, Llama, or Mistral assist programmers as copilots and knowledge sources to make the coding process faster and more efficient. This paper aims to improve the copilot performance by implementing different self-alignment processes and retrieval-augmented generation (RAG) pipelines, as well as their combination. To test the effectiveness of all approaches, we create a dataset and apply a model-based evaluation, using LLM as a judge. It is designed to check the model’s abilities to understand the source code semantics, the dependency between files, and the overall meta-information about the repository. We also compare our approach with other existing solutions, e.g. ChatGPT-3.5, and evaluate on the existing benchmarks. Code and dataset are available online (https://anonymous.4open.science/r/ma_llm-382D).

This paper presents the Sövereign submission for the shared task on perspective argument retrieval for the Argument Mining Workshop 2024. The main challenge is to perform argument retrieval considering socio-cultural aspects such as political interests, occupation, age, and gender. To address the challenge, we apply open-access Large Language Models (Mistral-7b) in a zero-shot fashion for re-ranking and explicit similarity scoring. Additionally, we combine different features in an ensemble setup using logistic regression. Our system ranks second in the competition for all test set rounds on average for the logistic regression approach using LLM similarity scores as a feature. In addition to the description of the approach, we also provide further results of our ablation study. Our code will be open-sourced upon acceptance.

Comparative Question Answering (CompQA) is a Natural Language Processing task that combines Question Answering and Argument Mining approaches to answer subjective comparative questions in an efficient argumentative manner. In this paper, we present an end-to-end (full pipeline) system for answering comparative questions called CAM 2.0 as well as a public leaderboard called CompUGE that unifies the existing datasets under a single easy-to-use evaluation suite. As compared to previous web-form-based CompQA systems, it features question identification, object and aspect labeling, stance classification, and summarization using up-to-date models. We also select the most time- and memory-effective pipeline by comparing separately fine-tuned Transformer Encoder models which show state-of-the-art performance on the subtasks with Generative LLMs in few-shot and LoRA setups. We also conduct a user study for a whole-system evaluation.

Extracting who says what to whom is a crucial part in analyzing human communication in today’s abundance of data such as online news articles. Yet, the lack of annotated data for this task in German news articles severely limits the quality and usability of possible systems. To remedy this, we present a new, freely available, creative-commons-licensed dataset for quotation attribution in German news articles based on WIKINEWS. The dataset provides curated, high-quality annotations across 1000 documents (250,000 tokens) in a fine-grained annotation schema enabling various downstream uses for the dataset. The annotations not only specify who said what but also how, in which context, to whom and define the type of quotation. We specify our annotation schema, describe the creation of the dataset and provide a quantitative analysis. Further, we describe suitable evaluation metrics, apply two existing systems for quotation attribution, discuss their results to evaluate the utility of our dataset and outline use cases of our dataset in downstream tasks.

In this system demonstration paper, we describe the Whiteboards extension for an existing web-based platform for digital qualitative discourse analysis. Whiteboards comprise interactive graph-based interfaces to organize and manipulate objects, which can be qualitative research data, such as documents, images, etc., and analyses of these research data, such as annotations, tags, and code structures. The proposed extension offers a customizable view of the material and a wide range of actions that enable new ways of interacting and working with such resources. We show that the visualizations facilitate various use cases of qualitative data analysis, including reflection of the research process through sampling maps, creation of actor networks, and refining code taxonomies.

With the emergence of numerous Large Language Models (LLM), the usage of such models in various Natural Language Processing (NLP) applications is increasing extensively. Counterspeech generation is one such key task where efforts are made to develop generative models by fine-tuning LLMs with hatespeech - counterspeech pairs, but none of these attempts explores the intrinsic properties of large language models in zero-shot settings. In this work, we present a comprehensive analysis of the performances of four LLMs namely GPT-2, DialoGPT, ChatGPT and FlanT5 in zero-shot settings for counterspeech generation, which is the first of its kind. For GPT-2 and DialoGPT, we further investigate the deviation in performance with respect to the sizes (small, medium, large) of the models. On the other hand, we propose three different prompting strategies for generating different types of counterspeech and analyse the impact of such strategies on the performance of the models. Our analysis shows that there is an improvement in generation quality for two datasets (17%), however the toxicity increase (25%) with increase in model size. Considering type of model, GPT-2 and FlanT5 models are significantly better in terms of counterspeech quality but also have high toxicity as compared to DialoGPT. ChatGPT are much better at generating counter speech than other models across all metrics. In terms of prompting, we find that our proposed strategies help in improving counter speech generation across all the models.

A wide body of research is concerned with the semantics of narratives, both in terms of understanding narratives and generating fictional narratives and stories. We provide a dataset of summaries to be used as a proxy for entire stories or for the analysis of the summaries themselves. Our dataset consists of a total of 96,831 individual summaries across 29,505 stories. We intend for the dataset to be used for training and evaluation of embedding representations for stories, specifically the stories’ narratives. The summary data is harvested from five different language versions of Wikipedia. Our dataset comes with rich metadata, which we extract from Wikidata, enabling a wide range of applications that operate on story summaries in conjunction with metadata. To set baseline results, we run retrieval experiments on the dataset, exploring the capability of similarity models in retrieving summaries of the same story. For this retrieval, a crucial element is to not place too much emphasis on the named entities, as this can enable retrieval of other summaries for the same work without taking the narrative into account.

The objective of the SemEval-2023 Task 1: Visual Word Sense Disambiguation (VWSD) is to identify the image illustrating the indented meaning of a target word and some minimal additional context. The omnipresence of textual and visual data in the task strongly suggests the utilization of the recent advances in multi-modal machine learning, i.e., pretrained visiolinguistic models (VLMs). Often referred to as foundation models due to their strong performance on many vision-language downstream tasks, these models further demonstrate powerful zero-shot capabilities. In this work, we utilize various pertained VLMs in a zero-shot fashion for multiple approaches using external knowledge sources to enrich the contextual information. Further, we evaluate our methods on the final test data and extensively analyze the suitability of different knowledge sources, the influence of training data, model sizes, multi-linguality, and different textual prompting strategies. Although we are not among the best-performing systems (rank 20 of 56), our experiments described in this work prove competitive results. Moreover, we aim to contribute meaningful insights and propel multi-modal machine learning tasks like VWSD.

This work introduces the D-WISE Tool Suite (DWTS), a novel working environment for digital qualitative discourse analysis in the Digital Humanities (DH). The DWTS addresses limitations of current DH tools induced by the ever-increasing amount of heterogeneous, unstructured, and multi-modal data in which the discourses of contemporary societies are encoded. To provide meaningful insights from such data, our system leverages and combines state-of-the-art machine learning technologies from Natural Language Processing and Com-puter Vision. Further, the DWTS is conceived and developed by an interdisciplinary team ofcultural anthropologists and computer scientists to ensure the tool’s usability for modernDH research. Central features of the DWTS are: a) import of multi-modal data like text, image, audio, and video b) preprocessing pipelines for automatic annotations c) lexical and semantic search of documents d) manual span, bounding box, time-span, and frame annotations e) documentation of the research process.

In this work, we analyse the role of output vocabulary for text-to-text (T2T) models on the task of SPARQL semantic parsing. We perform experiments within the the context of knowledge graph question answering (KGQA), where the task is to convert questions in natural language to the SPARQL query language. We observe that the query vocabulary is distinct from human vocabulary. Language Models (LMs) are pre-dominantly trained for human language tasks, and hence, if the query vocabulary is replaced with a vocabulary more attuned to the LM tokenizer, the performance of models may improve. We carry out carefully selected vocabulary substitutions on the queries and find absolute gains in the range of 17% on the GrailQA dataset.

WebAnno is one of the most popular annotation tools that supports generic annotation types and distributive annotation with multiple user roles. However, WebAnno focuses on annotating span-level mentions and relations among them, making document-level annotation complicated. When it comes to the annotation and analysis of social science materials, it usually involves the creation of codes to categorize a given document. The codes, which are known as codebooks, are typically hierarchical, which enables to code the document either with a general category or more fine-grained subcategories. CodeAnno is forked from WebAnno and designed to solve the coding problems faced by many social science researchers with the following main functionalities. 1) Creation of hierarchical codebooks, with functionality to move and sort categories in the hierarchy 2) an interactive UI for codebook annotation 3) import and export of annotations in CSV format, hence being compatible with existing annotations conducted using spreadsheet applications 4) integration of an external automation component to facilitate coding using machine learning 5) project templating that allows duplicating a project structure without copying the actual documents. We present different use-cases to demonstrate the capability of CodeAnno. A shot demonstration video of the system is available here: https://www.youtube.com/watch?v=RmCdTghBe-s

The rise of social media eases the spread of hateful content, especially racist content with severe consequences. In this paper, we analyze the tweets targeting the death of George Floyd in May 2020 as the event accelerated debates on racism globally. We focus on the tweets published in French for a period of one month since the death of Floyd. Using the Yandex Toloka platform, we annotate the tweets into categories as hate, offensive or normal. Tweets that are offensive or hateful are further annotated as racial or non-racial. We build French hate speech detection models based on the multilingual BERT and CamemBERT and apply transfer learning by fine-tuning the HateXplain model. We compare different approaches to resolve annotation ties and find that the detection model based on CamemBERT yields the best results in our experiments.

In this paper, we present a study of efficient data selection and annotation strategies for Amharic hate speech. We also build various classification models and investigate the challenges of hate speech data selection, annotation, and classification for the Amharic language. From a total of over 18 million tweets in our Twitter corpus, 15.1k tweets are annotated by two independent native speakers, and a Cohen’s kappa score of 0.48 is achieved. A third annotator, a curator, is also employed to decide on the final gold labels. We employ both classical machine learning and deep learning approaches, which include fine-tuning AmFLAIR and AmRoBERTa contextual embedding models. Among all the models, AmFLAIR achieves the best performance with an F1-score of 72%. We publicly release the annotation guidelines, keywords/lexicon entries, datasets, models, and associated scripts with a permissive license.

One of the many advantages of pre-trained language models (PLMs) such as BERT and RoBERTa is their flexibility and contextual nature. These features give PLMs strong capabilities for representing lexical semantics. However, PLMs seem incapable of capturing high-level semantics in terms of compositionally. We show that when augmented with the relevant semantic knowledge, PMLs learn to capture a higher degree of lexical compositionality. We annotate a large dataset from Wikidata highlighting a type of semantic inference that is easy for humans to understand but difficult for PLMs, like the correlation between age and date of birth. We use this resource for finetuning DistilBERT, BERT large and RoBERTa. Our results show that the performance of PLMs against the test data continuously improves when augmented with such a rich resource. Our results are corroborated by a consistent improvement over most GLUE benchmark natural language understanding tasks.

We present a novel approach to modeling narratives using narrative chain embeddings.A new dataset of narrative chains extracted from German news texts is presented. With neural methods, we produce models for both German and English that achieve state-of-the-art performance on the Multiple Choice Narrative Cloze task. Subsequently, we perform an extrinsic evaluation of the embeddings our models produce and show that they perform rather poorly in identifying narratively similar texts. We explore some of the reasons for this underperformance and discuss the upsides of our approach. We provide an outlook on alternative ways to model narratives, as well as techniques for evaluating such models.

In this work, we build a Question Answering (QA) classification dataset from a social media platform, namely the Telegram public channel called @AskAnythingEthiopia. The channel has more than 78k subscribers and has existed since May 31, 2019. The platform allows asking questions that belong to various domains, like politics, economics, health, education, and so on. Since the questions are posed in a mixed-code, we apply different strategies to pre-process the dataset. Questions are posted in Amharic, English, or Amharic but in a Latin script. As part of the pre-processing tools, we build a Latin to Ethiopic Script transliteration tool. We collect 8k Amharic and 24K transliterated questions and develop deep learning-based questions answering classifiers that attain as high as an F-score of 57.29 in 20 different question classes or categories. The datasets and pre-processing scripts are open-sourced to facilitate further research on the Amharic community-based question answering.

Referential gaze is a fundamental phenomenon for psycholinguistics and human-human communication. However, modeling referential gaze for real-world scenarios, e.g. for task-oriented communication, is lacking the well-deserved attention from the NLP community. In this paper, we address this challenging issue by proposing a novel multimodal NLP task; namely predicting when the gaze is referential. We further investigate how to model referential gaze and transfer gaze features to adapt to unseen situated settings that target different referential complexities than the training environment. We train (i) a sequential attention-based LSTM model and (ii) a multivariate transformer encoder architecture to predict whether the gaze is on a referent object. The models are evaluated on the three complexity datasets. The results indicate that the gaze features can be transferred not only among various similar tasks and scenes but also across various complexity levels. Taking the referential complexity of a scene into account is important for successful target prediction using gaze parameters especially when there is not much data for fine-tuning.

We present a dataset containing source code solutions to algorithmic programming exercises solved by hundreds of Bachelor-level students at the University of Hamburg. These solutions were collected during the winter semesters 2019/2020, 2020/2021 and 2021/2022. The dataset contains a set of solutions to a total of 21 tasks written in Java as well as Python and a total of over 1500 individual solutions. All solutions were submitted through Moodle and the Coderunner plugin and passed a number of test cases (including randomized tests), such that they can be considered as working correctly. All students whose solutions are included in the dataset gave their consent into publishing their solutions. The solutions are pseudonymized with a random solution ID. Included in this paper is a short analysis of the dataset containing statistical data and highlighting a few anomalies (e.g. the number of solutions per task decreases for the last few tasks due to grading rules). We plan to extend the dataset with tasks and solutions from upcoming courses.

MOTIF (MultimOdal ConTextualized Images For Language Learners) is a multimodal dataset that consists of 1125 comprehension texts retrieved from Wikipedia Simple Corpus. Allowing multimodal processing or enriching the context with multimodal information has proven imperative for many learning tasks, specifically for second language (L2) learning. In this respect, several traditional NLP approaches can assist L2 readers in text comprehension processes, such as simplifying text or giving dictionary descriptions for complex words. As nicely stated in the well-known proverb, sometimes “a picture is worth a thousand words” and an image can successfully complement the verbal message by enriching the representation, like in Pictionary books. This multimodal support can also assist on-the-fly text reading experience by providing a multimodal tool that chooses and displays the most relevant images for the difficult words, given the text context. This study mainly focuses on one of the key components to achieving this goal; collecting a multimodal dataset enriched with complex word annotation and validated image match.

Comparative Question Answering (cQA) is the task of providing concrete and accurate responses to queries such as: “Is Lyft cheaper than a regular taxi?” or “What makes a mortgage different from a regular loan?”. In this paper, we propose two new open-domain real-world datasets for identifying and labeling comparative questions. While the first dataset contains instances of English questions labeled as comparative vs. non-comparative, the second dataset provides additional labels including the objects and the aspects of comparison. We conduct several experiments that evaluate the soundness of our datasets. The evaluation of our datasets using various classifiers show promising results that reach close-to-human results on a binary classification task with a neural model using ALBERT embeddings. When approaching the unsupervised sequence labeling task, some headroom remains.

The multi-modal foundation model CLIP computes representations from texts and images that achieved unprecedented performance on tasks such as zero-shot image classification. However, CLIP was pretrained on public internet data. Thus it lacks highly domain-specific knowledge. We investigate the adaptation of CLIP-based models to the chest radiography domain using the MIMIC-CXR dataset. We show that the features of the pretrained CLIP models do not transfer to this domain. We adapt CLIP to the chest radiography domain using contrastive language supervision and show that this approach yields a model that outperforms supervised learning on labels on the MIMIC-CXR dataset while also generalizing to the CheXpert and RSNA Pneumonia datasets. Furthermore, we do a detailed ablation study of the batch and dataset size. Finally, we show that language supervision allows for better explainability by using the multi-modal model to generate images from texts such that experts can inspect what the model has learned.

The corona pandemic and countermeasures such as social distancing and lockdowns have confronted individuals with new challenges for their mental health and well-being. It can be assumed that the Jungian psychology types of extraverts and introverts react differently to these challenges. We propose a Bi-LSTM model with an attention mechanism for classifying introversion and extraversion from German tweets, which is trained on hand-labeled data created by 335 participants. With this work, we provide this novel dataset for free use and validation. The proposed model achieves solid performance with F1 = .72. Furthermore, we created a feature engineered logistic model tree (LMT) trained on hand-labeled tweets, to which the data is also made available with this work. With this second model, German tweets before and during the pandemic have been investigated. Extraverts display more positive emotions, whilst introverts show more insight and higher rates of anxiety. Even though such a model can not replace proper psychological diagnostics, it can help shed light on linguistic markers and to help understand introversion and extraversion better for a variety of applications and investigations.

Recent research using pre-trained language models for multi-document summarization task lacks deep investigation of potential erroneous cases and their possible application on other languages. In this work, we apply a pre-trained language model (BART) for multi-document summarization (MDS) task using both fine-tuning and without fine-tuning. We use two English datasets and one German dataset for this study. First, we reproduce the multi-document summaries for English language by following one of the recent studies. Next, we show the applicability of the model to German language by achieving state-of-the-art performance on German MDS. We perform an in-depth error analysis of the followed approach for both languages, which leads us to identifying most notable errors, from made-up facts and topic delimitation, and quantifying the amount of extractiveness.

Lexical complexity is a highly subjective notion, yet this factor is often neglected in lexical simplification and readability systems which use a ”one-size-fits-all” approach. In this paper, we investigate which aspects contribute to the notion of lexical complexity in various groups of readers, focusing on native and non-native speakers of English, and how the notion of complexity changes depending on the proficiency level of a non-native reader. To facilitate reproducibility of our approach and foster further research into these aspects, we release a dataset of complex words annotated by readers with different backgrounds.

ActiveAnno is an annotation tool focused on document-level annotation tasks developed both for industry and research settings. It is designed to be a general-purpose tool with a wide variety of use cases. It features a modern and responsive web UI for creating annotation projects, conducting annotations, adjudicating disagreements, and analyzing annotation results. ActiveAnno embeds a highly configurable and interactive user interface. The tool also integrates a RESTful API that enables integration into other software systems, including an API for machine learning integration. ActiveAnno is built with extensible design and easy deployment in mind, all to enable users to perform annotation tasks with high efficiency and high-quality annotation results.

In this thesis proposal, we explore the application of event extraction to literary texts. Considering the lengths of literary documents modeling events in different granularities may be more adequate to extract meaningful information, as individual elements contribute little to the overall semantics. We adapt the concept of schemas as sequences of events all describing a single process, connected through shared participants extending it to for multiple schemas in a document. Segmentation of event sequences into schemas is approached by modeling event sequences, on such task as the narrative cloze task, the prediction of missing events in sequences. We propose building on sequences of event embeddings to form schema embeddings, thereby summarizing sections of documents using a single representation. This approach will allow for the comparisons of different sections of documents and entire literary works. Literature is a challenging domain based on its variety of genres, yet the representation of literary content has received relatively little attention.

In primary school, children’s books, as well as in modern language learning apps, multi-modal learning strategies like illustrations of terms and phrases are used to support reading comprehension. Also, several studies in educational psychology suggest that integrating cross-modal information will improve reading comprehension. We claim that state-of- he-art multi-modal transformers, which could be used in a language learner context to improve human reading, will perform poorly because of the short and relatively simple textual data those models are trained with. To prove our hypotheses, we collected a new multi-modal image-retrieval dataset based on data from Wikipedia. In an in-depth data analysis, we highlight the differences between our dataset and other popular datasets. Additionally, we evaluate several state-of-the-art multi-modal transformers on text-image retrieval on our dataset and analyze their meager results, which verify our claims.

With the increasing number of user comments in diverse domains, including comments on online journalism and e-commerce websites, the manual content analysis of these comments becomes time-consuming and challenging. However, research showed that user comments contain useful information for different domain experts, which is thus worth finding and utilizing. This paper introduces Forum 4.0, an open-source framework to semi-automatically analyze, aggregate, and visualize user comments based on labels defined by domain experts. We demonstrate the applicability of Forum 4.0 with comments analytics scenarios within the domains of online journalism and app stores. We outline the underlying container architecture, including the web-based user interface, the machine learning component, and the task manager for time-consuming tasks. We finally conduct machine learning experiments with simulated annotations and different sampling strategies on existing datasets from both domains to evaluate Forum 4.0’s performance. Forum 4.0 achieves promising classification results (ROC-AUC ≥ 0.9 with 100 annotated samples), utilizing transformer-based embeddings with a lightweight logistic regression model. We explain how Forum 4.0’s architecture is applicable for millions of user comments in real-time, yet at feasible training and classification costs.

We present Sense Clustering over Time (SCoT), a novel network-based tool for analysing lexical change. SCoT represents the meanings of a word as clusters of similar words. It visualises their formation, change, and demise. There are two main approaches to the exploration of dynamic networks: the discrete one compares a series of clustered graphs from separate points in time. The continuous one analyses the changes of one dynamic network over a time-span. SCoT offers a new hybrid solution. First, it aggregates time-stamped documents into intervals and calculates one sense graph per discrete interval. Then, it merges the static graphs to a new type of dynamic semantic neighbourhood graph over time. The resulting sense clusters offer uniquely detailed insights into lexical change over continuous intervals with model transparency and provenance. SCoT has been successfully used in a European study on the changing meaning of ‘crisis’.

We present a system for answering comparative questions (Is X better than Y with respect to Z?) in natural language. Answering such questions is important for assisting humans in making informed decisions. The key component of our system is a natural language interface for comparative QA that can be used in personal assistants, chatbots, and similar NLP devices. Comparative QA is a challenging NLP task, since it requires collecting support evidence from many different sources, and direct comparisons of rare objects may be not available even on the entire Web. We take the first step towards a solution for such a task offering a testbed for comparative QA in natural language by probing several methods, making the three best ones available as an online demo.

To build machine learning-based applications for sensitive domains like medical, legal, etc. where the digitized text contains private information, anonymization of text is required for preserving privacy. Sequence tagging, e.g. as done in Named Entity Recognition (NER) can help to detect private information. However, to train sequence tagging models, a sufficient amount of labeled data are required but for privacy-sensitive domains, such labeled data also can not be shared directly. In this paper, we investigate the applicability of a privacy-preserving framework for sequence tagging tasks, specifically NER. Hence, we analyze a framework for the NER task, which incorporates two levels of privacy protection. Firstly, we deploy a federated learning (FL) framework where the labeled data are not shared with the centralized server as well as the peer clients. Secondly, we apply differential privacy (DP) while the models are being trained in each client instance. While both privacy measures are suitable for privacy-aware models, their combination results in unstable models. To our knowledge, this is the first study of its kind on privacy-aware sequence tagging models.

Pretrained language models (PTLMs) yield state-of-the-art performance on many natural language processing tasks, including syntax, semantics and commonsense. In this paper, we focus on identifying to what extent do PTLMs capture semantic attributes and their values, e.g., the correlation between rich and high net worth. We use PTLMs to predict masked tokens using patterns and lists of items from Wikidata in order to verify how likely PTLMs encode semantic attributes along with their values. Such inferences based on semantics are intuitive for humans as part of our language understanding. Since PTLMs are trained on large amount of Wikipedia data we would assume that they can generate similar predictions, yet our findings reveal that PTLMs are still much worse than humans on this task. We show evidence and analysis explaining how to exploit our methodology to integrate better context and semantics into PTLMs using knowledge bases.

This paper presents the study of sentiment analysis for Amharic social media texts. As the number of social media users is ever-increasing, social media platforms would like to understand the latent meaning and sentiments of a text to enhance decision-making procedures. However, low-resource languages such as Amharic have received less attention due to several reasons such as lack of well-annotated datasets, unavailability of computing resources, and fewer or no expert researchers in the area. This research addresses three main research questions. We first explore the suitability of existing tools for the sentiment analysis task. Annotation tools are scarce to support large-scale annotation tasks in Amharic. Also, the existing crowdsourcing platforms do not support Amharic text annotation. Hence, we build a social-network-friendly annotation tool called ‘ASAB’ using the Telegram bot. We collect 9.4k tweets, where each tweet is annotated by three Telegram users. Moreover, we explore the suitability of machine learning approaches for Amharic sentiment analysis. The FLAIR deep learning text classifier, based on network embeddings that are computed from a distributional thesaurus, outperforms other supervised classifiers. We further investigate the challenges in building a sentiment analysis system for Amharic and we found that the widespread usage of sarcasm and figurative speech are the main issues in dealing with the problem. To advance the sentiment analysis research in Amharic and other related low-resource languages, we release the dataset, the annotation tool, source code, and models publicly under a permissive.

The COVID-19 pandemic has caused international social tension and unrest. Besides the crisis itself, there are growing signs of rising conflict potential of societies around the world. Indicators of global mood changes are hard to detect and direct questionnaires suffer from social desirability biases. However, so-called implicit methods can reveal humans intrinsic desires from e.g. social media texts. We present psychologically validated social unrest predictors and replicate scalable and automated predictions, setting a new state of the art on a recent German shared task dataset. We employ this model to investigate a change of language towards social unrest during the COVID-19 pandemic by comparing established psychological predictors on samples of tweets from spring 2019 with spring 2020. The results show a significant increase of the conflict indicating psychometrics. With this work, we demonstrate the applicability of automated NLP-based approaches to quantitative psychological research.

Fine-tuning of pre-trained transformer networks such as BERT yield state-of-the-art results for text classification tasks. Typically, fine-tuning is performed on task-specific training datasets in a supervised manner. One can also fine-tune in unsupervised manner beforehand by further pre-training the masked language modeling (MLM) task. Hereby, in-domain data for unsupervised MLM resembling the actual classification target dataset allows for domain adaptation of the model. In this paper, we compare current pre-trained transformer networks with and without MLM fine-tuning on their performance for offensive language detection. Our MLM fine-tuned RoBERTa-based classifier officially ranks 1st in the SemEval 2020 Shared Task 12 for the English language. Further experiments with the ALBERT model even surpass this result.

We present the first approach to automatically building resources for academic writing. The aim is to build a writing aid system that automatically edits a text so that it better adheres to the academic style of writing. On top of existing academic resources, such as the Corpus of Contemporary American English (COCA) academic Word List, the New Academic Word List, and the Academic Collocation List, we also explore how to dynamically build such resources that would be used to automatically identify informal or non-academic words or phrases. The resources are compiled using different generic approaches that can be extended for different domains and languages. We describe the evaluation of resources with a system implementation. The system consists of an informal word identification (IWI), academic candidate paraphrase generation, and paraphrase ranking components. To generate candidates and rank them in context, we have used the PPDB and WordNet paraphrase resources. We use the Concepts in Context (CoInCO) “All-Words” lexical substitution dataset both for the informal word identification and paraphrase generation experiments. Our informal word identification component achieves an F-1 score of 82%, significantly outperforming a stratified classifier baseline. The main contribution of this work is a domain-independent methodology to build targeted resources for writing aids.

Disambiguation of word senses in context is easy for humans, but is a major challenge for automatic approaches. Sophisticated supervised and knowledge-based models were developed to solve this task. However, (i) the inherent Zipfian distribution of supervised training instances for a given word and/or (ii) the quality of linguistic knowledge representations motivate the development of completely unsupervised and knowledge-free approaches to word sense disambiguation (WSD). They are particularly useful for under-resourced languages which do not have any resources for building either supervised and/or knowledge-based models. In this paper, we present a method that takes as input a standard pre-trained word embedding model and induces a fully-fledged word sense inventory, which can be used for disambiguation in context. We use this method to induce a collection of sense inventories for 158 languages on the basis of the original pre-trained fastText word embeddings by Grave et al., (2018), enabling WSD in these languages. Models and system are available online.

The corpus, from which a predictive language model is trained, can be considered the experience of a semantic system. We recorded everyday reading of two participants for two months on a tablet, generating individual corpus samples of 300/500K tokens. Then we trained word2vec models from individual corpora and a 70 million-sentence newspaper corpus to obtain individual and norm-based long-term memory structure. To test whether individual corpora can make better predictions for a cognitive task of long-term memory retrieval, we generated stimulus materials consisting of 134 sentences with uncorrelated individual and norm-based word probabilities. For the subsequent eye tracking study 1-2 months later, our regression analyses revealed that individual, but not norm-corpus-based word probabilities can account for first-fixation duration and first-pass gaze duration. Word length additionally affected gaze duration and total viewing duration. The results suggest that corpora representative for an individual’s long-term memory structure can better explain reading performance than a norm corpus, and that recently acquired information is lexically accessed rapidly.

Multi-task learning (MTL) and transfer learning (TL) are techniques to overcome the issue of data scarcity when training state-of-the-art neural networks. However, finding beneficial auxiliary datasets for MTL or TL is a time- and resource-consuming trial-and-error approach. We propose new methods to automatically assess the similarity of sequence tagging datasets to identify beneficial auxiliary data for MTL or TL setups. Our methods can compute the similarity between any two sequence tagging datasets, they do not need to be annotated with the same tagset or multiple labels in parallel. Additionally, our methods take tokens and their labels into account, which is more robust than only using either of them as an information source, as conducted in prior work. We empirically show that our similarity measures correlate with the change in test score of neural networks that use the auxiliary dataset for MTL to increase the main task performance. We provide an efficient, open-source implementation.

Semantic frames are formal linguistic structures describing situations/actions/events, e.g. Commercial transfer of goods. Each frame provides a set of roles corresponding to the situation participants, e.g. Buyer and Goods, and lexical units (LUs) – words and phrases that can evoke this particular frame in texts, e.g. Sell. The scarcity of annotated resources hinders wider adoption of frame semantics across languages and domains. We investigate a simple yet effective method, lexical substitution with word representation models, to automatically expand a small set of frame-annotated sentences with new words for their respective roles and LUs. We evaluate the expansion quality using FrameNet. Contextualized models demonstrate overall superior performance compared to the non-contextualized ones on roles. However, the latter show comparable performance on the task of LU expansion.

We present path2vec, a new approach for learning graph embeddings that relies on structural measures of pairwise node similarities. The model learns representations for nodes in a dense space that approximate a given user-defined graph distance measure, such as e.g. the shortest path distance or distance measures that take information beyond the graph structure into account. Evaluation of the proposed model on semantic similarity and word sense disambiguation tasks, using various WordNet-based similarity measures, show that our approach yields competitive results, outperforming strong graph embedding baselines. The model is computationally efficient, being orders of magnitude faster than the direct computation of graph-based distances.

We present our system for semantic frame induction that showed the best performance in Subtask B.1 and finished as the runner-up in Subtask A of the SemEval 2019 Task 2 on unsupervised semantic frame induction (Qasem-iZadeh et al., 2019). Our approach separates this task into two independent steps: verb clustering using word and their context embeddings and role labeling by combining these embeddings with syntactical features. A simple combination of these steps shows very competitive results and can be extended to process other datasets and languages.

We present a neural network based approach of transfer learning for offensive language detection. For our system, we compare two types of knowledge transfer: supervised and unsupervised pre-training. Supervised pre-training of our bidirectional GRU-3-CNN architecture is performed as multi-task learning of parallel training of five different tasks. The selected tasks are supervised classification problems from public NLP resources with some overlap to offensive language such as sentiment detection, emoji classification, and aggressive language classification. Unsupervised transfer learning is performed with a thematic clustering of 40M unlabeled tweets via LDA. Based on this dataset, pre-training is performed by predicting the main topic of a tweet. Results indicate that unsupervised transfer from large datasets performs slightly better than supervised training on small ‘near target category’ datasets. In the SemEval Task, our system ranks 14 out of 103 participants.

The compositionality degree of multiword expressions indicates to what extent the meaning of a phrase can be derived from the meaning of its constituents and their grammatical relations. Prediction of (non)-compositionality is a task that has been frequently addressed with distributional semantic models. We introduce a novel technique to blend hierarchical information with distributional information for predicting compositionality. In particular, we use hypernymy information of the multiword and its constituents encoded in the form of the recently introduced Poincaré embeddings in addition to the distributional information to detect compositionality for noun phrases. Using a weighted average of the distributional similarity and a Poincaré similarity function, we obtain consistent and substantial, statistically significant improvement across three gold standard datasets over state-of-the-art models based on distributional information only. Unlike traditional approaches that solely use an unsupervised setting, we have also framed the problem as a supervised task, obtaining comparable improvements. Further, we publicly release our Poincaré embeddings, which are trained on the output of handcrafted lexical-syntactic patterns on a large corpus.

Graph measures, such as node distances, are inefficient to compute. We explore dense vector representations as an effective way to approximate the same information. We introduce a simple yet efficient and effective approach for learning graph embeddings. Instead of directly operating on the graph structure, our method takes structural measures of pairwise node similarities into account and learns dense node representations reflecting user-defined graph distance measures, such as e.g. the shortest path distance or distance measures that take information beyond the graph structure into account. We demonstrate a speed-up of several orders of magnitude when predicting word similarity by vector operations on our embeddings as opposed to directly computing the respective path-based measures, while outperforming various other graph embeddings on semantic similarity and word sense disambiguation tasks.

We introduce the use of Poincaré embeddings to improve existing state-of-the-art approaches to domain-specific taxonomy induction from text as a signal for both relocating wrong hyponym terms within a (pre-induced) taxonomy as well as for attaching disconnected terms in a taxonomy. This method substantially improves previous state-of-the-art results on the SemEval-2016 Task 13 on taxonomy extraction. We demonstrate the superiority of Poincaré embeddings over distributional semantic representations, supporting the hypothesis that they can better capture hierarchical lexical-semantic relationships than embeddings in the Euclidean space.

De-identification is the task of detecting protected health information (PHI) in medical text. It is a critical step in sanitizing electronic health records (EHR) to be shared for research. Automatic de-identification classifiers can significantly speed up the sanitization process. However, obtaining a large and diverse dataset to train such a classifier that works well across many types of medical text poses a challenge as privacy laws prohibit the sharing of raw medical records. We introduce a method to create privacy-preserving shareable representations of medical text (i.e. they contain no PHI) that does not require expensive manual pseudonymization. These representations can be shared between organizations to create unified datasets for training de-identification models. Our representation allows training a simple LSTM-CRF de-identification model to an F1 score of 97.4%, which is comparable to a strong baseline that exposes private information in its representation. A robust, widely available de-identification classifier based on our representation could potentially enable studies for which de-identification would otherwise be too costly.

Entity Disambiguation (ED) is the task of linking an ambiguous entity mention to a corresponding entry in a knowledge base. Current methods have mostly focused on unstructured text data to learn representations of entities, however, there is structured information in the knowledge base itself that should be useful to disambiguate entities. In this work, we propose a method that uses graph embeddings for integrating structured information from the knowledge base with unstructured information from text-based representations. Our experiments confirm that graph embeddings trained on a graph of hyperlinks between Wikipedia articles improve the performances of simple feed-forward neural ED model and a state-of-the-art neural ED system.

Capsule networks have been shown to demonstrate good performance on structured data in the area of visual inference. In this paper we apply and compare simple shallow capsule networks for hierarchical multi-label text classification and show that they can perform superior to other neural networks, such as CNNs and LSTMs, and non-neural network architectures such as SVMs. For our experiments, we use the established Web of Science (WOS) dataset and introduce a new real-world scenario dataset, the BlurbGenreCollection (BGC). Our results confirm the hypothesis that capsule networks are especially advantageous for rare events and structurally diverse categories, which we attribute to their ability to combine latent encoded information.

We present TARGER, an open source neural argument mining framework for tagging arguments in free input texts and for keyword-based retrieval of arguments from an argument-tagged web-scale corpus. The currently available models are pre-trained on three recent argument mining datasets and enable the use of neural argument mining without any reproducibility effort on the user’s side. The open source code ensures portability to other domains and use cases.

Expert finding is the task of ranking persons for a predefined topic or search query. Finding experts for a specified area is an important task and has attracted much attention in the information retrieval community. Most approaches for this task are evaluated in a supervised fashion, which depend on predefined topics of interest as well as gold standard expert rankings. Famous representatives of such datasets are enriched versions of DBLP provided by the ArnetMiner projet or the W3C Corpus of TREC. However, manually ranking experts can be considered highly subjective and detailed rankings are hardly distinguishable. Evaluating these datasets does not necessarily guarantee a good or bad performance of the system. Particularly for dynamic systems, where topics are not predefined but formulated as a search query, we believe a more informative approach is to perform user studies for directly comparing different methods in the same view. In order to accomplish this in a user-friendly way, we present the LT Expert Finder web-application, which is equipped with various query-based expert finding methods that can be easily extended, a detailed expert profile view, detailed evidence in form of relevant documents and statistics, and an evaluation component that allows the qualitative comparison between different rankings.

We present a detailed theoretical and computational analysis of the Watset meta-algorithm for fuzzy graph clustering, which has been found to be widely applicable in a variety of domains. This algorithm creates an intermediate representation of the input graph, which reflects the “ambiguity” of its nodes. Then, it uses hard clustering to discover clusters in this “disambiguated” intermediate graph. After outlining the approach and analyzing its computational complexity, we demonstrate that Watset shows competitive results in three applications: unsupervised synset induction from a synonymy graph, unsupervised semantic frame induction from dependency triples, and unsupervised semantic class induction from a distributional thesaurus. Our algorithm is generic and can also be applied to other networks of linguistic data.

In this paper, we propose a language-agnostic deep neural network architecture for aspect-based sentiment analysis. The proposed approach is based on Bidirectional Long Short-Term Memory (Bi-LSTM) network, which is further assisted with extra hand-crafted features. We define three different architectures for the successful combination of word embeddings and hand-crafted features. We evaluate the proposed approach for six languages (i.e. English, Spanish, French, Dutch, German and Hindi) and two problems (i.e. aspect term extraction and aspect sentiment classification). Experiments show that the proposed model attains state-of-the-art performance in most of the settings.

Implicit motives allow for the characterization of behavior, subsequent success and long-term development. While this has been operationalized in the operant motive test, research on motives has declined mainly due to labor-intensive and costly human annotation. In this study, we analyze over 200,000 labeled data items from 40,000 participants and utilize them for engineering features for training a logistic model tree machine learning model. It captures manually assigned motives well with an F-score of 80%, coming close to the pairwise annotator intraclass correlation coefficient of r = .85. In addition, we found a significant correlation of r = .2 between subsequent academic success and data automatically labeled with our model in an extrinsic evaluation.

We tackle the tasks of automatically identifying comparative sentences and categorizing the intended preference (e.g., “Python has better NLP libraries than MATLAB” → Python, better, MATLAB). To this end, we manually annotate 7,199 sentences for 217 distinct target item pairs from several domains (27% of the sentences contain an oriented comparison in the sense of “better” or “worse”). A gradient boosting model based on pre-trained sentence embeddings reaches an F1 score of 85% in our experimental evaluation. The model can be used to extract comparative sentences for pro/con argumentation in comparative / argument search engines or debating technologies.

Job boards and professional social networks heavily use recommender systems in order to better support users in exploring job advertisements. Detecting the similarity between job advertisements is important for job recommendation systems as it allows, for example, the application of item-to-item based recommendations. In this work, we research the usage of dense vector representations to enhance a large-scale job recommendation system and to rank German job advertisements regarding their similarity. We follow a two-folded evaluation scheme: (1) we exploit historic user interactions to automatically create a dataset of similar jobs that enables an offline evaluation. (2) In addition, we conduct an online A/B test and evaluate the best performing method on our platform reaching more than 1 million users. We achieve the best results by combining job titles with full-text job descriptions. In particular, this method builds dense document representation using words of the titles to weigh the importance of words of the full-text description. In the online evaluation, this approach allows us to increase the click-through rate on job recommendations for active users by 8.0%.

We use dependency triples automatically extracted from a Web-scale corpus to perform unsupervised semantic frame induction. We cast the frame induction problem as a triclustering problem that is a generalization of clustering for triadic data. Our replicable benchmarks demonstrate that the proposed graph-based approach, Triframes, shows state-of-the art results on this task on a FrameNet-derived dataset and performing on par with competitive methods on a verb class clustering task.

Learning from a real-world data stream and continuously updating the model without explicit supervision is a new challenge for NLP applications with machine learning components. In this work, we have developed an adaptive learning system for text simplification, which improves the underlying learning-to-rank model from usage data, i.e. how users have employed the system for the task of simplification. Our experimental result shows that, over a period of time, the performance of the embedded paraphrase ranking model increases steadily improving from a score of 62.88% up to 75.70% based on the NDCG@10 evaluation metrics. To our knowledge, this is the first study where an NLP component is adaptively improved through usage.

Depending on downstream applications, it is advisable to extend the notion of tokenization from low-level character-based token boundary detection to identification of meaningful and useful language units. This entails both identifying units composed of several single words that form a several single words that form a, as well as splitting single-word compounds into their meaningful parts. In this article, we introduce unsupervised and knowledge-free methods for these two tasks. The main novelty of our research is based on the fact that methods are primarily based on distributional similarity, of which we use two flavors: a sparse count-based and a dense neural-based distributional semantic model. First, we introduce DRUID, which is a method for detecting MWEs. The evaluation on MWE-annotated data sets in two languages and newly extracted evaluation data sets for 32 languages shows that DRUID compares favorably over previous methods not utilizing distributional information. Second, we present SECOS, an algorithm for decompounding close compounds. In an evaluation of four dedicated decompounding data sets across four languages and on data sets extracted from Wiktionary for 14 languages, we demonstrate the superiority of our approach over unsupervised baselines, sometimes even matching the performance of previous language-specific and supervised methods. In a final experiment, we show how both decompounding and MWE information can be used in information retrieval. Here, we obtain the best results when combining word information with MWEs and the compound parts in a bag-of-words retrieval set-up. Overall, our methodology paves the way to automatic detection of lexical units beyond standard tokenization techniques without language-specific preprocessing steps such as POS tagging.

We report the findings of the second Complex Word Identification (CWI) shared task organized as part of the BEA workshop co-located with NAACL-HLT’2018. The second CWI shared task featured multilingual and multi-genre datasets divided into four tracks: English monolingual, German monolingual, Spanish monolingual, and a multilingual track with a French test set, and two tasks: binary classification and probabilistic classification. A total of 12 teams submitted their results in different task/track combinations and 11 of them wrote system description papers that are referred to in this report and appear in the BEA workshop proceedings.

In this paper, we present Par4Sem, a semantic writing aid tool based on adaptive paraphrasing. Unlike many annotation tools that are primarily used to collect training examples, Par4Sem is integrated into a real word application, in this case a writing aid tool, in order to collect training examples from usage data. Par4Sem is a tool, which supports an adaptive, iterative, and interactive process where the underlying machine learning models are updated for each iteration using new training examples from usage data. After motivating the use of ever-learning tools in NLP applications, we evaluate Par4Sem by adopting it to a text simplification task through mere usage.

We introduce an advanced information extraction pipeline to automatically process very large collections of unstructured textual data for the purpose of investigative journalism. The pipeline serves as a new input processor for the upcoming major release of our New/s/leak 2.0 software, which we develop in cooperation with a large German news organization. The use case is that journalists receive a large collection of files up to several Gigabytes containing unknown contents. Collections may originate either from official disclosures of documents, e.g. Freedom of Information Act requests, or unofficial data leaks.

This paper describes BomJi, a supervised system for capturing discriminative attributes in word pairs (e.g. yellow as discriminative for banana over watermelon). The system relies on an XGB classifier trained on carefully engineered graph-, pattern- and word embedding-based features. It participated in the SemEval-2018 Task 10 on Capturing Discriminative Attributes, achieving an F1 score of 0.73 and ranking 2nd out of 26 participant systems.

In this paper we present the system for Answer Selection and Ranking in Community Question Answering, which we build as part of our participation in SemEval-2017 Task 3. We develop a Support Vector Machine (SVM) based system that makes use of textual, domain-specific, word-embedding and topic-modeling features. In addition, we propose a novel method for dialogue chain identification in comment threads. Our primary submission won subtask C, outperforming other systems in all the primary evaluation metrics. We performed well in other English subtasks, ranking third in subtask A and eighth in subtask B. We also developed open source toolkits for all the three English subtasks by the name cQARank [https://github.com/TitasNandi/cQARank].

This paper reports the STS-UHH participation in the SemEval 2017 shared Task 1 of Semantic Textual Similarity (STS). Overall, we submitted 3 runs covering monolingual and cross-lingual STS tracks. Our participation involves two approaches: unsupervised approach, which estimates a word alignment-based similarity score, and supervised approach, which combines dependency graph similarity and coverage features with lexical similarity measures using regression methods. We also present a way on ensembling both models. Out of 84 submitted runs, our team best multi-lingual run has been ranked 12th in overall performance with correlation of 0.61, 7th among 31 participating teams.

This paper reports team IITPB’s participation in the SemEval 2017 Task 5 on ‘Fine-grained sentiment analysis on financial microblogs and news’. We developed 2 systems for the two tracks. One system was based on an ensemble of Support Vector Classifier and Logistic Regression. This system relied on Distributional Thesaurus (DT), word embeddings and lexicon features to predict a floating sentiment value between -1 and +1. The other system was based on Support Vector Regression using word embeddings, lexicon features, and PMI scores as features. The system was ranked 5th in track 1 and 8th in track 2.

Complex word identification (CWI) is an important task in text accessibility. However, due to the scarcity of CWI datasets, previous studies have only addressed this problem on Wikipedia sentences and have solely taken into account the needs of non-native English speakers. We collect a new CWI dataset (CWIG3G2) covering three text genres News, WikiNews, and Wikipedia) annotated by both native and non-native English speakers. Unlike previous datasets, we cover single words, as well as complex phrases, and present them for judgment in a paragraph context. We present the first study on cross-genre and cross-group CWI, showing measurable influences in native language and genre types.

Complex Word Identification (CWI) is an important task in lexical simplification and text accessibility. Due to the lack of CWI datasets, previous works largely depend on Simple English Wikipedia and edit histories for obtaining ‘gold standard’ annotations, which are of doubtable quality, and limited only to English. We collect complex words/phrases (CP) for English, German and Spanish, annotated by both native and non-native speakers, and propose language independent features that can be used to train multilingual and cross-lingual CWI models. We show that the performance of cross-lingual CWI systems (using a model trained on one language and applying it on the other languages) is comparable to the performance of monolingual CWI systems.

Interpretability of a predictive model is a powerful feature that gains the trust of users in the correctness of the predictions. In word sense disambiguation (WSD), knowledge-based systems tend to be much more interpretable than knowledge-free counterparts as they rely on the wealth of manually-encoded elements representing word senses, such as hypernyms, usage examples, and images. We present a WSD system that bridges the gap between these two so far disconnected groups of methods. Namely, our system, providing access to several state-of-the-art WSD models, aims to be interpretable as a knowledge-based system while it remains completely unsupervised and knowledge-free. The presented tool features a Web interface for all-word disambiguation of texts that makes the sense predictions human readable by providing interpretable word sense inventories, sense representations, and disambiguation results. We provide a public API, enabling seamless integration.

The current trend in NLP is the use of highly opaque models, e.g. neural networks and word embeddings. While these models yield state-of-the-art results on a range of tasks, their drawback is poor interpretability. On the example of word sense induction and disambiguation (WSID), we show that it is possible to develop an interpretable model that matches the state-of-the-art models in accuracy. Namely, we present an unsupervised, knowledge-free WSID approach, which is interpretable at three levels: word sense inventory, sense feature representations, and disambiguation procedure. Experiments show that our model performs on par with state-of-the-art word sense embeddings and other unsupervised systems while offering the possibility to justify its decisions in human-readable form.

In this paper, we present ContrastMedium, an algorithm that transforms noisy semantic networks into full-fledged, clean taxonomies. ContrastMedium is able to identify the embedded taxonomy structure from a noisy knowledge graph without explicit human supervision such as, for instance, a set of manually selected input root and leaf concepts. This is achieved by leveraging structural information from a companion reference taxonomy, to which the input knowledge graph is linked (either automatically or manually). When used in conjunction with methods for hypernym acquisition and knowledge base linking, our methodology provides a complete solution for end-to-end taxonomy induction. We conduct experiments using automatically acquired knowledge graphs, as well as a SemEval benchmark, and show that our method is able to achieve high performance on the task of taxonomy induction.

We present a new approach to extraction of hypernyms based on projection learning and word embeddings. In contrast to classification-based approaches, projection-based methods require no candidate hyponym-hypernym pairs. While it is natural to use both positive and negative training examples in supervised relation extraction, the impact of positive examples on hypernym prediction was not studied so far. In this paper, we show that explicit negative examples used for regularization of the model significantly improve performance compared to the state-of-the-art approach of Fu et al. (2014) on three datasets from different languages.

We introduce a new method for unsupervised knowledge-based word sense disambiguation (WSD) based on a resource that links two types of sense-aware lexical networks: one is induced from a corpus using distributional semantics, the other is manually constructed. The combination of two networks reduces the sparsity of sense representations used for WSD. We evaluate these enriched representations within two lexical sample sense disambiguation benchmarks. Our results indicate that (1) features extracted from the corpus-based resource help to significantly outperform a model based solely on the lexical resource; (2) our method achieves results comparable or better to four state-of-the-art unsupervised knowledge-based WSD systems including three hybrid systems that also rely on text corpora. In contrast to these hybrid methods, our approach does not require access to web search engines, texts mapped to a sense inventory, or machine translation systems.

In this paper, we describe the concept of entity-centric information access for the biomedical domain. With entity recognition technologies approaching acceptable levels of accuracy, we put forward a paradigm of document browsing and searching where the entities of the domain and their relations are explicitly modeled to provide users the possibility of collecting exhaustive information on relations of interest. We describe three working prototypes along these lines: NEW/S/LEAK, which was developed for investigative journalists who need a quick overview of large leaked document collections; STORYFINDER, which is a personalized organizer for information found in web pages that allows adding entities as well as relations, and is capable of personalized information management; and adaptive annotation capabilities of WEBANNO, which is a general-purpose linguistic annotation tool. We will discuss future steps towards the adaptation of these tools to biomedical data, which is subject to a recently started project on biomedical knowledge acquisition. A key difference to other approaches is the centering around the user in a Human-in-the-Loop machine learning approach, where users define and extend categories and enable the system to improve via feedback and interaction.

This paper presents a new graph-based approach that induces synsets using synonymy dictionaries and word embeddings. First, we build a weighted graph of synonyms extracted from commonly available resources, such as Wiktionary. Second, we apply word sense induction to deal with ambiguous words. Finally, we cluster the disambiguated version of the ambiguous input graph into synsets. Our meta-clustering approach lets us use an efficient hard clustering algorithm to perform a fuzzy clustering of the graph. Despite its simplicity, our approach shows excellent results, outperforming five competitive state-of-the-art methods in terms of F-score on three gold standard datasets for English and Russian derived from large-scale manually constructed lexical resources.

We introduce the third major release of WebAnno, a generic web-based annotation tool for distributed teams. New features in this release focus on semantic annotation tasks (e.g. semantic role labelling or event annotation) and allow the tight integration of semantic annotations with syntactic annotations. In particular, we introduce the concept of slot features, a novel constraint mechanism that allows modelling the interaction between semantic and syntactic annotations, as well as a new annotation user interface. The new features were developed and used in an annotation project for semantic roles on German texts. The paper briefly introduces this project and reports on experiences performing annotations with the new tool. On a comparative evaluation, our tool reaches significant speedups over WebAnno 2 for a semantic annotation task.

Distributional Semantic Models (DSMs) have recently received increased attention, together with the rise of neural architectures for scalable training of dense vector embeddings. While some of the literature even includes terms like ‘vectors’ and ‘dimensionality’ in the definition of DSMs, there are some good reasons why we should consider alternative formulations of distributional models. As an instance, I present a scalable graph-based solution to distributional semantics. The model belongs to the family of ‘count-based’ DSMs, keeps its representation sparse and explicit, and thus fully interpretable. I will highlight some important differences between sparse graph-based and dense vector approaches to DSMs: while dense vector-based models are computationally easier to handle and provide a nice uniform representation that can be compared and combined in many ways, they lack interpretability, provenance and robustness. On the other hand, graph-based sparse models have a more straightforward interpretation, handle sense distinctions more naturally and can straightforwardly be linked to knowledge bases, while lacking the ability to compare arbitrary lexical units and a compositionality operation. Since both representations have their merits, I opt for exploring their combination in the outlook.

Language production is largely a matter of words which, in the case of access problems, can be searched for in an external resource (lexicon, thesaurus). In this kind of dialogue the user provides the momentarily available knowledge concerning the target and the system responds with the best guess(es) it can make given this input. As tip-of-the-tongue (ToT)-studies have shown, people always have some knowledge concerning the target (meaning fragments, number of syllables, ...) even if its complete form is eluding them. We will show here how to tap on this knowledge to build a resource likely to help authors (speakers/writers) to overcome the ToT-problem. Yet, before doing so we need a better understanding of the various kinds of knowledge people have when looking for a word. To this end, we asked crowdworkers to provide some cues to describe a given target and to specify then how each one of them relates to the target, in the hope that this could help others to find the elusive word. Next, we checked how well a given search strategy worked when being applied to differently built lexical networks. The results showed quite dramatic differences, which is not really surprising. After all, different networks are built for different purposes; hence each one of them is more or less suited for a given task. What was more surprising though is the fact that the relational information given by the users did not allow us to find the elusive word in WordNet better than without it.

This work presents a straightforward method for extending or creating in-domain web corpora by focused webcrawling. The focused webcrawler uses statistical N-gram language models to estimate the relatedness of documents and weblinks and needs as input only N-grams or plain texts of a predefined domain and seed URLs as starting points. Two experiments demonstrate that our focused crawler is able to stay focused in domain and language. The first experiment shows that the crawler stays in a focused domain, the second experiment demonstrates that language models trained on focused crawls obtain better perplexity scores on in-domain corpora. We distribute the focused crawler as open source software.

Semantic relations play an important role in linguistic knowledge representation. Although their role is relevant in the context of written text, there is no approach or dataset that makes use of contextuality of classic semantic relations beyond the boundary of one sentence. We present the SemRelData dataset that contains annotations of semantic relations between nominals in the context of one paragraph. To be able to analyse the universality of this context notion, the annotation was performed on a multi-lingual and multi-genre corpus. To evaluate the dataset, it is compared to large, manually created knowledge resources in the respective languages. The comparison shows that knowledge bases not only have coverage gaps; they also do not account for semantic relations that are manifested in particular contexts only, yet still play an important role for text cohesion.

We present Ambient Search, an open source system for displaying and retrieving relevant documents in real time for speech input. The system works ambiently, that is, it unobstructively listens to speech streams in the background, identifies keywords and keyphrases for query construction and continuously serves relevant documents from its index. Query terms are ranked with Word2Vec and TF-IDF and are continuously updated to allow for ongoing querying of a document collection. The retrieved documents, in our case Wikipedia articles, are visualized in real time in a browser interface. Our evaluation shows that Ambient Search compares favorably to another implicit information retrieval system on speech streams. Furthermore, we extrinsically evaluate multiword keyphrase generation, showing positive impact for manual transcriptions.

In this demonstration paper we describe Ambient Search, a system that displays and retrieves documents in real time based on speech input. The system operates continuously in ambient mode, i.e. it generates speech transcriptions and identifies main keywords and keyphrases, while also querying its index to display relevant documents without explicit query. Without user intervention, the results are dynamically updated; users can choose to interact with the system at any time, employing a conversation protocol that is enriched with the ambient information gathered continuously. Our evaluation shows that Ambient Search outperforms another implicit speech-based information retrieval system. Ambient search is available as open source software.

This paper introduces a distributional thesaurus and sense clusters computed on the complete Google Syntactic N-grams, which is extracted from Google Books, a very large corpus of digitized books published between 1520 and 2008. We show that a thesaurus computed on such a large text basis leads to much better results than using smaller corpora like Wikipedia. We also provide distributional thesauri for equal-sized time slices of the corpus. While distributional thesauri can be used as lexical resources in NLP tasks, comparing word similarities over time can unveil sense change of terms across different decades or centuries, and can serve as a resource for diachronic lexicography. Thesauri and clusters are available for download.

We describe the annotation of a new dataset for German Named Entity Recognition (NER). The need for this dataset is motivated by licensing issues and consistency issues of existing datasets. We describe our approach to creating annotation guidelines based on linguistic and semantic considerations, and how we iteratively refined and tested them in the early stages of annotation in order to arrive at the largest publicly available dataset for German NER, consisting of over 31,000 manually annotated sentences (over 591,000 tokens) from German Wikipedia and German online news. We provide a number of statistics on the dataset, which indicate its high quality, and discuss legal aspects of distributing the data as a compilation of citations. The data is released under the permissive CC-BY license, and will be fully available for download in September 2014 after it has been used for the GermEval 2014 shared task on NER. We further provide the full annotation guidelines and links to the annotation tool used for the creation of this resource.

This article describes a lexical substitution dataset for German. The whole dataset contains 2,040 sentences from the German Wikipedia, with one target word in each sentence. There are 51 target nouns, 51 adjectives, and 51 verbs randomly selected from 3 frequency groups based on the lemma frequency list of the German WaCKy corpus. 200 sentences have been annotated by 4 professional annotators and the remaining sentences by 1 professional annotator and 5 additional annotators who have been recruited via crowdsourcing. The resulting dataset can be used to evaluate not only lexical substitution systems, but also different sense inventories and word sense disambiguation systems.

This paper presents the Turk Bootstrap Word Sense Inventory (TWSI) 2.0. This lexical resource, created by a crowdsourcing process using Amazon Mechanical Turk (http://www.mturk.com), encompasses a sense inventory for lexical substitution for 1,012 highly frequent English common nouns. Along with each sense, a large number of sense-annotated occurrences in context are given, as well as a weighted list of substitutions. Sense distinctions are not motivated by lexicographic considerations, but driven by substitutability: two usages belong to the same sense if their substitutions overlap considerably. After laying out the need for such a resource, the data is characterized in terms of organization and quantity. Then, we briefly describe how this data was used to create a system for lexical substitutions. Training a supervised lexical substitution system on a smaller version of the resource resulted in well over 90% acceptability for lexical substitutions provided by the system. Thus, this resource can be used to set up reliable, enabling technologies for semantic natural language processing (NLP), some of which we discuss briefly.

We present a study of the word interaction networks of Bengali in the framework of complex networks. The topological properties of these networks reveal interesting insights into the morpho-syntax of the language, whereas clustering helps in the induction of the natural word classes leading to a principled way of designing POS tagsets. We compare different network construction techniques and clustering algorithms based on the cohesiveness of the word clusters. Cohesiveness is measured against two gold-standard tagsets by means of the novel metric of tag-entropy. The approach presented here is a generic one that can be easily extended to any language.

ASV Toolbox is a modular collection of tools for the exploration of written language data both for scientific and educational purposes. It includes modules that operate on word lists or texts and allow to perform various linguistic annotation, classification and clustering tasks, including language detection, POS-tagging, base form reduction, named entity recognition, and terminology extraction. On a more abstract level, the algorithms deal with various kinds of word similarity, using pattern-based and statistical approaches. The collection can be used to work on large real-world data sets as well as for studying the underlying algorithms. Each module of the ASV Toolbox is designed to work either on a plain text files or with a connection to a MySQL database. While it is especially designed to work with corpora of the Leipzig Corpora Collection, it can easily be adapted to other sources.

A simple and flexible schema for storing and presenting monolingual language resources is proposed. In this format, data for 18 different languages is already available in various sizes. The data is provided free of charge for online use and download. The main target is to ease the application of algorithms for monolingual and interlingual studies.