Generative AI systems have become ubiquitous for all kinds of modalities, which makes the issue of the evaluation of such models more pressing. One popular approach is preference ratings, where the generated outputs of different systems are shown to evaluators who choose their preferences. In recent years the field shifted towards the development of automated (trained) metrics to assess generated outputs, which can be used to create preference ratings automatically. In this work, we investigate the evaluation of the metrics themselves, which currently rely on measuring the correlation to human judgments or computing sign accuracy scores. These measures only assess how well the metric agrees with the human ratings. However, our research shows that this does not tell the whole story. Most metrics exhibit a disagreement with human system assessments which is often skewed in favor of particular text generation systems, exposing a degree of favoritism in automated metrics. This paper introduces a formal definition of favoritism in preference metrics, and derives the Favi-Score, which measures this phenomenon. In particular we show that favoritism is strongly related to errors in final system rankings. Thus, we propose that preference-based metrics ought to be evaluated on both sign accuracy scores and favoritism.

One of the central skills that language learners need to practice is speaking the language. Currently, students in school do not get enough speaking opportunities and lack conversational practice. The recent advances in speech technology and natural language processing allow the creation of novel tools to practice their speaking skills. In this work, we tackle the first component of such a pipeline, namely, the automated speech recognition module (ASR). State-of-the-art models are often trained on adult read-aloud data by native speakers and do not transfer well to young language learners’ speech. Second, most ASR systems contain a powerful language model, which smooths out mistakes made by the speakers. To give corrective feedback, which is a crucial part of language learning, the ASR systems in our setting need to preserve the mistakes made by the language learners. In this work, we build an ASR system that satisfies these requirements: it works on spontaneous speech by young language learners and preserves their mistakes. For this, we collected a corpus containing around 85 hours of English audio spoken by Swiss learners from grades 4 to 6 on different language learning tasks, which we used to train an ASR model. Our experiments show that our model benefits from direct fine-tuning of children’s voices and has a much higher error preservation rate.

We present STT4SG-350, a corpus of Swiss German speech, annotated with Standard German text at the sentence level. The data is collected using a web app in which the speakers are shown Standard German sentences, which they translate to Swiss German and record. We make the corpus publicly available. It contains 343 hours of speech from all dialect regions and is the largest public speech corpus for Swiss German to date. Application areas include automatic speech recognition (ASR), text-to-speech, dialect identification, and speaker recognition. Dialect information, age group, and gender of the 316 speakers are provided. Genders are equally represented and the corpus includes speakers of all ages. Roughly the same amount of speech is provided per dialect region, which makes the corpus ideally suited for experiments with speech technology for different dialects. We provide training, validation, and test splits of the data. The test set consists of the same spoken sentences for each dialect region and allows a fair evaluation of the quality of speech technologies in different dialects. We train an ASR model on the training set and achieve an average BLEU score of 74.7 on the test set. The model beats the best published BLEU scores on 2 other Swiss German ASR test sets, demonstrating the quality of the corpus.

A major challenge in the field of Text Generation is evaluation: Human evaluations are cost-intensive, and automated metrics often display considerable disagreements with human judgments. In this paper, we propose to apply automated metrics for Text Generation in a preference-based evaluation protocol. The protocol features a statistical model that incorporates various levels of uncertainty to account for the error-proneness of the metrics. We show that existing metrics are generally over-confident in assigning significant differences between systems. As a remedy, the model allows to combine human ratings with automated ratings. We show that it can reduce the required amounts of human ratings to arrive at robust and statistically significant results by more than 50%, while yielding the same evaluation outcome as the pure human evaluation in 95% of cases. We showcase the benefits of the evaluation protocol for three text generation tasks: dialogue systems, machine translation, and text summarization.

This paper investigates the challenges in building Swiss German speech translation systems, specifically focusing on the impact of dialect diversity and differences between Swiss German and Standard German. Swiss German is a spoken language with no formal writing system, it comprises many diverse dialects and is a low-resource language with only around 5 million speakers. The study is guided by two key research questions: how does the inclusion and exclusion of dialects during the training of speech translation models for Swiss German impact the performance on specific dialects, and how do the differences between Swiss German and Standard German impact the performance of the systems? We show that dialect diversity and linguistic differences pose significant challenges to Swiss German speech translation, which is in line with linguistic hypotheses derived from empirical investigations.

This paper describes the reproduction of a human evaluation in Language-Agnostic Meta- Learning for Low-Resource Text-to-Speech with Articulatory Features reported in Lux and Vu (2022). It is a contribution to the ReproNLP 2023 Shared Task on Reproducibility of Evaluations in NLP. The original evaluation assessed the naturalness of audio generated by different Text-to-Speech (TTS) systems for German, and our goal was to repeat the experiment with a different set of evaluators. We reproduced the evaluation based on data and instructions provided by the original authors, with some uncertainty concerning the randomisation of question order. Evaluators were recruited via email to relevant mailing lists and we received 157 responses over the course of three weeks. Our initial results show low reproducibility, but when we assume that the systems of the original and repeat evaluation experiment have been transposed, the reproducibility assessment improves markedly. We do not know if and at what point such a transposition happened; however, an initial analysis of our audio and video files provides some evidence that the system assignment in our repeat experiment is correct.

We report our efforts in identifying a set of previous human evaluations in NLP that would be suitable for a coordinated study examining what makes human evaluations in NLP more/less reproducible. We present our results and findings, which include that just 13% of papers had (i) sufficiently low barriers to reproduction, and (ii) enough obtainable information, to be considered for reproduction, and that all but one of the experiments we selected for reproduction was discovered to have flaws that made the meaningfulness of conducting a reproduction questionable. As a result, we had to change our coordinated study design from a reproduce approach to a standardise-then-reproduce-twice approach. Our overall (negative) finding that the great majority of human evaluations in NLP is not repeatable and/or not reproducible and/or too flawed to justify reproduction, paints a dire picture, but presents an opportunity for a rethink about how to design and report human evaluations in NLP.

This paper introduces an adversarial method to stress-test trained metrics for the evaluation of conversational dialogue systems. The method leverages Reinforcement Learning to find response strategies that elicit optimal scores from the trained metrics. We apply our method to test recently proposed trained metrics. We find that they all are susceptible to giving high scores to responses generated by rather simple and obviously flawed strategies that our method converges on. For instance, simply copying parts of the conversation context to form a response yields competitive scores or even outperforms responses written by humans.

A major challenge in the field of Text Generation is evaluation, because we lack a sound theory that can be leveraged to extract guidelines for evaluation campaigns. In this work, we propose a first step towards such a theory that incorporates different sources of uncertainty, such as imperfect automated metrics and insufficiently sized test sets. The theory has practical applications, such as determining the number of samples needed to reliably distinguish the performance of a set of Text Generation systems in a given setting. We showcase the application of the theory on the WMT 21 and Spot-The-Bot evaluation data and outline how it can be leveraged to improve the evaluation protocol regarding the reliability, robustness, and significance of the evaluation outcome.

We present SDS-200, a corpus of Swiss German dialectal speech with Standard German text translations, annotated with dialect, age, and gender information of the speakers. The dataset allows for training speech translation, dialect recognition, and speech synthesis systems, among others. The data was collected using a web recording tool that is open to the public. Each participant was given a text in Standard German and asked to translate it to their Swiss German dialect before recording it. To increase the corpus quality, recordings were validated by other participants. The data consists of 200 hours of speech by around 4000 different speakers and covers a large part of the Swiss German dialect landscape. We release SDS-200 alongside a baseline speech translation model, which achieves a word error rate (WER) of 30.3 and a BLEU score of 53.1 on the SDS-200 test set. Furthermore, we use SDS-200 to fine-tune a pre-trained XLS-R model, achieving 21.6 WER and 64.0 BLEU.

Dialogue summarization is a long-standing task in the field of NLP, and several data sets with dialogues and associated human-written summaries of different styles exist. However, it is unclear for which type of dialogue which type of summary is most appropriate. For this reason, we apply a linguistic model of dialogue types to derive matching summary items and NLP tasks. This allows us to map existing dialogue summarization data sets into this model and identify gaps and potential directions for future work. As part of this process, we also provide an extensive overview of existing dialogue summarization data sets.

We describe our approaches for the Social Media Geolocation (SMG) task at the VarDial Evaluation Campaign 2020. The goal was to predict geographical location (latitudes and longitudes) given an input text. There were three subtasks corresponding to German-speaking Switzerland (CH), Germany and Austria (DE-AT), and Croatia, Bosnia and Herzegovina, Montenegro and Serbia (BCMS). We submitted solutions to all subtasks but focused our development efforts on the CH subtask, where we achieved third place out of 16 submissions with a median distance of 15.93 km and had the best result of 14 unconstrained systems. In the DE-AT subtask, we ranked sixth out of ten submissions (fourth of 8 unconstrained systems) and for BCMS we achieved fourth place out of 13 submissions (second of 11 unconstrained systems).

We present LEDGAR, a multilabel corpus of legal provisions in contracts. The corpus was crawled and scraped from the public domain (SEC filings) and is, to the best of our knowledge, the first freely available corpus of its kind. Since the corpus was constructed semi-automatically, we apply and discuss various approaches to noise removal. Due to the rather large labelset of over 12’000 labels annotated in almost 100’000 provisions in over 60’000 contracts, we believe the corpus to be of interest for research in the field of Legal NLP, (large-scale or extreme) text classification, as well as for legal studies. We discuss several methods to sample subcopora from the corpus and implement and evaluate different automatic classification approaches. Finally, we perform transfer experiments to evaluate how well the classifiers perform on contracts stemming from outside the corpus.

Transliteration is the process of expressing a proper name from a source language in the characters of a target language (e.g. from Cyrillic to Latin characters). We present TRANSLIT, a large-scale corpus with approx. 1.6 million entries in more than 180 languages with about 3 million variations of person and geolocation names. The corpus is based on various public data sources, which have been transformed into a unified format to simplify their usage, plus a newly compiled dataset from Wikipedia. In addition, we apply several machine learning methods to establish baselines for automatically detecting transliterated names in various languages. Our best systems achieve an accuracy of 92% on identification of transliterated pairs.

In this paper, we present CEASR, a Corpus for Evaluating the quality of Automatic Speech Recognition (ASR). It is a data set based on public speech corpora, containing metadata along with transcripts generated by several modern state-of-the-art ASR systems. CEASR provides this data in a unified structure, consistent across all corpora and systems, with normalised transcript texts and metadata. We use CEASR to evaluate the quality of ASR systems by calculating an average Word Error Rate (WER) per corpus, per system and per corpus-system pair. Our experiments show a substantial difference in accuracy between commercial versus open-source ASR tools as well as differences up to a factor ten for single systems on different corpora. Using CEASR allowed us to very efficiently and easily obtain these results. Our corpus enables researchers to perform ASR-related evaluations and various in-depth analyses with noticeably reduced effort, i.e. without the need to collect, process and transcribe the speech data themselves.

In this paper, we introduce a novel methodology to efficiently construct a corpus for question answering over structured data. For this, we introduce an intermediate representation that is based on the logical query plan in a database, called Operation Trees (OT). This representation allows us to invert the annotation process without loosing flexibility in the types of queries that we generate. Furthermore, it allows for fine-grained alignment of the tokens to the operations. Thus, we randomly generate OTs from a context free grammar and annotators just have to write the appropriate question and assign the tokens. We compare our corpus OTTA (Operation Trees and Token Assignment), a large semantic parsing corpus for evaluating natural language interfaces to databases, to Spider and LC-QuaD 2.0 and show that our methodology more than triples the annotation speed while maintaining the complexity of the queries. Finally, we train a state-of-the-art semantic parsing model on our data and show that our dataset is a challenging dataset and that the token alignment can be leveraged to significantly increase the performance.

The goal of this work is to build conversational Question Answering (QA) interfaces for the large body of domain-specific information available in FAQ sites. We present DoQA, a dataset with 2,437 dialogues and 10,917 QA pairs. The dialogues are collected from three Stack Exchange sites using the Wizard of Oz method with crowdsourcing. Compared to previous work, DoQA comprises well-defined information needs, leading to more coherent and natural conversations with less factoid questions and is multi-domain. In addition, we introduce a more realistic information retrieval (IR) scenario where the system needs to find the answer in any of the FAQ documents. The results of an existing, strong, system show that, thanks to transfer learning from a Wikipedia QA dataset and fine tuning on a single FAQ domain, it is possible to build high quality conversational QA systems for FAQs without in-domain training data. The good results carry over into the more challenging IR scenario. In both cases, there is still ample room for improvement, as indicated by the higher human upperbound.

The lack of time efficient and reliable evalu-ation methods is hampering the development of conversational dialogue systems (chat bots). Evaluations that require humans to converse with chat bots are time and cost intensive, put high cognitive demands on the human judges, and tend to yield low quality results. In this work, we introduce Spot The Bot, a cost-efficient and robust evaluation framework that replaces human-bot conversations with conversations between bots. Human judges then only annotate for each entity in a conversation whether they think it is human or not (assuming there are humans participants in these conversations). These annotations then allow us to rank chat bots regarding their ability to mimic conversational behaviour of humans. Since we expect that all bots are eventually recognized as such, we incorporate a metric that measures which chat bot is able to uphold human-like be-havior the longest, i.e.Survival Analysis. This metric has the ability to correlate a bot’s performance to certain of its characteristics (e.g.fluency or sensibleness), yielding interpretable results. The comparably low cost of our frame-work allows for frequent evaluations of chatbots during their evaluation cycle. We empirically validate our claims by applying Spot The Bot to three domains, evaluating several state-of-the-art chat bots, and drawing comparisonsto related work. The framework is released asa ready-to-use tool.

We report our ongoing work about a new deep architecture working in tandem with a statistical test procedure for jointly training texts and their label descriptions for multi-label and multi-class classification tasks. A statistical hypothesis testing method is used to extract the most informative words for each given class. These words are used as a class description for more label-aware text classification. Intuition is to help the model to concentrate on more informative words rather than more frequent ones. The model leverages the use of label descriptions in addition to the input text to enhance text classification performance. Our method is entirely data-driven, has no dependency on other sources of information than the training data, and is adaptable to different classification problems by providing appropriate training data without major hyper-parameter tuning. We trained and tested our system on several publicly available datasets, where we managed to improve the state-of-the-art on one set with a high margin and to obtain competitive results on all other ones.

We describe our approaches for the German Dialect Identification (GDI) and the Cuneiform Language Identification (CLI) tasks at the VarDial Evaluation Campaign 2019. The goal was to identify dialects of Swiss German in GDI and Sumerian and Akkadian in CLI. In GDI, the system should distinguish four dialects from the German-speaking part of Switzerland. Our system for GDI achieved third place out of 6 teams, with a macro averaged F-1 of 74.6%. In CLI, the system should distinguish seven languages written in cuneiform script. Our system achieved third place out of 8 teams, with a macro averaged F-1 of 74.7%.

We study how language on social media is linked to mortal diseases such as atherosclerotic heart disease (AHD), diabetes and various types of cancer. Our proposed model leverages state-of-the-art sentence embeddings, followed by a regression model and clustering, without the need of additional labelled data. It allows to predict community-level medical outcomes from language, and thereby potentially translate these to the individual level. The method is applicable to a wide range of target variables and allows us to discover known and potentially novel correlations of medical outcomes with life-style aspects and other socioeconomic risk factors.

We present “AutoJudge”, an automated evaluation method for conversational dialogue systems. The method works by first generating dialogues based on self-talk, i.e. dialogue systems talking to itself. Then, it uses human ratings on these dialogues to train an automated judgement model. Our experiments show that AutoJudge correlates well with the human ratings and can be used to automatically evaluate dialogue systems, even in deployed systems. In a second part, we attempt to apply AutoJudge to improve existing systems. This works well for re-ranking a set of candidate utterances. However, our experiments show that AutoJudge cannot be applied as reward for reinforcement learning, although the metric can distinguish good from bad dialogues. We discuss potential reasons, but state here already that this is still an open question for further research.

We describe our approaches used in the German Dialect Identification (GDI) task at the VarDial Evaluation Campaign 2018. The goal was to identify to which out of four dialects spoken in German speaking part of Switzerland a sentence belonged to. We adopted two different meta classifier approaches and used some data mining insights to improve the preprocessing and the meta classifier parameters. Especially, we focused on using different feature extraction methods and how to combine them, since they influenced very differently the performance of the system. Our system achieved second place out of 8 teams, with a macro averaged F-1 of 64.6%.

Natural Language Generation plays an important role in the domain of dialogue systems as it determines how users perceive the system. Recently, deep-learning based systems have been proposed to tackle this task, as they generalize better and require less amounts of manual effort to implement them for new domains. However, deep learning systems usually adapt a very homogeneous sounding writing style which expresses little variation. In this work, we present our system for Natural Language Generation where we control various aspects of the surface realization in order to increase the lexical variability of the utterances, such that they sound more diverse and interesting. For this, we use a Semantically Controlled Long Short-term Memory Network (SC-LSTM), and apply its specialized cell to control various syntactic features of the generated texts. We present an in-depth human evaluation where we show the effects of these surface manipulation on the perception of potential users.

In this paper we propose a system for reranking answers for a given question. Our method builds on a siamese CNN architecture which is extended by two attention mechanisms. The approach was evaluated on the datasets of the SemEval-2017 competition for Community Question Answering (cQA), where it achieved 7th place obtaining a MAP score of 86:24 points on the Question-Comment Similarity subtask.

In this paper, we propose a classifier for predicting topic-specific sentiments of English Twitter messages. Our method is based on a 2-layer CNN.With a distant supervised phase we leverage a large amount of weakly-labelled training data. Our system was evaluated on the data provided by the SemEval-2017 competition in the Topic-Based Message Polarity Classification subtask, where it ranked 4th place.

In this paper we investigate the cross-domain performance of a current state-of-the-art sentiment analysis systems. For this purpose we train a convolutional neural network (CNN) on data from different domains and evaluate its performance on other domains. Furthermore, we evaluate the usefulness of combining a large amount of different smaller annotated corpora to a large corpus. Our results show that more sophisticated approaches are required to train a system that works equally well on various domains.

In this paper we present SB10k, a new corpus for sentiment analysis with approx. 10,000 German tweets. We use this new corpus and two existing corpora to provide state-of-the-art benchmarks for sentiment analysis in German: we implemented a CNN (based on the winning system of SemEval-2016) and a feature-based SVM and compare their performance on all three corpora. For the CNN, we also created German word embeddings trained on 300M tweets. These word embeddings were then optimized for sentiment analysis using distant-supervised learning. The new corpus, the German word embeddings (plain and optimized), and source code to re-run the benchmarks are publicly available.

We present our system for the WNUT 2017 Named Entity Recognition challenge on Twitter data. We describe two modifications of a basic neural network architecture for sequence tagging. First, we show how we exploit additional labeled data, where the Named Entity tags differ from the target task. Then, we propose a way to incorporate sentence level features. Our system uses both methods and ranked second for entity level annotations, achieving an F1-score of 40.78, and second for surface form annotations, achieving an F1-score of 39.33.

In this paper, we analyze the quality of several commercial tools for sentiment detection. All tools are tested on nearly 30,000 short texts from various sources, such as tweets, news, reviews etc. The best commercial tools have average accuracy of 60%. We then apply machine learning techniques (Random Forests) to combine all tools, and show that this results in a meta-classifier that improves the overall performance significantly.