This paper describes our submission to theSemEval 2023 Task 3 on two subtasks: detectingpersuasion techniques and framing. Bothsubtasks are multi-label classification problems. We present a set of experiments, exploring howto get robust performance across languages usingpre-trained RoBERTa models. We test differentoversampling strategies, a strategy ofadding textual features from predictions obtainedwith related models, and present bothinconclusive and negative results. We achievea robust ranking across languages and subtaskswith our best ranking being nr. 1 for Subtask 3on Spanish.

In this paper, we describe our approach to Task 4 in SemEval 2023. Our pipeline tries to solve the problem of multi-label text classification of human values in English-written arguments. We propose a label-aware system where we reframe the multi-label task into a binary task resembling an NLI task. We propose to include the semantic description of the human values by comparing each description to each argument and ask whether there is entailment or not.

Few-shot classification is a powerful technique, but training requires substantial computing power and data. We propose an efficient method with small model sizes and less training data with only 2-8 training instances per class. Our proposed method, AncSetFit, targets low data scenarios by anchoring the task and label information through sentence embeddings in fine-tuning a Sentence Transformer model. It uses contrastive learning and a triplet loss to enforce training instances of a class to be closest to its own textual semantic label information in the embedding space - and thereby learning to embed different class instances more distinct. AncSetFit obtains strong performance in data-sparse scenarios compared to existing methods across SST-5, Emotion detection, and AG News data, even with just two examples per class.

It is important to understand how people use words to persuade each other. This helps understand debate, and detect persuasive narratives in regard to e.g. misinformation. While computational modelling of some aspects of persuasion has received some attention, a way to unify and describe the overall phenomenon of when persuasion becomes undesired and problematic, is missing. In this paper, we attempt to address this by proposing a taxonomy of computational persuasion. Drawing upon existing research and resources, this paper shows how to re-frame and re-organise current work into a coherent framework targeting the misuse of rhetorical appeals. As a study to validate these re-framings, we then train and evaluate models of persuasion adapted to our taxonomy. Our results show an application of our taxonomy, and we are able to detecting misuse of rhetorical appeals, finding that these are more often used in misinformative contexts than in true ones.

Language Identification is the task of identifying a document’s language. For applications like automatic spell checker selection, language identification must use very short strings such as text message fragments. In this work, we reproduce a language identification architecture that Apple briefly sketched in a blog post. We confirm the bi-LSTM model’s performance and find that it outperforms current open-source language identifiers. We further find that its language identification mistakes are due to confusion between related languages.

In this work we present a corpus for the evaluation of sensitive information detection approaches that addresses the need for real world sensitive information for empirical studies. Our sentence corpus contains different notions of complex sensitive information that correspond to different aspects of concern in a current trial of the Monsanto company. This paper describes the annotations process, where we both employ human annotators and furthermore create automatically inferred labels regarding technical, legal and informal communication within and with employees of Monsanto, drawing on a classification of documents by lawyers involved in the Monsanto court case. We release corpus of high quality sentences and parse trees with these two types of labels on sentence level. We characterize the sensitive information via several representative sensitive information detection models, in particular both keyword-based (n-gram) approaches and recent deep learning models, namely, recurrent neural networks (LSTM) and recursive neural networks (RecNN). Data and code are made publicly available.

Word clustering groups words that exhibit similar properties. One popular method for this is Brown clustering, which uses short-range distributional information to construct clusters. Specifically, this is a hard hierarchical clustering with a fixed-width beam that employs bi-grams and greedily minimizes global mutual information loss. The result is word clusters that tend to outperform or complement other word representations, especially when constrained by small datasets. However, Brown clustering has high computational complexity and does not lend itself to parallel computation. This, together with the lack of efficient implementations, limits their applicability in NLP. We present efficient implementations of Brown clustering and the alternative Exchange clustering as well as a number of methods to accelerate the computation of both hierarchical and flat clusters. We show empirically that clusters obtained with the accelerated method match the performance of clusters computed using the original methods.

Word embeddings are an active topic in the NLP research community. State-of-the-art neural models achieve high performance on downstream tasks, albeit at the cost of computationally expensive training. Cost aware solutions require cheaper models that still achieve good performance. We present several reproduction studies of intrinsic evaluation tasks that evaluate non-contextual word representations in multiple languages. Furthermore, we present 50-8-8, a new data set for the outlier identification task, which avoids limitations of the original data set, such as ambiguous words, infrequent words, and multi-word tokens, while increasing the number of test cases. The data set is expanded to contain semantic and syntactic tests and is multilingual (English, German, and Italian). We provide an in-depth analysis of word embedding models with a range of hyper-parameters. Our analysis shows the suitability of different models and hyper-parameters for different tasks and the greater difficulty of representing German and Italian languages.

Brown and Exchange word clusters have long been successfully used as word representations in Natural Language Processing (NLP) systems. Their success has been attributed to their seeming ability to represent both semantic and syntactic information. Using corpora representing several language families, we test the hypothesis that Brown and Exchange word clusters are highly effective at encoding morphosyntactic information. Our experiments show that word clusters are highly capable at distinguishing Parts of Speech. We show that increases in Average Mutual Information, the clustering algorithms’ optimization goal, are highly correlated with improvements in encoding of morphosyntactic information. Our results provide empirical evidence that downstream NLP systems addressing tasks dependent on morphosyntactic information can benefit from word cluster features.

We present Abbreviation Explorer, a system that supports interactive exploration of abbreviations that are challenging for Unsupervised Abbreviation Disambiguation (UAD). Abbreviation Explorer helps to identify long-forms that are easily confused, and to pinpoint likely causes such as limitations of normalization, language switching, or inconsistent typing. It can also support determining which long-forms would benefit from additional input text for unsupervised abbreviation disambiguation. The system provides options for creating corrective rules that merge redundant long-forms with identical meaning. The identified rules can be easily applied to the already existing vector spaces used by UAD to improve disambiguation performance, while also avoiding the cost of retraining.

Abbreviations and acronyms are a part of textual communication in most domains. However, abbreviations are not necessarily defined in documents that employ them. Understanding all abbreviations used in a given document often requires extensive knowledge of the target domain and the ability to disambiguate based on context. This creates considerable entry barriers to newcomers and difficulties in automated document processing. Existing abbreviation expansion systems or tools require substantial technical knowledge for set up or make strong assumptions which limit their use in practice. Here, we present Abbreviation Expander, a system that builds on state of the art methods for identification of abbreviations, acronyms and their definitions and a novel disambiguator for abbreviation expansion in an easily accessible web-based solution.