Walter Daelemans

Victor De Marez, Luna De Bruyne, Walter Daelemans

Factual sycophancy occurs when a language model abandons a correct, verifiable answer under social pressure. Because a flip occurs only when pressure toward a false answer exceeds the model's neutral preference for the truth, flip rates conflate two mechanisms: the strength of that baseline preference (truth margin), and how far pressure shifts it (manipulation sensitivity). We decompose factual sycophancy into these channels and use them to separate the effects of size and instruction tuning across 56 open-weight models spanning 0.3B-32B parameters and 13 manipulation types. We find that vulnerability is governed mainly by size, but instruction tuning changes how size acts: small instruction-tuned models can become less robust, whereas large instruction-tuned models usually become more robust. Instruction tuning primarily increases truth margin, but its behavioral effect depends on manipulation type. Scaling also changes the two channels differently: base models gain margin but become mildly more manipulation-sensitive, whereas instruction-tuned models gain margin faster and become less sensitive. Factual sycophancy is therefore not a single scalar property. Evaluations should report channel-specific, manipulation-specific, and size-conditioned robustness rather than flip rates alone.

Maxime De Bruyn, Ehsan Lotfi, Jeska Buhmann et al.

Automatic evaluation of open-domain dialogs remains an unsolved problem. Moreover, existing methods do not correlate strongly with human annotations. This paper presents a new automated evaluation method using follow-ups: we measure the probability that a language model will continue the conversation with a fixed set of follow-ups (e.g., not really relevant here, what are you trying to say). When compared against twelve existing methods, our new evaluation achieves the highest correlation with human evaluations.

Jens Lemmens, Jens Van Nooten, Tim Kreutz et al.

We present CoNTACT: a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT (Delobelle, 2020) by using 2.8M Dutch COVID-19 related tweets posted in 2021. In order to test the performance of the model and compare it to RobBERT, the two models were tested on two tasks: (1) binary vaccine hesitancy detection and (2) detection of arguments for vaccine hesitancy. For both tasks, not only Twitter but also Facebook data was used to show cross-genre performance. In our experiments, CoNTACT showed statistically significant gains over RobBERT in all experiments for task 1. For task 2, we observed substantial improvements in virtually all classes in all experiments. An error analysis indicated that the domain adaptation yielded better representations of domain-specific terminology, causing CoNTACT to make more accurate classification decisions.

Chris Emmery, Ákos Kádár, Grzegorz Chrupała et al.

A limited amount of studies investigates the role of model-agnostic adversarial behavior in toxic content classification. As toxicity classifiers predominantly rely on lexical cues, (deliberately) creative and evolving language-use can be detrimental to the utility of current corpora and state-of-the-art models when they are deployed for content moderation. The less training data is available, the more vulnerable models might become. This study is, to our knowledge, the first to investigate the effect of adversarial behavior and augmentation for cyberbullying detection. We demonstrate that model-agnostic lexical substitutions significantly hurt classifier performance. Moreover, when these perturbed samples are used for augmentation, we show models become robust against word-level perturbations at a slight trade-off in overall task performance. Augmentations proposed in prior work on toxicity prove to be less effective. Our results underline the need for such evaluations in online harm areas with small corpora. The perturbed data, models, and code are available for reproduction at https://github.com/cmry/augtox

Maxime De Bruyn, Ehsan Lotfi, Jeska Buhmann et al.

Knowledgeable FAQ chatbots are a valuable resource to any organization. While powerful and efficient retrieval-based models exist for English, it is rarely the case for other languages for which the same amount of training data is not available. In this paper, we propose a novel pre-training procedure to adapt ConveRT, an English conversational retriever model, to other languages with less training data available. We apply it for the first time to the task of Dutch FAQ answering related to the COVID-19 vaccine. We show it performs better than an open-source alternative in both a low-data regime and a high-data regime.

Madhumita Sushil, Simon Šuster, Walter Daelemans

Understanding the behavior of a trained network and finding explanations for its outputs is important for improving the network's performance and generalization ability, and for ensuring trust in automated systems. Several approaches have previously been proposed to identify and visualize the most important features by analyzing a trained network. However, the relations between different features and classes are lost in most cases. We propose a technique to induce sets of if-then-else rules that capture these relations to globally explain the predictions of a network. We first calculate the importance of the features in the trained network. We then weigh the original inputs with these feature importance scores, simplify the transformed input space, and finally fit a rule induction model to explain the model predictions. We find that the output rule-sets can explain the predictions of a neural network trained for 4-class text classification from the 20 newsgroups dataset to a macro-averaged F-score of 0.80. We make the code available at https://github.com/clips/interpret_with_rules.

Stéphan Tulkens, Lisa Hilte, Elise Lodewyckx et al.

We present a dictionary-based approach to racism detection in Dutch social media comments, which were retrieved from two public Belgian social media sites likely to attract racist reactions. These comments were labeled as racist or non-racist by multiple annotators. For our approach, three discourse dictionaries were created: first, we created a dictionary by retrieving possibly racist and more neutral terms from the training data, and then augmenting these with more general words to remove some bias. A second dictionary was created through automatic expansion using a \texttt{word2vec} model trained on a large corpus of general Dutch text. Finally, a third dictionary was created by manually filtering out incorrect expansions. We trained multiple Support Vector Machines, using the distribution of words over the different categories in the dictionaries as features. The best-performing model used the manually cleaned dictionary and obtained an F-score of 0.46 for the racist class on a test set consisting of unseen Dutch comments, retrieved from the same sites used for the training set. The automated expansion of the dictionary only slightly boosted the model's performance, and this increase in performance was not statistically significant. The fact that the coverage of the expanded dictionaries did increase indicates that the words that were automatically added did occur in the corpus, but were not able to meaningfully impact performance. The dictionaries, code, and the procedure for requesting the corpus are available at: https://github.com/clips/hades

Stéphan Tulkens, Simon Šuster, Walter Daelemans

In this paper, we report a knowledge-based method for Word Sense Disambiguation in the domains of biomedical and clinical text. We combine word representations created on large corpora with a small number of definitions from the UMLS to create concept representations, which we then compare to representations of the context of ambiguous terms. Using no relational information, we obtain comparable performance to previous approaches on the MSH-WSD dataset, which is a well-known dataset in the biomedical domain. Additionally, our method is fast and easy to set up and extend to other domains. Supplementary materials, including source code, can be found at https: //github.com/clips/yarn

Nikolay Banar, Ehsan Lotfi, Walter Daelemans

Zero-shot evaluation of information retrieval (IR) models is often performed using BEIR; a large and heterogeneous benchmark composed of multiple datasets, covering different retrieval tasks across various domains. Although BEIR has become a standard benchmark for the zero-shot setup, its exclusively English content reduces its utility for underrepresented languages in IR, including Dutch. To address this limitation and encourage the development of Dutch IR models, we introduce BEIR-NL by automatically translating the publicly accessible BEIR datasets into Dutch. Using BEIR-NL, we evaluated a wide range of multilingual dense ranking and reranking models, as well as the lexical BM25 method. Our experiments show that BM25 remains a competitive baseline, and is only outperformed by the larger dense models trained for retrieval. When combined with reranking models, BM25 achieves performance on par with the best dense ranking models. In addition, we explored the impact of translation on the data by back-translating a selection of datasets to English, and observed a performance drop for both dense and lexical methods, indicating the limitations of translation for creating benchmarks. BEIR-NL is publicly available on the Hugging Face hub.

Ehsan Lotfi, Maxime De Bruyn, Jeska Buhmann et al.

The new wave of Large Language Models (LLM) has offered an efficient tool to curate sizeable conversational datasets. So far studies have mainly focused on task-oriented or generic open-domain dialogs, and have not fully explored the ability of LLMs in following complicated prompts. In this work, we focus on personalization, and employ LLMs to curate a dataset which is difficult and costly to crowd-source: PersonalityChat is a synthetic conversational dataset based upon the popular PersonaChat dataset, but conditioned on both personas and (Big-5) personality traits. Evaluating models fine-tuned on this dataset, we show that the personality trait labels can be used for trait-based personalization of generative dialogue models. We also perform a head-to-head comparison between PersonalityChat and PersonaChat, and show that training on the distilled dataset results in more fluent and coherent dialog agents in the small-model regime.

Ehsan Lotfi, Nikolay Banar, Nerses Yuzbashyan et al.

Statutory article retrieval plays a crucial role in making legal information more accessible to both laypeople and legal professionals. Multilingual countries like Belgium present unique challenges for retrieval models due to the need for handling legal issues in multiple languages. Building on the Belgian Statutory Article Retrieval Dataset (BSARD) in French, we introduce the bilingual version of this dataset, bBSARD. The dataset contains parallel Belgian statutory articles in both French and Dutch, along with legal questions from BSARD and their Dutch translation. Using bBSARD, we conduct extensive benchmarking of retrieval models available for Dutch and French. Our benchmarking setup includes lexical models, zero-shot dense models, and fine-tuned small foundation models. Our experiments show that BM25 remains a competitive baseline compared to many zero-shot dense models in both languages. We also observe that while proprietary models outperform open alternatives in the zero-shot setting, they can be matched or surpassed by fine-tuning small language-specific models. Our dataset and evaluation code are publicly available.

Ine Gevers, Victor De Marez, Luna De Bruyne et al.

In this study, we take a closer look at how Winograd schema challenges can be used to evaluate common sense reasoning in LLMs. Specifically, we evaluate generative models of different sizes on the popular WinoGrande benchmark. We release WinoWhat, a new corpus, in which each instance of the WinoGrande validation set is paraphrased. Additionally, we evaluate the performance on the challenge across five common sense knowledge categories, giving more fine-grained insights on what types of knowledge are more challenging for LLMs. Surprisingly, all models perform significantly worse on WinoWhat, implying that LLM reasoning capabilities are overestimated on WinoGrande. To verify whether this is an effect of benchmark memorization, we match benchmark instances to LLM trainingdata and create two test-suites. We observe that memorization has a minimal effect on model performance on WinoGrande.

Ine Gevers, Walter Daelemans

Large language models (LLMs) have achieved striking successes on many benchmarks, yet recent studies continue to expose fundamental weaknesses. In particular, tasks that require abstract reasoning remain challenging, often because they use representations such as grids, symbols, or visual patterns that differ from the natural language data LLMs are trained on. In this paper, we introduce Concept, a simple word-guessing board game, as a benchmark for probing abductive reasoning in a representation that is much closer to LLM pre-training data: natural language. Our results show that this game, easily solved by humans (with a success rate of over 90\%), is still very challenging for state-of-the-art LLMs (no model exceeds 40\% success rate). Specifically, we observe that LLMs struggle with interpreting other players' strategic intents, and with correcting initial hypotheses given sequential information updates. In addition, we extend the evaluation across multiple languages, and find that the LLM performance drops further in lower-resource languages (Dutch, French, and Spanish) compared to English.

Jens Van Nooten, Andriy Kosar, Guy De Pauw et al.

Distance-based unsupervised text classification is a method within text classification that leverages the semantic similarity between a label and a text to determine label relevance. This method provides numerous benefits, including fast inference and adaptability to expanding label sets, as opposed to zero-shot, few-shot, and fine-tuned neural networks that require re-training in such cases. In multi-label distance-based classification and information retrieval algorithms, thresholds are required to determine whether a text instance is "similar" to a label or query. Similarity between a text and label is determined in a dense embedding space, usually generated by state-of-the-art sentence encoders. Multi-label classification complicates matters, as a text instance can have multiple true labels, unlike in multi-class or binary classification, where each instance is assigned only one label. We expand upon previous literature on this underexplored topic by thoroughly examining and evaluating the ability of sentence encoders to perform distance-based classification. First, we perform an exploratory study to verify whether the semantic relationships between texts and labels vary across models, datasets, and label sets by conducting experiments on a diverse collection of realistic multi-label text classification (MLTC) datasets. We find that similarity distributions show statistically significant differences across models, datasets and even label sets. We propose a novel method for optimizing label-specific thresholds using a validation set. Our label-specific thresholding method achieves an average improvement of 46% over normalized 0.5 thresholding and outperforms uniform thresholding approaches from previous work by an average of 14%. Additionally, the method demonstrates strong performance even with limited labeled examples.

Nikolay Banar, Ehsan Lotfi, Jens Van Nooten et al.

Recently, embedding resources, including models, benchmarks, and datasets, have been widely released to support a variety of languages. However, the Dutch language remains underrepresented, typically comprising only a small fraction of the published multilingual resources. To address this gap and encourage the further development of Dutch embeddings, we introduce new resources for their evaluation and generation. First, we introduce the Massive Text Embedding Benchmark for Dutch (MTEB-NL), which includes both existing Dutch datasets and newly created ones, covering a wide range of tasks. Second, we provide a training dataset compiled from available Dutch retrieval datasets, complemented with synthetic data generated by large language models to expand task coverage beyond retrieval. Finally, we release a series of E5-NL models compact yet efficient embedding models that demonstrate strong performance across multiple tasks. We make our resources publicly available through the Hugging Face Hub and the MTEB package.

Ine Gevers, Walter Daelemans

This study aims to acquire more insights into the continuous pre-training phase of BERT regarding entity knowledge, using the COVID-19 pandemic as a case study. Since the pandemic emerged after the last update of BERT's pre-training data, the model has little to no entity knowledge about COVID-19. Using continuous pre-training, we control what entity knowledge is available to the model. We compare the baseline BERT model with the further pre-trained variants on the fact-checking benchmark Check-COVID. To test the robustness of continuous pre-training, we experiment with several adversarial methods to manipulate the input data, such as training on misinformation and shuffling the word order until the input becomes nonsensical. Surprisingly, our findings reveal that these methods do not degrade, and sometimes even improve, the model's downstream performance. This suggests that continuous pre-training of BERT is robust against misinformation. Furthermore, we are releasing a new dataset, consisting of original texts from academic publications in the LitCovid repository and their AI-generated false counterparts.

Ehsan Lotfi, Maxime De Bruyn, Jeska Buhmann et al.

Knowledge Grounded Conversation Models (KGCM) are usually based on a selection/retrieval module and a generation module, trained separately or simultaneously, with or without having access to a gold knowledge option. With the introduction of large pre-trained generative models, the selection and generation part have become more and more entangled, shifting the focus towards enhancing knowledge incorporation (from multiple sources) instead of trying to pick the best knowledge option. These approaches however depend on knowledge labels and/or a separate dense retriever for their best performance. In this work we study the unsupervised selection abilities of pre-trained generative models (e.g. BART) and show that by adding a score-and-aggregate module between encoder and decoder, they are capable of learning to pick the proper knowledge through minimising the language modelling loss (i.e. without having access to knowledge labels). Trained as such, our model - K-Mine - shows competitive selection and generation performance against models that benefit from knowledge labels and/or separate dense retriever.

Maxime De Bruyn, Ehsan Lotfi, Jeska Buhmann et al.

In this paper, we present the first multilingual FAQ dataset publicly available. We collected around 6M FAQ pairs from the web, in 21 different languages. Although this is significantly larger than existing FAQ retrieval datasets, it comes with its own challenges: duplication of content and uneven distribution of topics. We adopt a similar setup as Dense Passage Retrieval (DPR) and test various bi-encoders on this dataset. Our experiments reveal that a multilingual model based on XLM-RoBERTa achieves the best results, except for English. Lower resources languages seem to learn from one another as a multilingual model achieves a higher MRR than language-specific ones. Our qualitative analysis reveals the brittleness of the model on simple word changes. We publicly release our dataset, model and training script.

Nikolay Banar, Walter Daelemans, Mike Kestemont

Neural machine translation (NMT) is nowadays commonly applied at the subword level, using byte-pair encoding. A promising alternative approach focuses on character-level translation, which simplifies processing pipelines in NMT considerably. This approach, however, must consider relatively longer sequences, rendering the training process prohibitively expensive. In this paper, we discuss a novel, Transformer-based approach, that we compare, both in speed and in quality to the Transformer at subword and character levels, as well as previously developed character-level models. We evaluate our models on 4 language pairs from WMT'15: DE-EN, CS-EN, FI-EN and RU-EN. The proposed novel architecture can be trained on a single GPU and is 34% percent faster than the character-level Transformer; still, the obtained results are at least on par with it. In addition, our proposed model outperforms the subword-level model in FI-EN and shows close results in CS-EN. To stimulate further research in this area and close the gap with subword-level NMT, we make all our code and models publicly available.

Madhumita Sushil, Simon Šuster, Walter Daelemans

Several previous studies on explanation for recurrent neural networks focus on approaches that find the most important input segments for a network as its explanations. In that case, the manner in which these input segments combine with each other to form an explanatory pattern remains unknown. To overcome this, some previous work tries to find patterns (called rules) in the data that explain neural outputs. However, their explanations are often insensitive to model parameters, which limits the scalability of text explanations. To overcome these limitations, we propose a pipeline to explain RNNs by means of decision lists (also called rules) over skipgrams. For evaluation of explanations, we create a synthetic sepsis-identification dataset, as well as apply our technique on additional clinical and sentiment analysis datasets. We find that our technique persistently achieves high explanation fidelity and qualitatively interpretable rules.

Georg Rehm, Katrin Marheinecke, Stefanie Hegele et al.

Multilingualism is a cultural cornerstone of Europe and firmly anchored in the European treaties including full language equality. However, language barriers impacting business, cross-lingual and cross-cultural communication are still omnipresent. Language Technologies (LTs) are a powerful means to break down these barriers. While the last decade has seen various initiatives that created a multitude of approaches and technologies tailored to Europe's specific needs, there is still an immense level of fragmentation. At the same time, AI has become an increasingly important concept in the European Information and Communication Technology area. For a few years now, AI, including many opportunities, synergies but also misconceptions, has been overshadowing every other topic. We present an overview of the European LT landscape, describing funding programmes, activities, actions and challenges in the different countries with regard to LT, including the current state of play in industry and the LT market. We present a brief overview of the main LT-related activities on the EU level in the last ten years and develop strategic guidance with regard to four key dimensions.

Chris Emmery, Ben Verhoeven, Guy De Pauw et al.

The detection of online cyberbullying has seen an increase in societal importance, popularity in research, and available open data. Nevertheless, while computational power and affordability of resources continue to increase, the access restrictions on high-quality data limit the applicability of state-of-the-art techniques. Consequently, much of the recent research uses small, heterogeneous datasets, without a thorough evaluation of applicability. In this paper, we further illustrate these issues, as we (i) evaluate many publicly available resources for this task and demonstrate difficulties with data collection. These predominantly yield small datasets that fail to capture the required complex social dynamics and impede direct comparison of progress. We (ii) conduct an extensive set of experiments that indicate a general lack of cross-domain generalization of classifiers trained on these sources, and openly provide this framework to replicate and extend our evaluation criteria. Finally, we (iii) present an effective crowdsourcing method: simulating real-life bullying scenarios in a lab setting generates plausible data that can be effectively used to enrich real data. This largely circumvents the restrictions on data that can be collected, and increases classifier performance. We believe these contributions can aid in improving the empirical practices of future research in the field.

Simon Šuster, Madhumita Sushil, Walter Daelemans

Memory networks have been a popular choice among neural architectures for machine reading comprehension and question answering. While recent work revealed that memory networks can't truly perform multi-hop reasoning, we show in the present paper that vanilla memory networks are ineffective even in single-hop reading comprehension. We analyze the reasons for this on two cloze-style datasets, one from the medical domain and another including children's fiction. We find that the output classification layer with entity-specific weights, and the aggregation of passage information with relatively flat attention distributions are the most important contributors to poor results. We propose network adaptations that can serve as simple remedies. We also find that the presence of unseen answers at test time can dramatically affect the reported results, so we suggest controlling for this factor during evaluation.

Janneke van de Loo, Guy De Pauw, Walter Daelemans

This paper describes continuing work on semantic frame slot filling for a command and control task using a weakly-supervised approach. We investigate the advantages of using retraining techniques that take the output of a hierarchical hidden markov model as input to two inductive approaches: (1) discriminative sequence labelers based on conditional random fields and memory-based learning and (2) probabilistic context-free grammar induction. Experimental results show that this setup can significantly improve F-scores without the need for additional information sources. Furthermore, qualitative analysis shows that the weakly supervised technique is able to automatically induce an easily interpretable and syntactically appropriate grammar for the domain and task at hand.

Janneke van de Loo, Jort F. Gemmeke, Guy De Pauw et al.

We present a framework for the induction of semantic frames from utterances in the context of an adaptive command-and-control interface. The system is trained on an individual user's utterances and the corresponding semantic frames representing controls. During training, no prior information on the alignment between utterance segments and frame slots and values is available. In addition, semantic frames in the training data can contain information that is not expressed in the utterances. To tackle this weakly supervised classification task, we propose a framework based on Hidden Markov Models (HMMs). Structural modifications, resulting in a hierarchical HMM, and an extension called expression sharing are introduced to minimize the amount of training time and effort required for the user. The dataset used for the present study is PATCOR, which contains commands uttered in the context of a vocally guided card game, Patience. Experiments were carried out on orthographic and phonetic transcriptions of commands, segmented on different levels of n-gram granularity. The experimental results show positive effects of all the studied system extensions, with some effect differences between the different input representations. Moreover, evaluation experiments on held-out data with the optimal system configuration show that the extended system is able to achieve high accuracies with relatively small amounts of training data.

Tom De Smedt, Sylvia Jaki, Eduan Kotzé et al.

In this report, we present a study of eight corpora of online hate speech, by demonstrating the NLP techniques that we used to collect and analyze the jihadist, extremist, racist, and sexist content. Analysis of the multilingual corpora shows that the different contexts share certain characteristics in their hateful rhetoric. To expose the main features, we have focused on text classification, text profiling, keyword and collocation extraction, along with manual annotation and qualitative study.

Madhumita Sushil, Simon Šuster, Kim Luyckx et al.

We have three contributions in this work: 1. We explore the utility of a stacked denoising autoencoder and a paragraph vector model to learn task-independent dense patient representations directly from clinical notes. To analyze if these representations are transferable across tasks, we evaluate them in multiple supervised setups to predict patient mortality, primary diagnostic and procedural category, and gender. We compare their performance with sparse representations obtained from a bag-of-words model. We observe that the learned generalized representations significantly outperform the sparse representations when we have few positive instances to learn from, and there is an absence of strong lexical features. 2. We compare the model performance of the feature set constructed from a bag of words to that obtained from medical concepts. In the latter case, concepts represent problems, treatments, and tests. We find that concept identification does not improve the classification performance. 3. We propose novel techniques to facilitate model interpretability. To understand and interpret the representations, we explore the best encoded features within the patient representations obtained from the autoencoder model. Further, we calculate feature sensitivity across two networks to identify the most significant input features for different classification tasks when we use these pretrained representations as the supervised input. We successfully extract the most influential features for the pipeline using this technique.

Simon Šuster, Walter Daelemans

We present a new dataset for machine comprehension in the medical domain. Our dataset uses clinical case reports with around 100,000 gap-filling queries about these cases. We apply several baselines and state-of-the-art neural readers to the dataset, and observe a considerable gap in performance (20% F1) between the best human and machine readers. We analyze the skills required for successful answering and show how reader performance varies depending on the applicable skills. We find that inferences using domain knowledge and object tracking are the most frequently required skills, and that recognizing omitted information and spatio-temporal reasoning are the most difficult for the machines.

Cynthia Van Hee, Gilles Jacobs, Chris Emmery et al.

While social media offer great communication opportunities, they also increase the vulnerability of young people to threatening situations online. Recent studies report that cyberbullying constitutes a growing problem among youngsters. Successful prevention depends on the adequate detection of potentially harmful messages and the information overload on the Web requires intelligent systems to identify potential risks automatically. The focus of this paper is on automatic cyberbullying detection in social media text by modelling posts written by bullies, victims, and bystanders of online bullying. We describe the collection and fine-grained annotation of a training corpus for English and Dutch and perform a series of binary classification experiments to determine the feasibility of automatic cyberbullying detection. We make use of linear support vector machines exploiting a rich feature set and investigate which information sources contribute the most for this particular task. Experiments on a holdout test set reveal promising results for the detection of cyberbullying-related posts. After optimisation of the hyperparameters, the classifier yields an F1-score of 64% and 61% for English and Dutch respectively, and considerably outperforms baseline systems based on keywords and word unigrams.

Madhumita Sushil, Simon Šuster, Kim Luyckx et al.

We have two main contributions in this work: 1. We explore the usage of a stacked denoising autoencoder, and a paragraph vector model to learn task-independent dense patient representations directly from clinical notes. We evaluate these representations by using them as features in multiple supervised setups, and compare their performance with those of sparse representations. 2. To understand and interpret the representations, we explore the best encoded features within the patient representations obtained from the autoencoder model. Further, we calculate the significance of the input features of the trained classifiers when we use these pretrained representations as input.

Pieter Fivez, Simon Šuster, Walter Daelemans

We present an unsupervised context-sensitive spelling correction method for clinical free-text that uses word and character n-gram embeddings. Our method generates misspelling replacement candidates and ranks them according to their semantic fit, by calculating a weighted cosine similarity between the vectorized representation of a candidate and the misspelling context. To tune the parameters of this model, we generate self-induced spelling error corpora. We perform our experiments for two languages. For English, we greatly outperform off-the-shelf spelling correction tools on a manually annotated MIMIC-III test set, and counter the frequency bias of a noisy channel model, showing that neural embeddings can be successfully exploited to improve upon the state-of-the-art. For Dutch, we also outperform an off-the-shelf spelling correction tool on manually annotated clinical records from the Antwerp University Hospital, but can offer no empirical evidence that our method counters the frequency bias of a noisy channel model in this case as well. However, both our context-sensitive model and our implementation of the noisy channel model obtain high scores on the test set, establishing a state-of-the-art for Dutch clinical spelling correction with the noisy channel model.

Simon Šuster, Stéphan Tulkens, Walter Daelemans

Clinical NLP has an immense potential in contributing to how clinical practice will be revolutionized by the advent of large scale processing of clinical records. However, this potential has remained largely untapped due to slow progress primarily caused by strict data access policies for researchers. In this paper, we discuss the concern for privacy and the measures it entails. We also suggest sources of less sensitive data. Finally, we draw attention to biases that can compromise the validity of empirical research and lead to socially harmful applications.

Stéphan Tulkens, Chris Emmery, Walter Daelemans

Word embeddings have recently seen a strong increase in interest as a result of strong performance gains on a variety of tasks. However, most of this research also underlined the importance of benchmark datasets, and the difficulty of constructing these for a variety of language-specific tasks. Still, many of the datasets used in these tasks could prove to be fruitful linguistic resources, allowing for unique observations into language use and variability. In this paper we demonstrate the performance of multiple types of embeddings, created with both count and prediction-based architectures on a variety of corpora, in two language-specific tasks: relation evaluation, and dialect identification. For the latter, we compare unsupervised methods with a traditional, hand-crafted dictionary. With this research, we provide the embeddings themselves, the relation evaluation task benchmark for use in further research, and demonstrate how the benchmarked embeddings prove a useful unsupervised linguistic resource, effectively used in a downstream task.

Vincent Van Asch, Walter Daelemans

The goal of this paper is to investigate the connection between the performance gain that can be obtained by selftraining and the similarity between the corpora used in this approach. Self-training is a semi-supervised technique designed to increase the performance of machine learning algorithms by automatically classifying instances of a task and adding these as additional training material to the same classifier. In the context of language processing tasks, this training material is mostly an (annotated) corpus. Unfortunately self-training does not always lead to a performance increase and whether it will is largely unpredictable. We show that the similarity between corpora can be used to identify those setups for which self-training can be beneficial. We consider this research as a step in the process of developing a classifier that is able to adapt itself to each new test corpus that it is presented with.

Claudia Peersman, Walter Daelemans, Reinhild Vandekerckhove et al.

We present a corpus-based analysis of the effects of age, gender and region of origin on the production of both "netspeak" or "chatspeak" features and regional speech features in Flemish Dutch posts that were collected from a Belgian online social network platform. The present study shows that combining quantitative and qualitative approaches is essential for understanding non-standard linguistic variation in a CMC corpus. It also presents a methodology that enables the systematic study of this variation by including all non-standard words in the corpus. The analyses resulted in a convincing illustration of the Adolescent Peak Principle. In addition, our approach revealed an intriguing correlation between the use of regional speech features and chatspeak features.