Jens Lemmens - Academia.edu (original) (raw)

Papers by Jens Lemmens

arXiv (Cornell University), Mar 14, 2022

We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model... more We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT [3] 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.

We study the usefulness of hateful metaphors as features for the identification of the type and t... more We study the usefulness of hateful metaphors as features for the identification of the type and target of hate speech in Dutch Facebook comments. For this purpose, all hateful metaphors in the Dutch LiLaH corpus were annotated and interpreted in line with Conceptual Metaphor Theory and Critical Metaphor Analysis. We provide SVM and BERT/RoBERTa results, and investigate the effect of different metaphor information encoding methods on hate speech type and target detection accuracy. The results of the conducted experiments show that hateful metaphor features improve model performance for the both tasks. To our knowledge, it is the first time that the effectiveness of hateful metaphors as an information source for hate speech classification is investigated.

Proceedings of the 13th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis

Social media provide a rich source of data that can be mined and used for a wide variety of resea... more Social media provide a rich source of data that can be mined and used for a wide variety of research purposes. However, annotating this data can be expensive, yet necessary for state-ofthe-art pre-trained language models to achieve high prediction performance. Therefore, we combine pool-based active learning based on prediction uncertainty (an established method for reducing annotation costs) with unsupervised task adaptation through Masked Language Modeling (MLM). The results on three different datasets (two social media corpora, one benchmark dataset) show that task adaptation significantly improves results and that with only a fraction of the available training data, this approach reaches similar F1-scores as those achieved by an upper-bound baseline model fine-tuned on all training data. We hereby contribute to the scarce corpus of research on active learning with pre-trained language models and propose a cost-efficient annotation sampling and fine-tuning approach that can be applied to a wide variety of tasks and datasets.

In this paper, we present emotion lexicons of Croatian, Dutch and Slovene, based on manually corr... more In this paper, we present emotion lexicons of Croatian, Dutch and Slovene, based on manually corrected automatic translations of the English NRC Emotion lexicon. We evaluate the impact of the translation changes by measuring the change in supervised classification results of socially unacceptable utterances when lexicon information is used for feature construction. We further showcase the usage of the lexicons by calculating the difference in emotion distributions in texts containing and not containing socially unacceptable discourse, comparing them across four languages (English, Croatian, Dutch, Slovene) and two topics (migrants and LGBT). We show significant and consistent improvements in automatic classification across all languages and topics, as well as consistent (and expected) emotion distributions across all languages and topics, proving for the manually corrected lexicons to be a useful addition to the severely lacking area of emotion lexicons, the crucial resource for emo...

Cornell University - arXiv, Mar 14, 2022

We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model... more We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT [3] 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.

We study the usefulness of hateful metaphorsas features for the identification of the type and ta... more We study the usefulness of hateful metaphorsas features for the identification of the type and target of hate speech in Dutch Facebook comments. For this purpose, all hateful metaphors in the Dutch LiLaH corpus were annotated and interpreted in line with Conceptual Metaphor Theory and Critical Metaphor Analysis. We provide SVM and BERT/RoBERTa results, and investigate the effect of different metaphor information encoding methods on hate speech type and target detection accuracy. The results of the conducted experiments show that hateful metaphor features improve model performance for the both tasks. To our knowledge, it is the first time that the effectiveness of hateful metaphors as an information source for hatespeech classification is investigated.

Proceedings of the Second Workshop on Figurative Language Processing

arXiv (Cornell University), Mar 14, 2022

We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model... more We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT [3] 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.

We study the usefulness of hateful metaphors as features for the identification of the type and t... more We study the usefulness of hateful metaphors as features for the identification of the type and target of hate speech in Dutch Facebook comments. For this purpose, all hateful metaphors in the Dutch LiLaH corpus were annotated and interpreted in line with Conceptual Metaphor Theory and Critical Metaphor Analysis. We provide SVM and BERT/RoBERTa results, and investigate the effect of different metaphor information encoding methods on hate speech type and target detection accuracy. The results of the conducted experiments show that hateful metaphor features improve model performance for the both tasks. To our knowledge, it is the first time that the effectiveness of hateful metaphors as an information source for hate speech classification is investigated.

Proceedings of the 13th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis

Social media provide a rich source of data that can be mined and used for a wide variety of resea... more Social media provide a rich source of data that can be mined and used for a wide variety of research purposes. However, annotating this data can be expensive, yet necessary for state-ofthe-art pre-trained language models to achieve high prediction performance. Therefore, we combine pool-based active learning based on prediction uncertainty (an established method for reducing annotation costs) with unsupervised task adaptation through Masked Language Modeling (MLM). The results on three different datasets (two social media corpora, one benchmark dataset) show that task adaptation significantly improves results and that with only a fraction of the available training data, this approach reaches similar F1-scores as those achieved by an upper-bound baseline model fine-tuned on all training data. We hereby contribute to the scarce corpus of research on active learning with pre-trained language models and propose a cost-efficient annotation sampling and fine-tuning approach that can be applied to a wide variety of tasks and datasets.

In this paper, we present emotion lexicons of Croatian, Dutch and Slovene, based on manually corr... more In this paper, we present emotion lexicons of Croatian, Dutch and Slovene, based on manually corrected automatic translations of the English NRC Emotion lexicon. We evaluate the impact of the translation changes by measuring the change in supervised classification results of socially unacceptable utterances when lexicon information is used for feature construction. We further showcase the usage of the lexicons by calculating the difference in emotion distributions in texts containing and not containing socially unacceptable discourse, comparing them across four languages (English, Croatian, Dutch, Slovene) and two topics (migrants and LGBT). We show significant and consistent improvements in automatic classification across all languages and topics, as well as consistent (and expected) emotion distributions across all languages and topics, proving for the manually corrected lexicons to be a useful addition to the severely lacking area of emotion lexicons, the crucial resource for emo...

Cornell University - arXiv, Mar 14, 2022

We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model... more We present CoNTACT 1 : a Dutch language model adapted to the domain of COVID-19 tweets. The model was developed by continuing the pre-training phase of RobBERT [3] 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.

We study the usefulness of hateful metaphorsas features for the identification of the type and ta... more We study the usefulness of hateful metaphorsas features for the identification of the type and target of hate speech in Dutch Facebook comments. For this purpose, all hateful metaphors in the Dutch LiLaH corpus were annotated and interpreted in line with Conceptual Metaphor Theory and Critical Metaphor Analysis. We provide SVM and BERT/RoBERTa results, and investigate the effect of different metaphor information encoding methods on hate speech type and target detection accuracy. The results of the conducted experiments show that hateful metaphor features improve model performance for the both tasks. To our knowledge, it is the first time that the effectiveness of hateful metaphors as an information source for hatespeech classification is investigated.

Proceedings of the Second Workshop on Figurative Language Processing