UTFPR at SemEval-2019 Task 5: Hate Speech Identification with Recurrent Neural Networks

Paetzold, Gustavo Henrique; Zampieri, Marcos; Malmasi, Shervin

doi:10.18653/v1/s19-2093

Cited by 9 publications

(4 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Comparison with previous Results It is not straightforward to compare with previous results due to different data splits, or labeled tweets becoming unavailable over time. The best hate detection results on the SemEval (SE) dataset were reported to be 0.65 in macro-F1 (Paetzold et al, 2019). Our results are favorable, ranging from 0.68-0.80 in macro-F1.…”

Section: Cross-dataset Resultsmentioning

confidence: 47%

Fight Fire with Fire: Fine-tuning Hate Detectors using Large Samples of Generated Hate Speech

Wullach¹,

Adler²,

Minkov³

2021

Findings of the Association for Computational Linguistics: EMNLP 2021

View full text Add to dashboard Cite

Automatic hate speech detection is hampered by the scarcity of labeled datasetd, leading to poor generalization. We employ pretrained language models (LMs) to alleviate this data bottleneck. We utilize the GPT LM for generating large amounts of synthetic hate speech sequences from available labeled examples, and leverage the generated data in fine-tuning large pretrained LMs on hate detection. An empirical study using the models of BERT, RoBERTa and ALBERT, shows that this approach improves generalization significantly and consistently within and across data distributions. In fact, we find that generating relevant labeled hate speech sequences is preferable to using out-of-domain, and sometimes also within-domain, human-labeled examples.

show abstract

Section: Cross-dataset Resultsmentioning

confidence: 47%

Fight Fire with Fire: Fine-tuning Hate Detectors using Large Samples of Generated Hate Speech

Wullach¹,

Adler²,

Minkov³

2021

Findings of the Association for Computational Linguistics: EMNLP 2021

View full text Add to dashboard Cite

show abstract

“…Recurrent neural networks ( Paetzold, Zampieri & Malmasi, 2019 ) are a type of neural network in which the connections between nodes form a directed graph along a temporal sequence. Among the different variants of this type of network, the Long Short-Term Memory Network (LSTM) ( Talita & Wiguna, 2019 ; Bisht et al, 2020 ; Zhao et al, 2020 ; Zhang et al, 2021 ), which were specifically designed to avoid the problem of long-term dependency.…”

Section: Related Workmentioning

confidence: 99%

Detecting racism and xenophobia using deep learning models on Twitter data: CNN, LSTM and BERT

Benítez-Andrades

González-Jiménez

López-Brea

et al. 2022

PeerJ Computer Science

View full text Add to dashboard Cite

With the growth that social networks have experienced in recent years, it is entirely impossible to moderate content manually. Thanks to the different existing techniques in natural language processing, it is possible to generate predictive models that automatically classify texts into different categories. However, a weakness has been detected concerning the language used to train such models. This work aimed to develop a predictive model based on BERT, capable of detecting racist and xenophobic messages in tweets written in Spanish. A comparison was made with different Deep Learning models. A total of five predictive models were developed, two based on BERT and three using other deep learning techniques, CNN, LSTM and a model combining CNN + LSTM techniques. After exhaustively analyzing the results obtained by the different models, it was found that the one that got the best metrics was BETO, a BERT-based model trained only with texts written in Spanish. The results of our study show that the BETO model achieves a precision of 85.22% compared to the 82.00% precision of the mBERT model. The rest of the models obtained between 79.34% and 80.48% precision. On this basis, it has been possible to justify the vital importance of developing native transfer learning models for solving Natural Language Processing (NLP) problems in Spanish. Our main contribution is the achievement of promising results in the field of racism and hate speech in Spanish by applying different deep learning techniques.

show abstract

“…For instance, in order to break the barrier of language dependency in word embedding approach, [19] conducted an ensemble of RNN classifiers, incorporating various features associated with user related information. [20] experimented with a robust system based on compositional RNNs able to handle even substantially noisy inputs, and reached competitive results for HS detection in English texts. In [21] authors developed a system for Twitter HS text identification based on two CNNs and feature embeddings including one-hot encoded character n-gram vectors and word embeddings, and they reported that the use of character n-gram does not help in the detection.…”

Section: Related Workmentioning

confidence: 99%

A Multi-Task Learning Approach to Hate Speech Detection Leveraging Sentiment Analysis

et al. 2021

View full text Add to dashboard Cite

The rise of social media platforms has significantly changed the way our world communicates, and part of those changes includes a rise in inappropriate behaviors, such as the use of aggressive and hateful language online. Detecting such content is crucial to filtering or blocking inappropriate content on the Web. However, due to the huge amount of data posted every day, automatic methods are essential for identifying this type of content. Seeking to address this issue, the Natural Language Processing community is increasingly involved in testing a wide range of techniques for hate speech detection. While achieving promising results, these techniques consider hate speech detection as the sole optimization objective, without involving other related tasks such as polarity and emotion classification that are strongly linked to offensive behavior. In this paper, we propose the first Multi-task approach that leverages the shared affective knowledge to detect hate speech in Spanish tweets, using a well-known Transformer-based model. Our results show that the combination of both polarity and emotion knowledge helps to detect hate speech more accurately across datasets.

show abstract

UTFPR at SemEval-2019 Task 5: Hate Speech Identification with Recurrent Neural Networks

Cited by 9 publications

References 18 publications

Fight Fire with Fire: Fine-tuning Hate Detectors using Large Samples of Generated Hate Speech

Fight Fire with Fire: Fine-tuning Hate Detectors using Large Samples of Generated Hate Speech

Detecting racism and xenophobia using deep learning models on Twitter data: CNN, LSTM and BERT

A Multi-Task Learning Approach to Hate Speech Detection Leveraging Sentiment Analysis

Contact Info

Product

Resources

About