Tsegaye Misikir Tashu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Tsegaye Misikir Tashu has a total of 25 co-authors. Here are the top co-authors by number of shared publications:
- Tomáš Horváth
School of Computing, Engineering & Built Environment, Edinburgh Napier University, Scotland, UK
7 shared publications Tomáš Horváth
5 shared publications- Matias Valdenegro-Toro
University of Groningen
3 shared publications Tigist Ali Gebeyehu
ETHIOPIAN PUBLIC HEALTH INSTITUTE
3 shared publicationsHaddad Philip
Eötvös Loránd University
2 shared publicationsAleksandar Angelov
Department of Artificial Intelligence University of Groningen
1 shared publication
Latest Publications
- 19 October 2026
- Published by Springer Nature
- p. 488-507
Multilingual Behavioural Emotion Classification with Instruction-Tuned LLMs: Cross-Lingual Multi-label Evidence from XED
- 19 October 2026
- Published by Springer Nature
- p. 488-507
Multilingual Behavioural Emotion Classification with Instruction-Tuned LLMs: Cross-Lingual Multi-label Evidence from XED
- 19 October 2026
- Published by Springer Nature
- p. 419-437
Conversational Emotion-Cause Pair Extraction: A Single-Step Architecture and the Role of Emotion-Label Supervision
- 19 October 2026
- Published by Springer Nature
- p. 419-437
Conversational Emotion-Cause Pair Extraction: A Single-Step Architecture and the Role of Emotion-Label Supervision
- 19 October 2026
- Published by Springer Nature
- p. 38-56
Multimodal Fusion for Uncertainty-Aware Emotion Recognition in Conversations
- 19 October 2026
- Published by Springer Nature
- p. 38-56
Multimodal Fusion for Uncertainty-Aware Emotion Recognition in Conversations
- 6 August 2025
- Published by Springer Nature
- p. 67-78
Phrase-Level Adversarial Data Generation for Bias Mitigation in Automatic Essay Scoring
- 6 August 2025
- Published by Springer Nature
- p. 67-78
Phrase-Level Adversarial Data Generation for Bias Mitigation in Automatic Essay Scoring
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 120-137
Confidence Calibration in Large Language Model-Based Entity Matching
Iris Kamsteeg, Juan Cardenas-Cartagena, Floris van Beers, Tsegaye Misikir Tashu, Matias Valdenegro-Toro. Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025). 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 120-137
Open AccessConfidence Calibration in Large Language Model-Based Entity Matching
Iris Kamsteeg, Juan Cardenas-Cartagena, Floris van Beers, Tsegaye Misikir Tashu, Matias Valdenegro-Toro. Proceedings of the 2nd Workshop on Uncertainty-Aware NLP (UncertaiNLP 2025). 2025.
- Open Access
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 351-364
Difficulty Estimation in Natural Language Tasks with Action Scores
Aleksandar Angelov, Tsegaye Misikir Tashu, Matias Valdenegro-Toro. Proceedings of the 5th Workshop on Trustworthy NLP (TrustNLP 2025). 2025.
- 1 January 2025
- Published by Association for Computational Linguistics (ACL)
- p. 351-364
Open AccessDifficulty Estimation in Natural Language Tasks with Action Scores
Aleksandar Angelov, Tsegaye Misikir Tashu, Matias Valdenegro-Toro. Proceedings of the 5th Workshop on Trustworthy NLP (TrustNLP 2025). 2025.
- Open Access
Phrase-Level Adversarial Training for Mitigating Bias in Neural Network-based Automatic Essay Scoring
Automatic Essay Scoring (AES) is widely used to evaluate candidates for educational purposes. However, due to the lack of representative data, most existing AES systems are not robust, and their scoring predictions are biased towards the most represented data samples. In this study, we propose a model-agnostic phrase-level method to generate an adversarial essay set to address the biases and robustness of AES models. Specifically, we construct an attack test set comprising samples from the original test set and adversarially generated samples using our proposed method. To evaluate the effectiveness of the attack strategy and data augmentation, we conducted a comprehensive analysis utilizing various neural network scoring models. Experimental results show that the proposed approach significantly improves AES model performance in the presence of adversarial examples and scenarios without such attacks.
Open AccessPhrase-Level Adversarial Training for Mitigating Bias in Neural Network-based Automatic Essay Scoring
Automatic Essay Scoring (AES) is widely used to evaluate candidates for educational purposes. However, due to the lack of representative data, most existing AES systems are not robust, and their scoring predictions are biased towards the most represented data samples. In this study, we propose a model-agnostic phrase-level method to generate an adversarial essay set to address the biases and robustness of AES models. Specifically, we construct an attack test set comprising samples from the original test set and adversarially generated samples using our proposed method. To evaluate the effectiveness of the attack strategy and data augmentation, we conducted a comprehensive analysis utilizing various neural network scoring models. Experimental results show that the proposed approach significantly improves AES model performance in the presence of adversarial examples and scenarios without such attacks.
- Open Access
Mapping Transformer Leveraged Embeddings for Cross-Lingual Document Representation
Recommendation systems, for documents, have become tools to find relevant content on the Web. However, these systems have limitations when it comes to recommending documents in languages different from the query language, which means they might overlook resources in non-native languages. This research focuses on representing documents across languages by using Transformer Leveraged Document Representations (TLDRs) that are mapped to a cross-lingual domain. Four multilingual pre-trained transformer models (mBERT, mT5 XLM RoBERTa, ErnieM) were evaluated using three mapping methods across 20 language pairs representing combinations of five selected languages of the European Union. Metrics like Mate Retrieval Rate and Reciprocal Rank were used to measure the effectiveness of mapped TLDRs compared to non-mapped ones. The results highlight the power of cross-lingual representations achieved through pre-trained transformers and mapping approaches suggesting a promising direction for expanding beyond language connections, between two specific languages.
Open AccessMapping Transformer Leveraged Embeddings for Cross-Lingual Document Representation
Recommendation systems, for documents, have become tools to find relevant content on the Web. However, these systems have limitations when it comes to recommending documents in languages different from the query language, which means they might overlook resources in non-native languages. This research focuses on representing documents across languages by using Transformer Leveraged Document Representations (TLDRs) that are mapped to a cross-lingual domain. Four multilingual pre-trained transformer models (mBERT, mT5 XLM RoBERTa, ErnieM) were evaluated using three mapping methods across 20 language pairs representing combinations of five selected languages of the European Union. Metrics like Mate Retrieval Rate and Reciprocal Rank were used to measure the effectiveness of mapped TLDRs compared to non-mapped ones. The results highlight the power of cross-lingual representations achieved through pre-trained transformers and mapping approaches suggesting a promising direction for expanding beyond language connections, between two specific languages.
- 30 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 35, 1-7
Course Review Sentiment Analysis: A Comparative Study Of Machine Learning and Deep Learning Methods
In this ever-evolving world of education, one of the key tools for improving the quality of teaching and learning is understanding student sentiment and feedback through course reviews. This study analyzed more than 100,000 reviews of courses on the online learning platform Coursera, with the objective of comparing the efficacy of various sentiment analysis methods. The initial preprocessing steps involved data cleaning using different Natural Language Processing (NLP) techniques. The cleaned data subsequently underwent transformation into TF-IDF and word embeddings were created using a pre-trained model. Naive Bayes, Random Forest, and SVM models were trained on the vectorized TF-IDF data and fine-tuned using a grid search method. Meanwhile, the word embeddings were used to train the LSTM and GRU models, which were optimized through Bayesian methods. Additionally, the BERT model was incorporated for further comparison. The findings revealed that BERT outperformed all other models in metrics such as accuracy, precision, recall, and F1 score, making it the most effective for the sentiment analysis of course reviews in this study. This comprehensive analysis aims to offer valuable insight into the sentiments of course evaluation, ultimately striving to improve the educational experiences of students. It also discusses the significance of the findings and highlights potential areas of improvement for future research.
- 30 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 35, 1-7
Course Review Sentiment Analysis: A Comparative Study Of Machine Learning and Deep Learning Methods
In this ever-evolving world of education, one of the key tools for improving the quality of teaching and learning is understanding student sentiment and feedback through course reviews. This study analyzed more than 100,000 reviews of courses on the online learning platform Coursera, with the objective of comparing the efficacy of various sentiment analysis methods. The initial preprocessing steps involved data cleaning using different Natural Language Processing (NLP) techniques. The cleaned data subsequently underwent transformation into TF-IDF and word embeddings were created using a pre-trained model. Naive Bayes, Random Forest, and SVM models were trained on the vectorized TF-IDF data and fine-tuned using a grid search method. Meanwhile, the word embeddings were used to train the LSTM and GRU models, which were optimized through Bayesian methods. Additionally, the BERT model was incorporated for further comparison. The findings revealed that BERT outperformed all other models in metrics such as accuracy, precision, recall, and F1 score, making it the most effective for the sentiment analysis of course reviews in this study. This comprehensive analysis aims to offer valuable insight into the sentiments of course evaluation, ultimately striving to improve the educational experiences of students. It also discusses the significance of the findings and highlights potential areas of improvement for future research.
- 27 April 2023
- Published by Elsevier in Applied Soft Computing
- Vol. 142
NCC: Neural concept compression for multilingual document recommendation
- 27 April 2023
- Published by Elsevier in Applied Soft Computing
- Vol. 142
NCC: Neural concept compression for multilingual document recommendation