OCHADAI-KYOTO at SemEval-2021 Task 1: Enhancing Model Generalization and Robustness for Lexical Complexity Prediction

Taya, Yuki; Pereira, Lis Kanashiro; Cheng, Fei; Kobayashi, Ichiro

doi:10.18653/v1/2021.semeval-1.2

Cited by 4 publications

(3 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We evaluate the results of our system by applying the regression metrics in the execution of the supervised learning algorithms, specifically MAE, MSE, RMSE, and R2. We emphasize that we apply the methodologies of the winning teams, which are based on the application of language models based on the pre-trained and adjusted Transformers BERT and RoBERTa [9,34,35], together with the linguistic, syntactic and statistical characteristics, and the embedding results at the word and sentence level.…”

Section: Resultsmentioning

confidence: 99%

Combining Transformer Embeddings with Linguistic Features for Complex Word Identification

2022

View full text Add to dashboard Cite

Identifying which words present in a text may be difficult to understand by common readers is a well-known subtask in text complexity analysis. The advent of deep language models has also established the new state-of-the-art in this task by means of end-to-end semi-supervised (pre-trained) and downstream training of, mainly, transformer-based neural networks. Nevertheless, the usefulness of traditional linguistic features in combination with neural encodings is worth exploring, as the computational cost needed for training and running such networks is becoming more and more relevant with energy-saving constraints. This study explores lexical complexity prediction (LCP) by combining pre-trained and adjusted transformer networks with different types of traditional linguistic features. We apply these features over classical machine learning classifiers. Our best results are obtained by applying Support Vector Machines on an English corpus in an LCP task solved as a regression problem. The results show that linguistic features can be useful in LCP tasks and may improve the performance of deep learning systems.

show abstract

Section: Resultsmentioning

confidence: 99%

Combining Transformer Embeddings with Linguistic Features for Complex Word Identification

2022

View full text Add to dashboard Cite

show abstract

“…The lack of sufficient labeled data and domain and gender agnostic data limit the performance of those approaches . Considering the scarcity of annotated data and the problem of predicting the lexical complexity of single-word and multi-word expressions, (Taya et al, 2021) used an ensemble model over a set of transformer-based model with hand-crafted features to increase the model generalization and robustness. To improve the quality of the sentiment analysis task of lowresource languages such as Bangla, the authors Sultana et al, 2022) proposed aspect-based sentiment analysis using BOW and supervised machine learning techniques and provided two datasets for aspect-based sentiment analysis.…”

Section: Related Workmentioning

confidence: 99%

Proceedings of the First Workshop on Bangla Language Processing (BLP-2023)

2023

View full text Add to dashboard Cite

Although Spanish is one of the most spoken languages in the world, it was only until very recently that the development of linguistic technologies for it had a strong boost. However, this is not entirely true for some of its Latin American variants, such as the Mexican Spanish, which show phonetic, and also some lexical and semantic differences with respect to peninsular Spanish. This talk will focus on presenting the development of NLP for Mexican Spanish, emphasizing the path taken through the organization of different evaluation campaigns. It will present some data about Mexican Spanish as well as about the impact of the organization of shared tasks in the context of IberLEF for the development of the NLP area in our country, first as a mechanism to motivate more students to get involved, and then as a vehicle to build resources and design and implement specific methods. The talk will conclude by exposing some of the obstacles faced, our main achievements, and some plans for the coming years.Bio: Manuel Montes is Full Professor at the National Institute of Astrophysics, Optics and Electronics (INAOE) of Mexico. His research is on automatic text processing. He is author of more than 250 journal and conference papers in the fields of information retrieval, text mining and authorship analysis. He has been visiting professor at the Polytechnic University of Valencia (Spain), and the University of Alabama (USA). He is also member of the Mexican Academy of Sciences (AMC), and founding member of the Mexican Academy of Computer Science (AMEXCOMP), and the Mexican Association of Natural Language Processing (AMNLP). In the context of the latter, he has been the organizer of

show abstract

“…The lack of sufficient labeled data and domain and gender agnostic data limit the performance of those approaches (Islam et al, 2023a). Considering the scarcity of annotated data and the problem of predicting the lexical complexity of single-word and multi-word expressions, (Taya et al, 2021) used an ensemble model over a set of transformer-based model with hand-crafted features to increase the model generalization and robustness. To improve the quality of the sentiment analysis task of lowresource languages such as Bangla, the authors (Rahman and Kumar Dey, 2018; Sultana et al, 2022) proposed aspect-based sentiment analysis using BOW and supervised machine learning techniques and provided two datasets for aspect-based sentiment analysis.…”

Section: Related Workmentioning

confidence: 99%

Ushoshi2023 at BLP-2023 Task 2: A Comparison of Traditional to Advanced Linguistic Models to Analyze Sentiment in Bangla Texts

Khushbu,

Nur,

Ahmed

et al. 2023

Proceedings of the First Workshop on Bangla Language Processing (BLP-2023)

View full text Add to dashboard Cite

This article describes our analytical approach designed for BLP Workshop-2023 Task-2: in Sentiment Analysis. During actual task submission, we used DistilBERT. However, we later applied rigorous hyperparameter tuning and preprocessing, improving the result to 68% accuracy and a 68% F1 micro score with vanilla LSTM. Traditional machine learning models were applied to compare the result where 75% accuracy was achieved with traditional SVM. Our contributions are a) data augmentation using the oversampling method to remove data imbalance and b) attention masking for data encoding with masked language modeling to capture representations of language semantics effectively, by further demonstrating it with explainable AI. Originally, our system scored 0.26 micro-F1 in the competition and ranked 30th among the participants for a basic DistilBERT model, which we later improved to 0.68 and 0.65 with LSTM and XLM-RoBERTa-base models, respectively.

show abstract

OCHADAI-KYOTO at SemEval-2021 Task 1: Enhancing Model Generalization and Robustness for Lexical Complexity Prediction

Cited by 4 publications

References 18 publications

Combining Transformer Embeddings with Linguistic Features for Complex Word Identification

Combining Transformer Embeddings with Linguistic Features for Complex Word Identification

Proceedings of the First Workshop on Bangla Language Processing (BLP-2023)

Ushoshi2023 at BLP-2023 Task 2: A Comparison of Traditional to Advanced Linguistic Models to Analyze Sentiment in Bangla Texts

Contact Info

Product

Resources

About