Institutional Repository of Vilnius University
BERT modelio panaudojimas įvairiems natūralios kalbos apdorojimo uždaviniams spręsti /
Abstract
dc:descriptionIn this paper BERT and DistilBERT machine learning models are fine-tuned to solve two tasks - sentiment analysis and named entity recognition. The results of these models are compared with each other and with the most commonly used ways of solving each task. To compare the results, each method of solving the task is implemented with the same data set. BERT and DistilBERT results of the sentiment analysis task are compared with the results of TF-IDF, bag of words and Word2Vec vectorization algorithms combined with logistic regression, naive-Bayes, SVM and kNN classification algorithms. The models results of the named entity recognition task are compared to the results of the EntityRecognizer component of the SpaCy library. The obtained results show that the machine learning models fine-tuned to the sentiment analysis task provide higher accuracy than the combinations of vectorization and classification algorithms (the accuracy of the BERT model results – 93.5%). Meanwhile, machine learning models fine-tuned to the named entity recognition task provide worse results than the EntityRecognizer component of SpaCy library (F1 metric of BERT model results – 51.73%).
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Murza, Karolis,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD192842844&prefLang=en_US