Program code containing studies conducted as a part of thesis dedicated to aspect-based review summarization
This study is dedicated to aspect-based summarization. I fine-tuned LLMs ruBERT, mBERT and XLM-RoBERTa on reviews about restaurants and automobiles. There are two solutions to make summaries: use one of the model as a pair classificator or as a sequence tagger. Finally, there is clusterisation to obtain the result.
Jupyter Notebook, Python 3
- Jupyter Notebook, Python 3
- Each notebook contains installer for needed libraries
There are following notebooks:
- Preparation of laptop data
- Preparation of train data
- Fine-tuning of models
- Summarization pipelines
You can get laptop database from this link.
Yanina Khudina, currently enrolled in bachelor program "Fundamential and Computational Linguistics" at HSE University, Moscow.
- e-mail: [email protected]