University of Twente Student Theses
This laptop has great coffee : training a Dutch ABSA model from customer reviews
Middelraad, F. (2022) This laptop has great coffee : training a Dutch ABSA model from customer reviews.
PDF
418kB |
Abstract: | Aspect-based sentiment analysis (ABSA) finds the sentiment of a feature in a text. It can be useful for both businesses and customers to get an overview of product reviews. In this paper, I explore the possibility to use data extracted from Dutch Bol.com customer reviews rather than annotated data, since annotating data is a labour-intensive task. I try two different methods: in the first one, I fine-tune a Dutch GPT-2 model to generate aspects together with a sentiment label. In the second, I fine-tune the same GPT-2 model to generate only an aspect. I then fine-tune a Dutch BERT model (BERTje) to classify the sentiment. Using both an automatic and human evaluation, I find that the combination of GPT-2 and BERT outperforms the multi-tasked GPT-2. |
Item Type: | Essay (Bachelor) |
Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
Subject: | 54 computer science |
Programme: | Computer Science BSc (56964) |
Link to this item: | https://purl.utwente.nl/essays/91778 |
Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Repository Staff Only: item control page