A few years ago, I published TaPaCo, a corpus of sentential paraphrases for 73 languages (Scherrer, 2020). I could use some help to produce an updated and improved version of it. A Master's thesis could focus on some of the following points:
- Regenerating the corpus from a recent Tatoeba dump
- Improved filtering
- Automatic annotation of paraphrase types
- Extension of the methodology to other corpora (e.g. OpenSubtitles)
- Use of TaPaCo as training and/or evaluation corpus of paraphrasing models