Abstract
How can neural networks perform so well on compositional tasks even though they lack explicit compositional representations? We use a novel analysis technique called ROLE to show that recurrent neural networks perform well on such tasks by converging to solutions which implicitly represent symbolic structure. This method uncovers a symbolic structure which, when properly embedded in vector space, closely approximates the encodings of a standard seq2seq network trained to perform the compositional SCAN task. We verify the causal importance of the discovered symbolic structure by showing that, when we systematically manipulate hidden embeddings based on this symbolic structure, the model’s output is changed in the way predicted by our analysis.- Anthology ID:
- 2020.blackboxnlp-1.23
- Volume:
- Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
- Month:
- November
- Year:
- 2020
- Address:
- Online
- Editors:
- Afra Alishahi, Yonatan Belinkov, Grzegorz Chrupała, Dieuwke Hupkes, Yuval Pinter, Hassan Sajjad
- Venue:
- BlackboxNLP
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 238–254
- Language:
- URL:
- https://aclanthology.org/2020.blackboxnlp-1.23
- DOI:
- 10.18653/v1/2020.blackboxnlp-1.23
- Cite (ACL):
- Paul Soulos, R. Thomas McCoy, Tal Linzen, and Paul Smolensky. 2020. Discovering the Compositional Structure of Vector Representations with Role Learning Networks. In Proceedings of the Third BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP, pages 238–254, Online. Association for Computational Linguistics.
- Cite (Informal):
- Discovering the Compositional Structure of Vector Representations with Role Learning Networks (Soulos et al., BlackboxNLP 2020)
- PDF:
- https://preview.aclanthology.org/naacl24-info/2020.blackboxnlp-1.23.pdf
- Code
- psoulos/role-decomposition + additional community code
- Data
- SCAN, SST