Scwad – Compositional distributional modelling of Polish language semantics

Project factsheet

Project summary

Within the project, basic research will be conducted on compositional distributional semantics employed in modelling the meaning of phrases and sentences. A compositional distributional semantic model endeavours to determine the meaning of sentences or phrases based on the sophisticated procedure of composing distributional word vectors, and to generate a vector representation of this meaning. The degree of similarity between two vectors, which belong to the same vector space but represent meanings of different sentences, can be estimated with similarity measures. With respect to the Polish language, this scientific issue has been studied neither by us nor by other members of the natural language processing community in Poland. Within our pioneering studies, we will investigate whether it is possible to estimate compositional distributional semantic models for languages with a complex inflectional system and relatively free word order, such as Polish.

Achievements

• COMBO, the jointly trained neural tagger, morphological analyser, lemmatizer and dependency parser ranked 3rd/4th in the CoNLL 2018 UD shared task.
• COMBO won the shared task 1(A) in PolEval 2018 competition (Task 1(A) results).
• The morphosyntactic disambiguator Toygger won the shared task 1(A) in PolEval 2017 competition (Task 1(A) results).

Resources

• Polish CDSCorpus (Wróblewska and Krasnowska-Kieraś, 2017)
• AIDe - Corpus of Annotated Image Descriptions (Wróblewska, 2018a)
• Polish and English probing task data sets (Krasnowska-Kieraś and Wróblewska, 2019)

Tools and Demo

• Scwad-demo - wizualizacja wyników modeli parsowania zależnościowego, wynikania tekstowego i powiązania semantycznego opracowanych w projekcie.
• COMBO-pytorch - the reimplementation of COMBO in PyTorch library
• COMBO - the jointly trained neural tagger, morphological analyser, lemmatizer and dependency parser (Rybak and Wróblewska, 2018). The COMBO models for Polish trained on Polish Dependency Bank (Wróblewska, 2018b) are publicly available.
• Toygger - morphosyntactic disambiguator of Polish (Krasnowska-Kieraś, 2017)

Publications

• Katarzyna Krasnowska-Kieraś and Alina Wróblewska. 2019. Empirical Linguistic Study of Sentence Embeddings. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5729–5739. Association for Computational Linguistics, Florence, Italy.
• Alina Wróblewska and Piotr Rybak. 2019. Dependency Parsing of Polish. In Poznań Studies in Contemporary Linguistics, pages 305–337.
• Piotr Rybak and Alina Wróblewska. 2018. Semi-Supervised Neural System for Tagging, Parsing and Lematization. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 45–54. Association for Computational Linguistics.
• Piotr Rybak and Alina Wróblewska. 2018. Semi-Supervised Neural System for Tagging, Parsing and Lemmatization. Addendum. In Proceedings of the PolEval 2018 Workshop, pages 49–51. Institute of Computer Science, Polish Academy of Sciences.
• Alina Wróblewska. 2018. Polish Corpus of Annotated Descriptions of Images. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), pages 2141–2146. European Language Resources Association (ELRA).
• Alina Wróblewska. 2018. Extended and Enhanced Polish Dependency Bank in Universal Dependencies Format. In Proceedings of the Second Workshop on Universal Dependencies (UDW 2018), pages 173–182. Association for Computational Linguistics.
• Alina Wróblewska and Katarzyna Krasnowska-Kieraś. 2017. Polish Evaluation Dataset for Compositional Distributional Semantics Models. In Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 784–792. Association for Computational Linguistics, Vancouver, Canada.
• Alina Wróblewska, Katarzyna Krasnowska-Kieraś and Piotr Rybak. 2017. Towards the Evaluation of Feature Embedding Models of the Fusional Languages. In Proceedings of the 8th Language & Technology Conference: Human Language Technologies as a Challenge for Computer Science and Linguistics, pages 420–424. Fundacja Uniwersytetu im. Adama Mickiewicza w Poznaniu, Poznań, Poland.
• Katarzyna Krasnowska-Kieraś. 2017. Morphosyntactic Disambiguation for Polish with Bi-LSTM Neural Networks. In Proceedings of the 8th Language & Technology Conference: Human Language Technologies as a Challenge for Computer Science and Linguistics, pages 367–371. Fundacja Uniwersytetu im. Adama Mickiewicza w Poznaniu, Poznań, Poland.