CoDeS project

Project factsheet

Project summary

The general aim of the project is to evaluate existing methods and devise new techniques of computational distributional semantics (CDS) in the area of discrimination and disambiguation of senses for Polish. One of the basic problems which has to be solved while analyzing natural language utterances is semantic ambiguity of their elements. Our goal is to elaborate methods determining the meaning of a particular word in the context in which it appears and methods for automatic detection if a word is used in more than one sense in an analyzed text. We want to focus our attention on noun-adjectives constructions and investigate how the meaning of a whole phrase may contribute to creation of sense models for its elements. We also plan to extend these methods to recognize words or phrases which are not used in their literal sense but figuratively. One of the project’s goal is to look for answers for some of general questions, for example, to what degree we will be able to describe sense differences using CDS methods, and which types of models are better suited for such a task and Polish data.

Resources

Word embedings

Polish word embeddings http://dsmodels.nlp.ipipan.waw.pl/ (A. Mykowiecka, M. Marciniak, P. Rychlik, 2017)

DSmodels demo - web service for calculating word similarity using Polish word embeddings

SimLex for Polish

Polish version of SimLex-999 (A. Mykowiecka, M. Marciniak, P. Rychlik, 2018).

Plain text (UTF-8 encoded) of Polish SimLex-999.

Literal/Non-Literal Adjective-Noun Phrases

FigAN: a list of 1526 adjective-noun phrases labelled with one of three categories : L (literal phrase), M (non-literal phrase), B (ambiguous phrase)

FigSen-1: 1833 short fragments of text selected from the NKJP (National Corpus of Polish, (Przepiórkowski et al., 2012)) in which all grammatically correct occurrences of all adjective-noun phrases were annotated at the phrase level either as literal (L) or figurative (M).

FigSen-2: Word-level metaphors corpus annotated by Joanna March_la and Maciej Rosiński. The same 1833 fragments with alternative annotation. It contains two versions (FigSen-2_9_11 and FigSen-2_16_09) of word-level annotations. You may likely want to work with the final annotations only (9_11). Metaphors are marked for selected types of part-of-speech tags in a column named "Metafory". Metaphorically used words are marked as "M", literally used words as "L". More details can be found in the (LTC 2019) paper listed below.

Word Sense Disambiguation

Python package for the WSD method presented in (Rutkowski, Sz., P. Rychlik, and A. Mykowiecka, 2019).

Sense-annotated text for WSD testing.

Publications

• Mykowiecka, A., M. Marciniak (2019) Experiments with ad hoc ambiguous abbreviation expansion. In Eben Holderness, Antonio Jimeno Yepes, Alberto Lavelli, Anne-Lyse Minard, James Pustejovsky, and Fabio Rinaldi, editors, Proceedings of the Tenth International Workshop on Health Text Mining and Information Analysis (LOUHI 2019), pages 44–53, Hong Kong, 2019. Association for Computational Linguistics

• Mykowiecka, A., M. Marciniak (2020), Are White Ravens Ever White? - Non-Literal Adjective-Noun Phrases in Polish, Proceedings of the 12th Language Resources and Evaluation Conference (LREC)

• Wawer, A., M. Marciniak and A. Mykowiecka (2019) Detecting word level metaphors in Polish, Proceedings of the 9th Language and Technology Conference: Human Language Technologies as a Challenge for Computer Science and Linguistics (LTC 2019).

• Rutkowski, Sz., P. Rychlik, and A. Mykowiecka. Estimating senses with sets of lexically related words for Polish word sense disambiguation, Proceedings of the 10th Global WordNet Conference (GWC 2019).

• Mykowiecka, A., M. Marciniak, and A. Wawer. Literal, metaphorical or both? Detecting metaphoricity in isolated adjective-noun phrases. In Beata Beigman Klebanov, Ekaterina Shutova, Patricia Lichtenstein, Smaranda Muresan, and Chee Wee, editors, Proceedings of the Workshop on Figurative Language Processing, pages 27–33. Association for Computational Linguistics, 2018.

• Mykowiecka, A., A.Wawer, and M. Marciniak. Detecting figurative word occurrences using recurrent neural networks. In Beata Beigman Klebanov, Ekaterina Shutova, Patricia Lichtenstein, Smaranda Muresan, and Chee Wee, editors, Proceedings of the Workshop on Figurative Language Processing, pages 124–127. Association for Computational Linguistics, 2018.

• Mykowiecka, A., M. Marciniak and P. Rychlik (2018) SimLex-999 for Polish, LREC 2018.

• Marciniak, M., A. Mykowiecka, and P. Rychlik. Recognition of irrelevant phrases in automatically extracted lists of domain terms. Terminology, 24(1):66–90, 2018.

• Wawer, A. and A. Mykowiecka (2017) Supervised and Unsupervised Word Sense Disambiguation on Word Embedding Vectors of Unambiguous Synonyms, Proceedings of the 1st Workshop on Sense, Concept and Entity Representations and their Applications, pages 120–125, Valencia, Spain, April 4 2017

• Wawer, A. and A. Mykowiecka (2017) Detecting Metaphorical Phrases in the Polish Language, Proc. of the Recent Advances in Natural Language Conference (RANLP), Warna.

• Mykowiecka, A., M. Marciniak and P. Rychlik (2017) Testing word embeddings for Polish, Cogntive Studies (17), DOI: 10.11649/cs.1468

• Mykowiecka, A., M. Marciniak and P. Rychlik (2016) Recognition of non-domain phrases in automatically extracted lists of terms, Proceedings of the 5th International Workshop on Computational Terminology Computerm2016