Gazetteer for Polish Named Entities

The Gazetteer for Polish Named Entities was used within the SProUT platform, initially for information extraction from Polish texts, and then for the automatic pre-annotation of the National Corpus of Polish (NKJP) on the level of named entities. Its construction, contents and use have been described in:

• SAVARY, A., PISKORSKI, J. (2011). Language Resources for Named Entity Annotation in the National Corpus of Polish, in Control and Cybernetics 40(2), Systems Research Institute, Polish Academy of Sciences, Warsaw, Poland, pp. 361-391.

• SAVARY, A., PISKORSKI, J. (2010). Lexicons and Grammars for Named Entity Annotation in the National Corpus of Polish, in Proceedings of the 18th International Conference Intelligent Information Systems (IIS'10), Siedlce, Poland.

• PISKORSKI, J. (2005). Named-Entity Recognition for Polish with SProUT, in LNCS Vol 3490: Proceedings of IMTCI 2004, Warsaw, Poland.

The gazetteer contains 153,477 inflected entries of Polish (and some foreign) proper names and named entity components:

• forenames and surnames,
• city, country, mountain, region and river names,
• institution names,
• relational adjectives and inhabitant names stemming from country names,
• named entity triggers (months, days, positions, etc.).

The file DOES NOT contain inhabitant names and relational adjectives stemming from Polish settlements. These data, owned by the PWN publisher, were used within the NKJP project under a particular licence and are concerned by the copyright.

Authors

• Agata Savary - NKJP version version of the gazetteer; LMF format definition

• Michał Lenart - LMF conversion and validation

• Jakub Piskorski - earlier version of the gazetteer used for information extraction from Polish texts

License

The data are available under the 2-clause BSD licence.

Available resources

• Text version as used with Sprout for NKJP pre-annotation

• LMF format definition and conversion guidelines

• LMF-compliant version containing:

  • LMF format definition and conversion guidelines,
  • Relax NG schema, morphosyntax configuration file and validation scrypts,
  • gramatically complete gazetteer entries (9,060 lemmas and 95,359 word forms),
  • gramatically incomplete gazetteer entries (35,884 lemmas and 40,612 word forms).