Heuristics-Based Replenishment of Collocation Databases
PorTAL '02 Proceedings of the Third International Conference on Advances in Natural Language Processing
| Hi-index | 0.00 |
A very large Russian dictionary is described. It contains currently 3.6 million links between its 120,000 entries. The links are syntagmatic (collocations), paradigmatic (WordNet-like), or paronymic (words similar in letters or in morphs). The entries of the dictionary are single- or multiwords belonging to four main POS. The entries represent so-called grammemes rather than lexemes: e.g., nouns are represented as singular and plural; verbs are split into 'finite forms + infinitive', 'participle', and 'gerund'. The multiword entries in turn can be collocations---idiomatic free---whose parts are also entries of the same dictionary.