Lithuanian Word embeddings

  Authors

Bielinskienė, Agnė ; Boizou, Loïc ; Bumbulienė, Ieva ; Kovalevskaitė, Jolanta ; Krilavičius, Tomas ; Mandravickaitė, Justina ; Rimkutė, Erika ; Vilkaitė-Lozdienė, Laura

 Project URL

http://mwe.lt/

 Date issued

2019

 Type

lexicalConceptualResource

 Size

331 203 entries

 Language(s)

Lithuanian

 Description

GloVe type word vectors (embeddings) for Lithuanian. Delfi.lt corpus (~70 million words) and StanfordNLP were used for training. The training consisted of several stages: 1) the vocabulary was compiled, eliminating words the the frequency less than 5; 2) word co-occurrence matrix was generated with window size of 5; 3) this matrix was randomly shuffled; 4) word vectors were generated (100 iterations, 200 dimensions). The final result consists of 331 203 unique word vectors.

 Publisher

Baltic Institute of Advanced Technology

 

Vytautas Magnus University

 Acknowledgement

The Research Council of Lithuania

Project code: LIP-027/2016

Project name: Automatic Identification of Lithuanian Multi-word Expressions (PASTOVU)

 Subject(s)

word embeddings Lithuanian embeddings

 Collection(s)

CLARIN-LT

Show full item record