HJ-Ky-0.1: fastText Models

Abstract

Here we release fastText embeddings described in the paper as follows: [...] we trained fastText embeddings on the Leipzig Corpus data. The training scheme was alsoSkip-Gram Negative Sampling, with 10 epochs, vector dimensions of 100 and 300, a window size of 5,and 10 negative samples. Character n-grams of 3 to 6 characters and 2,000,000 hashing buckets wereused for the hashing trick.

More by Гульнара Кабаева

• Human-Annotated NER Dataset for the Kyrgyz Language2025 · Citations: 2
• Syntactic Transfer to Kyrgyz Using the Treebank Translation MethodarXiv · 2024
• HJ-Ky-0.1: an Evaluation Dataset for Kyrgyz Word EmbeddingsarXiv · 2024
• Peculiarities of Creating Computer LexicographyJournal of Education Culture and Society · 2024