Co-occurrence Weight Selection for Word Embeddings to Enhance Test Performance

Abstract

This study revisits the problem of maximizing the performance of mathematical word representations for a given task. It is aimed to improve performance in analogy and similarity tasks by suggesting innovative weights instead of the counting weights used conventionally in counting-based methods of generating word representations (adding the statistics of word co-occurrences to the account). The language of study was selected as Turkish. The root structures of Turkish words were managed during the compilation of corpus such that each word having a suffix was considered as a new word. The performance of the proposed co-occurrence weights are analyzed with respect to the varying parameter and the results are presented within the paper.

Keywords

• kelime temsilleri
• doğal dil işleme
• i̇statistiksel dilbilimi
• word embeddings
• natural language processing
• statistical linguistics