Persian Word Embedding Learning

Project description: In my bachelor’s thesis project, I aimed to learn embeddings for Persian words and assess the quality of them through different tasks. This task was not explored adequately for Persian words. As a thorough dataset was not accessible, I started with data collecting by crawling Persian news websites. Then, applied word2vec on the pre-processed dataset to achieve word embeddings. I extracted about hundreds of word relationships like synonymous, antonymous, gender, tense, etc. and qualitatively tested the embeddings using them. In addition, using Clustering algorithms, I successfuly clustered the word embeddings in meaningful categories like adverbs, places, countries, numbers, months, cinema, etc, and by taking use of Visualization techniques such as PCA, I visibly showed the effectiveness of the extracted embeddings.