Query Expansion for Bangla Search Engine Pipilika

Query expansion aims to extend the user query with related terms for improving the search results of a search engine like Google, Pipilika (a Bangla search engine), etc. In this paper, a query expansion method has been proposed for Pipilika to select expanded words that will be added with a search key to make the search result more accurate and efficient. Collected documents from corpus data have been ranked using BM25 document ranking method. The method selects additional words from ranked documents that have top syntagmatic relations with the search key. It generates similar words for each additional word and searches keywords using Word2Vec CBOW model which has a more semantic similarity with original words. It adds these similar words with search key and additional words. Finally, an expanded query will be generated. This can help the search engine to capture more user requirements for better search results. The calculated F1 score of the method is 0.74.