A Hybrid Model of Query Expansion using Word2Vec

Query expansion method is one of the most popular methods to reduce the vocabulary mismatch in Information retrieval tasks. Traditional methods of query expansion that use Pseudo relevance feedback are not much efficient for document retrieval from a large collection of documents. In the proposed method we will try to minimize the vocabulary mismatch using a hybrid method of query expansion using the word embedding technique. The proposed method uses both Word2Vec and a local method to predict the expansion terms. The mean average precision of the proposed method is 0.2992. The proposed model also compares with the original query and BM25 model.