Combining Vector Space Model and Word2Vec: a preliminary study

This paper investigates traditional Information Retrieval (IR) methods for syntax evaluation of documents. The main aim of this study is to search for a method to combine IR models with new studies in Natural Language Processing (NLP), specifically with Word2Vec model. These techniques provide a semantic distributed representation of terms which can be used to improve retrieval and documents ranking. The present work focus on the selection of words to be used for documents retrieval. A further point of interest is the research of a suitable weighting scheme to be applied for the evaluation of additional information, provided by the use of Word2Vec word embeddings