現今我們處於資訊爆炸的時代，在面臨龐大資料量時，如何有效率地獲取所需資訊是一個非常重要的課題，而資訊檢索 (Information Retrieval) 系統也就成為人們在篩選資料時最常用的工具之一。在相關回饋 (Relevance Feedback) 領域中，Rocchio演算法最廣為人知，該演算法藉由分析相關文件字詞及非相關文件字詞出現頻率，來產生新的查詢字詞，並加入到查詢擴展 (Query Expansion) 集合中，不過Rocchio僅以頻率之角度判斷，並未考量字詞間其他可以利用的資訊。近年來陸續也有語意搜索的研究被提出，概念為發掘字詞間隱含的語意關係，因此，本研究以使用者的原始查詢和查詢結果作為基礎，主要利用神經網路模型Word2Vec來分析原始查詢以及相關回饋中字詞間的語意資訊，並結合共現性分析，萃取出適合的相關字詞來擴展原始查詢字詞集合，使查詢關鍵字能夠更貼近使用者需求。最後透過實驗證明，本研究所提出之方法相較於其他方法能有較佳的檢索效果。;In an era of information explosion, to obtain the information efficiently is a very important issue when faced with huge data volume, and the information retrieval system has become one of the most commonly used tools. In the field of relevance feedback, Rocchio’s query expansion is a well-known method. The algorithm generates new query terms by analyzing the frequency of terms which residing in relevance documents and non-relevance documents. However, Rocchio’s method only focuses on term frequency and ignores information between terms. In recent years, the idea of semantic search is getting more and more popular. Therefore, based on the user′s original query and search results, our research uses Word2Vec which is a neural network model to analyze the semantic information between the original query and the relevance feedback, and combine the co-occurrence analysis to extract the appropriate query expansion terms. The results of experiments verify that the proposed method is effective in document retrieval.
