Combining Query Ambiguity and Query-URL Strength for Log-Based Query Suggestion

摘要

The ambiguity of query may have potential impact on the performance of Query Suggestion. For getting better candidates adapting to query's ambiguity, we propose an efficient log-based Query Suggestion method. Firstly we construct a Query-URL graph from logs and calculate the bidirectional transition probabilities between queries and URLs. Then, by taking URL's rank and order into consideration, we make a strength metric of the Query-URL edge. Besides, we conduct random walk with the edge strength and transition probability to measure the closeness among queries. To reflect the influence of query ambiguity, we exploit an entropy-based method to calculate the entropy of each query as a quantitative indicator for ambiguity, making a notion of ambiguity similarity as an available factor in relevance estimation. Finally we incorporate ambiguity similarity with closeness to derive a comprehensive relevance measurement. Experimental results show that our approach can achieve a good effect.