Abstract: When search logs encounter over sparse and timeliness problems，the quality of extensions cannot be guaranteed by the query expansion methods based on the search log， resulting in the impossibility to meet the needs of timeliness information. A query expansion method based on the search log and local context was introduced. By exploring relationships between user queries and relevant documents in search logs， candidate expansion terms were extracted， and local contexts from related documents were concentrated to extract words with higher frequencies and timeliness. By considering query performances and timeliness of expansions， expansions with higher qualities were provided with this method. Experimental results showed that precisions and recalls could be effectively improved， and better query performance to original queries could be obtained.

Key words: information retrieval, query expansion, search log, local context, query performance