Code Search Based on Alteration Intent
Code Search Based on Alteration Intent
复制标题
DOI：
10.1109/access.2019.2913560
复制
发表时间：
2019
期刊：
影响因子：
3.9
通讯作者：
Huaiguang Wu;Yang Yang-Yang
中科院分区：
文献类型：
--
作者：
Huaiguang Wu;Yang Yang-Yang
代码检索是根据用户的需要进行检索的方法。许多人认为查询是唯一反映用户需求的查询。因此，他们从查询开始，查找与查询词语义相似的同义词，以进行查询扩展。然而，他们忽略了一个事实：检索到的方法仍然需要修改，因为它不能直接满足用户的需求。这意味着在检索方法中所做的更改也反映了用户的需求。在本文中，我们开始与变更意图（检索方法后可能发生的变更），并提出预测变更意图，并将其用于查询扩展。实验结果表明，我们的方法优于CodeHow，使用API进行查询扩展的方法，由59.8%的精度得分为0.815。
Code search is to retrieve the method according to the user needs. Many people think of the query as the only one to reflect the user needs. So they start with the query and find the synonyms, semantically similar to the query terms, for query expansion. However, they overlook a fact: the retrieved methods still need to be altered because it does not meet the user needs directly. This implies that the alteration made in the retrieved methods also reflects the user needs. In this paper, we start with the alteration intent (the possible alterations after retrieving methods) and propose to predict the alteration intent and use it for query expansion. The experiment results show that our approach outperforms CodeHow, the approach to use APIs for query expansion, by 59.8% with a precision score of 0.815.