Zhongda Xie 15-05-25

来自cslt Wiki
2015年5月31日 (日) 11:21Xiezd讨论 | 贡献的版本

(差异) ←上一版本 | 最后版本 (差异) | 下一版本→ (差异)
跳转至: 导航搜索

Progress in Last Week:

Start for a new work: Detecting key entity in a query.

(e.g. In query "pickup for guitar", the key entity is "pickup".)

Plan: (1)Use web search co-click data to obtain training data, use query pair to determine key entity.

(2)for each query, extract the following information: stemmed word, POS, entity, modifier and so on.

(3)Use CRF model to learn from training data, using features that are extracted in step (2); finally predict key entity in unseen queries.