Soft-constrained inference for Named Entity Recognition
作者：
Highlights：
• Named Entity Recognition is addressed by constraining inference in CRF.
• An two phases integer linear programming approach is proposed.
• Complex relationships among labels are automatically extracted from data.
• Extracted relationships are introduced as soft constraints in the ILP formulation.
• The proposed method significantly outperforms the state of the art approach.
摘要
•Named Entity Recognition is addressed by constraining inference in CRF.•An two phases integer linear programming approach is proposed.•Complex relationships among labels are automatically extracted from data.•Extracted relationships are introduced as soft constraints in the ILP formulation.•The proposed method significantly outperforms the state of the art approach.
论文关键词：Conditional Random Fields,Named Entity Recognition,Rule extraction,Integer linear programming
论文评审过程：Received 30 April 2013, Revised 14 April 2014, Accepted 22 April 2014, Available online 17 May 2014.
论文官网地址：https://doi.org/10.1016/j.ipm.2014.04.005
