| 网络文本中文药名实体识别 |
| |
| 引用本文： | 王莉,刘大鹏,佟铁.网络文本中文药名实体识别[J].辽宁科技大学学报,2015(2):127-130,154. |
| |
| 作者姓名： | 王莉 刘大鹏 佟铁 |
| |
| 作者单位： | 1. 辽宁科技大学 电子与信息工程学院,辽宁 鞍山,114051;2. 鞍山市千山风景名胜区管理委员会 信访局,辽宁 鞍山,114045 |
| |
| 基金项目： | 国家自然科学基金(71472081) |
| |
| 摘 要： | 针对中文医药类网络文本的不规范性引起的药名实体识别性能下降,提出基于层次结构的多策略方法。首先使用条件随机场模型结合改进的最大匹配算法识别药名实体,然后在此基础上对其中的不规范药名实体利用最小编辑距离方法规范化药名实体并扩充药名词典。实验结果表明,改进的最大匹配算法结合统计模型有效地提升了药名实体识别的性能,同时为药名实体规范化扩展提供了新的思路。
|
| 关 键 词： | 命名实体识别 条件随机场 最大匹配算法 规范化 |
Chinese drug name entity recognition in web |
| |
| WANG Li;LIU Dapeng;TONG Tie.Chinese drug name entity recognition in web[J].Journal of University of Science and Technology Liaoning,2015(2):127-130,154. |
| |
| Authors: | WANG Li;LIU Dapeng;TONG Tie |
| |
| Institution: | WANG Li;LIU Dapeng;TONG Tie;School of Electronic and Information Engineering,University of Science and Technology Liaoning;Qianshan Mountain Scenic Area Management Committee; |
| |
| Abstract: | Due to drug name entity recognition(DNER)performance decreased by non-normative web in Chi-nese pharmaceutical website,the hybrid approach over hierarchical structure is proposed. The conditional ran-dom fields(CRFs)model is combined with improved maximum matching(IMM)algorithm to recognize the drug name entities and then the Levenshtein distance is used to normalize the drug name entities and expand the drug dictionary. The results show that IMM combined with statistical algorithm effectively improves the performance of DNER,and it provids a new way for drug name entities standardized extension. |
| |
| Keywords: | name entity recognition conditional random field maximum matching algorithm normalization |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
