论文部分内容阅读
军事命名实体(Military Named Entities,MNEs)内部嵌套关系复杂、语法区分不明显,从而影响实体识别效果,针对这一问题,提出了一种小粒度策略下基于条件随机场(Conditional Random Fields,CRFs)的MNEs识别方法。运用小粒度策略,结合手工构建的MNEs标注语料进行建模,采用CRFs模型识别出不可再分的小粒度MNEs,再通过对小粒度MNEs进行组合得到完整的MNEs。最后,通过实验对该方法进行了验证,结果表明:在作战文书语料的开放测试中,MNEs识别的召回率达到72%以上,准确率达到85%以上。
In order to solve this problem, military internal identities (Military Named Entities, MNEs) have complex internal nested relations and non-obvious grammatical distinction. A small granularity-based conditional random field (CRFs) ) MNEs recognition method. A small particle size strategy was used in combination with the constructed MNEs annotation corpus. The CRFs model was used to identify the small MNEs that could not be subdivided, and the complete MNEs were obtained by combining the small MNEs. Finally, the method is verified through experiments. The results show that the MNEs recognition recall rate is above 72% and the accuracy rate is more than 85% in the open test of combat documents corpus.