CAJ | 학술논문

与基于词典和基于规则的识别方法相比,统计机器学习方法更加适合被应用到命名实体的识别工作中来。本文主要在中文化学物质名称的识别工作中,考察两类统计机器学习模型识别效果及识别效率的优劣,实验结果表明,在所取训练语料与测试语料相同的情况下,以CRF模型为代表的条件概率模型可以展现出更好的实验性能。
여기우사전화기우규칙적식별방법상비,통계궤기학습방법경가괄합피응용도명명실체적식별공작중래。본문주요재중문화학물질명칭적식별공작중,고찰량류통계궤기학습모형식별효과급식별효솔적우렬,실험결과표명,재소취훈련어료여측시어료상동적정황하,이CRF모형위대표적조건개솔모형가이전현출경호적실험성능。
Comparing with the recognition methods based on dictionary or rule,the method based on machine learning is suitably to be applied to the research on NER（named entity recognition）.This article mainly evaluated the performance of two kinds of machine learning methods SVM and CRF in the course of recognizing Chinese chemical substance names,and the result of the experiment showed：in the condition of selecting same training sample and testing sample,the conditional models（Take the CRF for example）reveal the better performance.