为提高文本中时间信息识别和抽取的效率,提出一种基于CRF (条件随机场)的方法。根据时间信息表现出的一般特点,采用机器学习的方法,通过分析文本中相关词性、短语结构和上下文信息等,提取时间信息的外部特征,采用一种自训练的半监督方法,使用CRF进行识别和抽取。实验结果表明,该方法有效提升了时间识别的性能,在显性时间、隐性时间和总体时间上分别取得了96?25%、88?65%和93?97%的 F1值。
To improve the efficiency of extracting temporal information from the text ,one method based on conditional random fields (CRF) was proposed .According to the general characteristics shown by the temporal information ,the method of machine learning was adopted ,by analyzing a set of linguistic features of time phrases in text such as lexical features ,syntactic features and context information ,while using the semi‐supervised method of self‐training ,temporal information was recognized and ex‐tracted using CRF .Experimental results show a good performance reaching scores of 96?25% ,88?65% and 93?97% for F‐measure to dominant time ,recessive time and full time .