基于释义语料库的词汇语义引证知识抽取及分析研究
毕业论文关键词：释义语料库；知识抽取；条件随机场；实体识别
Knowledge Extraction and Analysis of Lexical Semantic Citation Based on Interpretation Corpus
Abstract： [Purpose/Significance] "Chinese Dictionary" is the largest Chinese language tool and it’s content is numerous. It is very important to use the dictionary to establish relevant citation databases and carry out related research and analysis. [Method/Process] Firstly, the internal and external information of lexical citation of lexical lexicon is obtained by rule-based extraction method. Then the cleaning and processing of external data descriptive statistical analysis to explore the specific sub-cases. The internal corpus after pretreatment was trained and tested by 10 - fold cross - validation with CRF condition random field. [Result/Conclusion] By analyzing the external information, we can see that the peak period of the development of ancient Chinese literature is the Tang and Song Dynasties, the Ming and Qing dynasties, the number of cited authors and works are concentrated in the four dynasties; by training the internal citation information, can be applied to the dictionary name, Office and other entities to identify the model, the highest accuracy rate of 87%, recall rate of 93%.
Key words: Interpretation corpus；Knowledge extraction；Conditional random field；Entity identification
目 录
摘要 1
关键词 1
Abstract. 1
Key words 1
引言： 2
一、 文献综述 2
二、 释义语料库简介 3
三、 词汇语义引证知识的自动抽取 4
（一）抽取规则 4
1．正则表达式简介 4
2．正则表达式的基本单位 4
（二）实体抽取 4
2．抽取数据再处理 5
1．规则构造与实体抽取 5
四、 词汇语义引证知识的分析 6
（一）引证外部信息分析 6
（二）引证内部信息分析 7
1．CRF模型介绍 7
2．实体识别实验 8
3.引证义项统计分布 11
五、 结语 11
致谢 11
表 1 正则表达式元字符 4
表 2 CRF文件格式 9
表 3 CRF分词后输出结果示例 10
表 4 训练评价 10
表 5 第八次实验特征评价 11
表 6 第九次实验特征评价 11
表 7 实体长度分布 11
图 1 释义语料库语料 4
图 2 朝代分布情况 6
图 3 作者分布情况 7
图 4 书名分布情况 7