A Study of Named Entity Recognition Based on Syntax Rule

• Main Authors: Chang-yu Yang, 楊長諭
• Other Authors: Chuen-min Huang
• Format: Others
• Language: zh-TW
• Published: 2013
• Online Access: http://ndltd.ncl.edu.tw/handle/18147596038192273740

碩士 === 國立雲林科技大學 === 資訊管理系碩士班 === 101 === As people gain more and more news on the Internet, news information needs to be processed and presented effectively. By the extraction technique of the features, it reduces the reading time of readers and enhances the benefits of knowledge acquisition. In the past, information extraction technique was used to name entity recognition, and it could extract the names of people, location, and organization from news articles. However, new named entities are continually created. If the corpus includes the rapidly generated named entities, it requires high artificial maintenance costs. The Syntax Rule adapts rules pairing mode. When the sentence matches the Syntax Rule, its features will be analyzed and extracted from the news articles. Applying Syntax Rule doesn’t consider the corpus covered rate. Therefore, this study applied the Syntax Rule in the news articles. Furthermore, we used Wikipedia’s open categories and infoboxes to characterize and examined the named entity to confirm the feature terms were precisely extracted. We used MET-2 Chinese data sets as the experiment data sets to evaluate the performance of experiments. In our overall system evaluation, the precision reaches to 86.32%, recall reaches to 75.65%, and F-measure reaches to 80.4%. It sufficiently shows that our extraction named entities recognition has a great performance.