An Optimized Comments Generation Model Using Situated Learning Theory

• Main Authors: CHANG, YU-CHIA, 張育嘉
• Other Authors: YANG, SHIH-TING
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/mq2fds

碩士 === 南華大學 === 資訊管理學系 === 107 === 　　Among the current large number of knowledge documents and comments on the issues, most of the documents and comments on the issues are presented in a large amount of text to facilitate the understanding of meanings expressed by documents or comments using the reading of a large amount of text. As a result, providing the public with the summaries of knowledge documents or comments on issues containing a large amount of text for reading enables them to learn of the main ideas of documents and comments on issues in the shortest period of time. Besides, situational elements, such as people, things, time, place, and objects, are usually missing from documents and comments on issues, so the public usually cannot understand the objectives of the posted documents and comments or misunderstand the actual meanings while reading them because the situations and meanings are not clearly expressed. 　　Therefore, this paper develops an Optimized Comments Generation Model using Situated Learning Theory. In this proposed model, firstly, this paper mainly combines the design principles of situated learning improved by CTGV (1992), scenario-based design, 8000 Chinese Vocabulary, and hierarchical clustering to judge situational comments and comment clusters with high similarity. Secondly, this paper uses techniques, such as summary extraction, abstraction of summary, and weighted directed graph, to analyze comment clusters with high similarity, and then combined situational information categories and 5W (Who, What, Where, When, Why) of scenario-based design to judge the presentation of the final situations. Finally, based on the technology proposed in this paper, a Web-based system can be constructed and real-world cases including PIXNET articles and udn News to verify the feasibility of the technology. The verification results show that firstly the recall, accuracy rates and F value are 37%、97% and 54% better than previous research for comments clustering analysis. Secondly, for reading time, reading effect and reading situation interpretation, all the P values are less than α (0.05). Therefore, all three have significant differences to confirm the feasibility of the model.