Department of Information Management , Shu-Te UniversityHybrid semantic roles and KNN approach to emotion word expansion and quantization for monitoring blogs

• Main Authors: Po-Yu Tsui, 崔博喻
• Other Authors: 陳璽煌
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/79714868438055041692

碩士 === 樹德科技大學 === 資訊管理系碩士班 === 100 === Campus was previously regarded as a simpler environment. It was also a territory with relatively more autonomy and more academic freedom. However, with the change of social environment, a student''s growing process is now suffering pressures from academic requirements, physical changes, inter-personal relationship, external temptation and independence from parents. These pressures often leave students into lingering in the midst of helplessness. A national college health evaluation conducted by the American College Health Association in 2006 indicated that 94% of students suffered from tremendous pressure because of things they were requested to do, 44% of students said they once felt very depressed and hard to lead a normal life and 1.3% of students said they thought about killing themselves. According to the annual campus incident statistics for various grades of schools conducted by the National Ministry of Education each year, the number of incidents reported hits record high each year. Take case analysis for student suicide and self-inflicted injuries in 2009 school year for example, there were 107 such cases for college, 132 for senior high school, 78 for junior high school and 16 for primary school. Additionally, there were a total of 8,792 reported cases of child abuse, violent deviance behavior and discipline conflict on campus. This has created a cumbersome burden to the society as a whole. With the popularization of internet on campus and at home in recent years, many students have their own blogs and facebook accounts. They will post their diaries documenting their mood on related platforms from time to time. Some of these articles posted described these students'' mood and their views on certain incidents. Accordingly, this study tries to design and develop a system to quantify and assess depressed mood. Through RSS, we''re able to subscribe characteristics of these articles posted on blogs and collect articles written by students. With the utilization of related techniques on natural language, we can estimate the degree of depression from these articles and point out, through trend lines, recent degree of depression on each article. These will serve as counseling psychologist''s references during counseling in order to reduce accident rates for students with the tendency of depression. This system includes setting up of a database for students'' articles depicting their mood on blogs, assessment expansion system for degrees of depressed mood and a quantify system on the depression degrees of articles. Students are able to write articles on their blogs and the system will regularly collect, through RSS, their articles on the blogs. Natural language analysis will be conducted using article depression quantifying system. With this, emotional terms and their responding linguistic characters on each paragraph will be obtained. An assessment expansion system for degrees of depressed mood will then be established using K-Nearest Neighbor Classifier method. Finally, a depression mood degree quantify system will be built up based on depression mood degrees and analysis on linguistic characters. According to the result of this study, for the group originally with the same reference terminology database depression degree, the emotional terminology depression degree closest to the center point was designated as the emotion terminology depression degree for the whole group. However, after the expansion on K-Nearest Neighbor Classifier method, the emotional terminology database re-designated new depression degrees onto respective emotional terminologies and effectively rectified negative emotional terminology depression degrees embedded in positive emotional terminologies. In the meantime, through analysis over Semantic Roles, emotion detection terminologies from non-writers were excluded accordingly. This has provided more accurate depression quantified degrees. Counseling psychologists will be able to monitor students'' recent emotional ups and downs using systematic depression quantifying trend line charts. This will ensure effective processing of related counseling operation and prevent the occurrence of student accidents and injuries.