Skip to content

chokchai/Expertise-Ranking-for-Online-Social-Network-Group-Members

Repository files navigation

การจัดลาดับความเชี่ยวชาญสาหรับสมาชิกกลุ่มเครือข่ายสังคมออนไลน์

เครือข่ายสังคมออนไลน์ได้กลายเป็นช่องทางหลักที่ผู้คนนิยมใช้สร้างความสัมพันธ์ แสดงความเห็น รวมถึงแลกเปลี่ยนข้อมูลข่าวสาร กลุ่มเครือข่ายสังคมถูกสร้างขึ้นมาเป็นชุมชนสาหรับกลุ่มคนที่มีความสนใจในเรื่องใดเรื่องหนึ่งร่วมกัน รูปแบบหนึ่งของปฏิสัมพันธ์ภายในกลุ่มเครือข่ายสังคมคือการถามตอบ โดยที่สมาชิกในกลุ่มสามารถตั้งคาถามหรือตอบคาถามของผู้อื่นได้ วิธีการวิเคราะห์เครือข่าย เช่น ขั้นตอนวิธีแบบเพจแรงค์ สามารถนามาใช้ในการวิเคราะห์รูปแบบความสัมพันธ์ระหว่างสมาชิกในกลุ่ม เพื่อระบุตัวผู้เชี่ยวชาญและจัดลาดับความเชี่ยวชาญของสมาชิกในกลุ่ม งานวิจัยนี้มุ่งเน้นไปที่การทดลองเพื่อหาว่าคุณภาพของความเห็นซึ่งเป็นคาตอบของคาถามนั้น หากนามาใช้ร่วมกับขั้นตอนวิธีแบบเพจแรงค์ จะมีผลต่อการระบุตัวและจัดลาดับผู้เชี่ยวชาญเพียงใด ปัจจัยทางคุณภาพของความเห็นที่พิจารณาได้แก่ ค่าความนิยมที่สมาชิกอื่นกาหนดให้กับความเห็น และคุณภาพของเนื้อหาซึ่งประกอบด้วย ความยาว ความซับซ้อน และความมีเนื้อหาสาระ งานวิจัยนี้ทาการทดลองโดยใช้ข้อมูลจากกลุ่มเฟซบุ๊กจาวา และประเมินความแม่นของการจัดลาดับความเชี่ยวชาญทั้งกรณีที่พิจารณาและไม่พิจารณาคุณภาพของความเห็น โดยเปรียบเทียบกับการจัดลาดับโดยผู้เชี่ยวชาญภาษาจาวา ผลการทดลองพบว่าคุณภาพของคาตอบแม้จะยังไม่ส่งผลเด่นชัดต่อการจัดลาดับความเชี่ยวชาญเนื่องด้วยข้อจากัดของข้อมูลทดสอบและการประเมินผล แต่มีแนวโน้มที่จะส่งผลดีต่อการจัดลาดับความเชี่ยวชาญ

โดย นายโชคชัย พุฒตาล
อาจารย์ที่ปรึกษาวิทยานิพนธ์หลัก รองศาสตราจารย์ ดร.ทวิตีย์ เสนีวงศ์ ณ อยุธยา
วิทยานิพนธ์นี้เป็นส่วนหนึ่งของการศึกษาตามหลักสูตรปริญญาวิทยาศาสตรมหาบัณฑิต
สาขาวิชาวิศวกรรมซอฟต์แวร์ ภาควิชาวิศวกรรมคอมพิวเตอร์
คณะวิศวกรรมศาสตร์ จุฬาลงกรณ์มหาวิทยาลัย
ปีการศึกษา 2555
ลิขสิทธิ์ของจุฬาลงกรณ์มหาวิทยาลัย

EXPERTISE RANKING FOR ONLINE SOCIAL NETWORK GROUP MEMBERS

Online social networks have become the major channel for people to maintain relationships, collaborate, and contribute shared information. A social network group can be created as a specific community for people who share an interest in a particular topic. One form of interaction within the group is the question-answer interaction by which the users in the group can ask questions or provide answers to others. A network analysis method, e.g., PageRank, can be used to analyze the interaction patterns between the users in order to identify and rank experts in the group. In this research, we are interested in experimenting on how the quality of the users’ comments, i.e., the answers to the questions, can take part in the identification and ranking of experts by a PageRank-like algorithm. The quality factors are community rating, that is given to the comments, and the content-based features of the comments, i.e., length, complexity, and informativeness. We conduct an experiment on a Java Facebook group and evaluate the accuracy of the ranking with and without comment quality consideration against expertise ranking by Java experts. Even though the experimental result does not clearly show the influence of comment quality on expertise ranking due to several limitations relating to the experimental data and evaluation procedure, there is a positive tendency that comment quality can benefit expertise ranking.

By Mr.Chokchai Puttan
Advisor Assoc. Prof. Twittie Senivongse
A Thesis Submitted in Partial Fulfillment of the Requirements
for the Degree of Master of Science Program in Software Engineering
Department of Computer Engineering
Faculty of Engineering
Chulalongkorn University
Academic Year 2012
Copyright of Chulalongkorn University


เครื่องมือที่ใช้ในการพัฒนา

  • ซับไลม์ 2 (Sublime 2)
  • ชุดเครื่องมือพัฒนาโหนดเจเอส 10.1 (NodeJS 10.1)
  • โหนดดีบี (Node-DB)
  • ฐานข้อมูลมายเอสคิวแอล 5.5.24 (MySQL Database 5.5.24)

ส่วนหนึ่งของวิทยานิพนธ์นี้ได้รับการตีพิมพ์เป็นบทความวิชาการในหัวข้อเรื่อง “Expertise Ranking in Question-Answer Social Network Groups” by Chokchai Puttan and Twittie Senivongse in The 10th International Joint Conference on Computer Science and Software Engineering (JCSSE'13), organized by Mahasarakam University, Thailand, 29-31 May 2013

About

Thesis @ Chula University

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published