實驗室成員‎ > ‎

鍾澤華

鍾澤華 Tse-HuaChung

碩士論文 (2010)

數位合作學習之自動化人格辨識研究
On Automatic Personality Pattern Recognition for Collaborative e-Learning

關鍵字 Keywords

數位合作式學習, 人格辨識, 說話風格擷取, 團隊組成
Collaborative e-Learning, Personality Classification, Speaking Styles Extraction, Team Formation

摘要

近年來,合作式學習已被廣泛採用並有不錯之成效。隨著網際網路技術之普及與電腦科技之進步,網路式合作學習(Collaborative e-Learning)也因應而生。數位合作式學習團隊成員在團隊中所擔任之角色常直接或間接影響整體團隊之學習成效。過去,透過問卷了解參與者人格特質,再安排其角色;然而,填寫問卷過程耗時,對分散且需頻繁討論的學習團隊來說,如何即時、有效率的辨識學習成員之人格特質,並依照其個性賦予適當之角色,為當前數位合作學習之重要課題。 本研究研發一數位合作學習之自動化人格辨識研究機制,透過分析數位合作式學習成員之對話內容找到其對應之人格類型(NEO-FFI),以利合作式團隊學習時角色適性化之安排。本研究首先建立數位合作式學習平台,以問卷方式取得學習者人格類型及對應談話內容,並以此資料作為人格分類器訓練資料;再將談話內容轉換成三種不同類型詞性,利用搭配詞(collocation)之概念找出人格類型在三種不同詞性組合下對應的說話風格,以此作為人格類型分類之特徵(feature);接著以term frequency–inverse document frequency作為特徵篩選方式並以支持向量機(Support Vector Machine,SVM)作為訓練人格分類器之方法,最後以此分類器辨識判斷學習者之人格類型,以利安排適合各學習者特質之團隊角色。根據本研究之分類器結果,在以上位詞為基礎之人格分類器下,整體分類準確度可達86%。

Abstract

Over the past few decades, collaborative learning has been widely utilized in learning with favorable and positive results. With the advanced Internet technology, collaborative e-Learning has become a trend of learning. Collaborative learning involves forming “learning teams.” However, traditional learning team formation based on time-consuming personality tests seems not practical for collaborative eLearning. Figuring out a method to effectively identify team members’ personality traits for role assignments is a challenge in collaborative e-Learning. The purpose of this study is to develop an Automatic Personality Pattern Recognition Mechanism (APPR) where, through the analysis of each learner’s dialogue, certain type of personality (NEO-FFI) is identified for role assignments in forming a learning team. Firstly, a collaborative e-learning platform was developed in which personality types with corresponding dialogue transcripts were collected. Secondly, these dialogue transcripts were converted into three different types of parts-of-speech, with which the concept of collocation was used to find the difference in how these parts-of-speech were used by people of different personality types. By them, personality types were categorized. Thirdly, term frequency–inverse document frequency was used as a screening method for features extraction and personality types were classified through SVM (Support Vector Machine) as a training stage. Lastly, using this personality identification mechanism, collaborative e-learning teams were formed with each individual learner assigned a role in a respective group in accord with his/her personality type. Empirical results showed that proposed APPR can identify personality types with overall classification accuracy up to 86%.


Comments