利用社群網路進行問題答覆者的搜尋研究

Abstract

利用社群網路進行問題答覆者的搜尋研究 近年來隨著行動裝置的普及和網路技術的成長，使得社群網路平台漸成為人際交流的重要工具。它不僅有助於朋友間或團體間的訊息分享，也往往提供快速的資訊需求解答。目前已知的社群問答系統例如Facebook Questions和Aardvark，均缺乏分析使用者互動訊息以致無法評估答覆者的信任度，而傳統建構在搜尋引擎上的問答系統對於解決較複雜的資訊需求在尋找答案的過程又相當費時。 有鑑於此，本計劃擬研製一個以社群網路為基礎的問題答覆者搜尋系統。我們將利用社群網路上的資訊豐富和即時特性，開發相關的篇章處理和意見分析技術來處理使用者互動內容，探勘他們的人際關係類型，進而加速答案取得的過程。本系統預計兩年完成。在第一年裡，我們將建立一個能夠隨著時間與內容變化的異質性社群網路，並在其上進行社群關係辨識。主要研究工作包含社群網路資料擷取、使用者互動訊息的意見分析、和社群間的關係辨識及強度計算。在第二年裡，我們將進行問題答覆者搜尋系統的建置，包括問題的主題辨識、答覆者的領域辨識和能力評估、社群成員的信任度計算、提問者與可能答覆者的信任度計算、以及系統成效評估。我們相信本計畫的執行不僅有助於社群網路關係的辨識和探討，也能對社群網路使用者的資訊需求提供一個有效且快速的服務機置。

A Question Answerer Search System Through Social Networks With the rapid growth of network technologies and popularity of mobile devices, social networks become important media for today’s social interaction. They not only facilitate message sharing among friends and groups, but also provide possible answers in response to users’ questions. For example, Facebook Questions and Aardvark are well-known question answering system built up on the basis of social networks. However, it is hard to justify the answers posted by the network users if their interaction content is not employed and analyzed. On the other hand, it might also be challenging for a search-engine based question answering system to resolve complex questions if user’s queries are not well formulated. Hence this project is proposed with the aim to facilitate the question answering process by employing the rich and real-time information available on social networks. We will explore novel discourse processing and opinion analysis methods applicable for interaction contents analysis and the relationship identification among system users. The proposed project will be implemented in two years. In the first year, we will build up a heterogeneous social network in which user’s and group’s information are recorded and updated as time passes. The main tasks will include information extraction from present social networks, opinion analysis on the extracted interaction contents, social relationship identification and its strength computation. In the second year, we will implement the proposed answerer search system. The kernel tasks will include question processing and its topic identification, answerer’s expertise identification and answering capability evaluation, the trustee computation among network users, candidate answerer ranking and the system evaluation. It is believed that the implementation of the proposed system will not only enhance the investigation of social networks but also provide an efficient and effective information service in response to the user’s information needs.