結合主題關聯與社群認同之內容推薦方法

Abstract

在現今的網路中，每個人都能成為資訊的提供者或傳播者，內容與知識分散在各種異質網路平台的狀況造成內容的尋找變得困難，然而隨著社群網路服務的興起，其中所包含人與人的互動資訊，都能即時反應使用者觀感以及使用者經驗。社群網路平台中的資料能夠提供更為即時且豐富的互動以及有效的影響力指標。這些資訊能夠提供良好的社群指標索引。本研究提出一結合主題關聯內容過濾與社群認同分析之模型，結合異質網路平台的社群指標以及內容主題關聯性針對內容提供者產生一社群認同指標。透過社群認同指標以及內容主題關聯量測因子則能夠增加不同領域中評估社群網路個體信任的精確度。 實驗結果顯示社群認同指標確能成功提高推薦結果的精確率以及使用者滿意度；同時使用於字義較廣泛之搜尋關鍵字時，社群認同指標的效果更為顯著。

The web evolves into an ecological platform of information as social web, with the increasing user generated content since the apparition of wildly read-write web. Users search content and knowledge in perplexity with the variety of contents and heterogeneous sources. The social networking portals like Twitter, Facebook, LinkedIn etc. are getting popular day by day among users' community by means of social data with people interaction history and personal information can aid in bootstrapping a topic relevant user model. This paper presents a content recommendation model with the integration of topic relevant content filtering and social network analysis. Measure the social endorsement of authors with the topic relevance is useful for estimating the content comprehensibility and also addresses the sparsity and cold start problem in recommendation systems. And experimental evaluation reveals that the social-endorsement-based model outperforms the non-social-endorsement-based model and enhances the performance efficiently.