使用隱含狄利克雷分佈和知識推理引擎之智慧型健康照護廣告方法

Abstract

隨著健康照護網站的日益興盛，人們開始對替這類網站設計專屬的醫療商品廣告系統產生興趣。模糊而簡略的使用者訊息讓使用者易接收低相關度的廣告，進而造成低點擊率和較差的使用體驗。本論文試圖利用推理引擎和語意分析技術改善廣告系統的效能。推測而得的使用者健康狀況以及訊息和廣告間的隱藏語意關係能豐富已知內容，提升廣告選擇的準確度。在實驗中，本論文提出的方法被證實有效且優於許多其他方法。根據實驗結果可推測，使用領域相關知識擴增內容資訊量能顯著的改善廣告系統的效能。

The growing access to healthcare websites has aroused the interest of designing a specific advertising system focusing on healthcare products. For the purpose of increasing the poor ad-selecting performance resulting from the succinctness and ambiguity of user messages, we introduced semantic analysis and an inference engine into our proposed system. By inferring the illness which users possibly suffer from and bridging the vocabulary gaps between user messages and ads, our method is confirmed to be able to outperform other approaches significantly in precision and coverage. Based on our findings, we conclude that advertising systems could be enhanced by using the domain-specific knowledge to augment the user messages and ads.