Learning User Profiles Research Articles

부정적 피드백과 강화학습을 이용한 사용자 프로파일 학습

정보추천은 동적인 사용자의 정보요구를 기준으로 문서들을 선별하여 제시한다. 사용자의 정보요구는 하나 이상의 단어들로 구성된 프로파일로 표현이 되며, 문서의 추천 과정 동안에 발생하는 사용자의 관련성 평가를 통해 구체적인 내용으로 변할 수 있다. 기존 연구의 경우 사용자는 자신이 직접 관련성 평가에 참여하여 평가 정보를 입력하고, 사용자가 평가한 긍정적 피드백 정보를 이용하여 사용자 프로파일을 학습한다. 본 연구는 사용자가 평가한 긍정적 피드백, 부정적 피드백 정보와 강화학습을 이용한 사용자 프로파일의 학습 방법을 제안한다. 제안된 방법과 기존의 긍정적 평가 정보만을 이용한 방법의 성능을 측정하기 위해 동일한 토픽에 대한 추천 성능을 비교 실험하였다. 실험한 결과 제안된 방식에 의한 추천성능이 긍정적 평가 정보만을 이용한 피드백 방법보다 나은 성능을 보였다. The information recommendation system offers selected documents according to information needs of dynamic users. User's needs are expressed as profiles consisting of one or more words and may be changed into some specifics through relevance feedback made by users during the recommendation process. In previous research, users have entered relevance information by taking part in explicit relevance feedbacks and learned user profiles using the positive relevance feedbacks. In this paper, we learn user profiles using not only positive relevance feedback but negative relevance feedback and reinforcement learning. To compare the proposed with previous method, we performed experiments to evaluate recommendation performance of the same topic. As a result, the former shows the improved performance than the latter does.

Bayesian graphical models for adaptive filtering

No abstract available.

Personalized web search for improving retrieval effectiveness

Current Web search engines are built to serve all users, independent of the special needs of any individual user. Personalization of Web search is to carry out retrieval for each user incorporating his/her interests. We propose a novel technique to learn user profiles from users' search histories. The user profiles are then used to improve retrieval effectiveness in Web search. A user profile and a general profile are learned from the user's search history and a category hierarchy, respectively. These two profiles are combined to map a user query into a set of categories which represent the user's search intention and serve as a context to disambiguate the words in the user's query. Web search is conducted based on both the user query and the set of categories. Several profile learning and category mapping algorithms and a fusion algorithm are provided and evaluated. Experimental results indicate that our technique to personalize Web search is both effective and efficient.

Learning for automatic personalization in a semantic taxonomy-based meta-search agent

Providing highly relevant page hits to the user is a major concern in Web search. To accomplish this goal, the user must be allowed to express his intent precisely. Secondly, page hit rating mechanisms should be used that take the user’s intent into account. Finally, a learning mechanism is needed that captures a user’s preferences in his Web search, even when those preferences are changing dynamically. To address the first two issues, we propose a semantic taxonomy-based meta-search agent approach that incorporates the user’s taxonomic search intent. It also addresses relevancy improvement issues of the resulting page hits by using user’s search intent and preference-based rating. To provide a learning mechanism, we first propose a connectionist model-based user profile representation approach, which can leverage all of the features of the semantic taxonomy-based information retrieval approach. A user profile learning algorithm is also devised for our proposed user profile representation framework by significantly modifying and extending a typical neural network learning algorithm. Finally, the entire methodology including this learning mechanism is implemented in an agent-based system, WebSifter II. Empirical results of learning performance are also discussed.

Personalized web-document filtering using reinforcement learning

Document filtering is increasingly deployed in Web environments to reduce information overload of users. We formulate online information filtering as a reinforcement learning problem, i.e., TD(0). The goal is to learn user profiles that best represent information needs and thus maximize the expected value of user relevance feedback. A method is then presented that acquires reinforcement signals automatically by estimating user's implicit feedback from direct observations of browsing behaviors. This "learning by observation" approach is contrasted with conventional relevance feedback methods which require explicit user feedbacks. Field tests have been performed that involved 10 users reading a total of 18,750 HTML documents during 45 days. Compared to the existing document filtering techniques, the proposed learning method showed superior performance in information quality and adaptation speed to user preferences in online filtering.