Relevance Feedback

The relevance feedback method is applicable, a sub-discipline of computer science and information science in information retrieval. It describes a method for gradually extending improving search results of a text-based search ( eg search engines).

The basic idea

The idea is to use the relevance already Documents found for the search of similar documents. Hence the word relevance feedback, because it describes an information feedback on the outcome of previous searches. The relevance feedback, building on existing search method on ( Probabilistic model or vector space model).

Method

• In the vector space model, a new query Q ', whose vector is the relevant documents and irrelevant documents similar dissimilar than Q,
• In the probabilistic model new conditional probabilities that represent the relationship between the occurrence of terms in Indexat and relevance assessment.

So the result is improved step by step.

Motivation

A problem for today's search engines are often too short queries. On average, these go much beyond 2 words. This leads to a high number of ambiguous requests. Another problem are inaccuracies and / or blurring in the content analysis of the documents. Special motivation wins the relevance feedback from the Multimedia Information Retrieval, since often a media break is to be overcome between query and document by approximately a textually defined search query is applied to image data. The relevance feedback helps to alleviate these problems, since the system can build increasingly better hypotheses about the requirements that the user provides to relevant documents.

Disadvantages

A disadvantage of relevance feedback is included in the costs, the demanding on repeated assessments of relevance to the user.

Blind Relevance Feedback

The Blind Relevance Feedback ( or also known as pseudo- relevance feedback ) overcomes these disadvantages of manual relevance feedback, but brings other disadvantages. The relevance of the search is not marked manually by the user, but automatically (hence the name "Blind "). The search system automatically associates a relevance for the respective result documents, whereupon the search query by query expansion is automatically expanded, and a new result with the expanded query is generated. In this method, no more manual intervention is necessary for the user, the results are often inaccurate.