Browse Prior Art Database

Query expansion based on the attention centric content

IP.com Disclosure Number: IPCOM000246224D
Publication Date: 2016-May-17
Document File: 6 page(s) / 56K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a method for enhancing QA systems performance by identifying focus and attention areas of the user in the presented content using eye tracker data analyzer. Method analyzes focus area content in the context of the current search and generates query extensions to enhance the search ranking, and formulates the feedback to the question classifier enhancing query classification model.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 19% of the total text.

Page 01 of 6

Query expansion based on the attention centric content

With increasing volume of available information and data, the information retrieval process becomes more and more challenging.

The process of information discovery most of the time performed iteratively: the search results are presented to the user, retrieved document is browsed by the user, who then performs additional search, etc.

    In the process of the information discovery and searching for the solution or just acquiring knowledge on the particular topic, the user's attention can be focused just on particular segments in the presented documents. However, this attention is not always clearly mapped to the originally searched criteria and keywords. By analyzing the user focus and attention content, the query expansions can be formulated and applied for higher ranking of more relevant to the user attention documents.

    QA systems rely on Question classification trained model. However, question classes or labels are not always accurately reflecting the intent of the user asking that question. This intent can be more predictable if the attention of the users to the particular content segments is detected and analyzed.

    The solution is required to optimize information retrieval process by analyzing the user focus and attention areas identified during his/her content browsing activity and translating this into relevance feedback to search engines, QA systems, information discovery systems, and knowledge repository browsers.

    Proposed here is a method for enhancing QA systems performance by identifying focus and attention areas of the user in the presented content using eye tracker data analyzer. Method analyzes focus area content in the context of the current search and generates query extensions to enhance the search ranking. The method formulates the feedback to the question classifier enhancing query classification model.

    During the process of browsing retrieved documents, the focus and attention of the user can be captured by eye tracker and the content under the focus can be processed for identifying the focus areas. The focus areas content is analyzed, the level of user attention is evaluated, and query expansion parameters are formulated. Feedback on the relevance and importance of the presented content to the user is formulated based on analysis of the user attention. If established, its statistical significance to the feedback is reflected in the QA classifier model.

    Evaluated user's attention toward particular segments is translated into the relevance feedback, improving runtime ranking of the search documents and precision of QA classifiers. The method is based on statistical learning attention model and its application in generating query expansions and question classes or labels. The proposed method produces relevant search query expansion and classification labels by following steps:


identifying the user focused areas


evaluating the intensity or level of the user attent...