Browse Prior Art Database

Method and System for User Intent Modeling and Intent summary Extraction Disclosure Number: IPCOM000213549D
Publication Date: 2011-Dec-21
Document File: 6 page(s) / 38K

Publishing Venue

The Prior Art Database

Related People

Benoit Dumoulin: INVENTOR [+4]


A method and system for User Intent modeling and Intent summary Extraction is disclosed.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 16% of the total text.

Method and System for User Intent Modeling and Intent summary Extraction


A method and system for User Intent modeling and Intent summary Extraction is disclosed. 


Disclosed is a method and system for user intent modeling and intent summary extraction.  Users interact with a search Engine issuing queries to get desired results.  Since search activity is performed due to lack of information about a particular topic, users might not have an adequate background that would enable them to express or clarify their needs.

The method disclosed herein captures signals that differentiate the users who are able to articulate their intents or information needed from users who have only a vague idea of their requirements.  In this method, three sets of different features are used.  The set includes features that are already used in literature and some features that are created on purpose, to train a set of automatic classifiers that are able to understand if the user was able to articulate user’s needs or intents.

The method automatically determines whether users are able to articulate their intent by their queries.  Thus, the method infers what would best summarize the intent of a user.  The search could be successful even for the cases where the intent is not clearly expressed by the user in terms of a query.  Consider, for example, the following sequence of queries: “Jackson president money → Jackson currency”.  The intent of the user is to get information about the currency reform of President Andrew Jackson.  This intent can be met by a few long clicks, as confirmed by the editors.  However, the intent is not clearly articulated in any of the queries.  Explaining by way of another example, a query: “toxicity → toxicity level high → high toxicity level”.  Initially, the query indicates that there is a central theme since the user is consistent with queries.  In this query, the intent is not explicitly stated and the context is not obvious.  For example, there is lack of clarity on an element to which the toxicity level is related.  Hence, after considering the content of clicked pages, the intent can be summarized by “lithium toxicity treatment for bipolar disorder”.  While the users are satisfied their goals in both cases, the intent was not clear from the submitted queries.  The above examples are cases where users were not able to completely articulate their needs due to various reasons.  For example, the users have a vague idea of what they are looking for or how to express it.  Despite these counter-examples, users often explicitly state their intents.  Consider the goal “dodge diesel → dodge diesel performance → dodge forum”.  Although the first and the last queries are quite general, the analysis of the entire sequence of queries and user actions (e.g. clicks) demonstrates that the user wants to collect information about the performance of “dodge diesel”.  Hence, the...