Browse Prior Art Database

A Method and System for Retrieving Geo-Location of Media Content with a Hierarchical Divide and Conquer Approach

IP.com Disclosure Number: IPCOM000238389D
Publication Date: 2014-Aug-22

Publishing Venue

The IP.com Prior Art Database

Related People

Michele Trevisiol: INVENTOR [+3]

Abstract

A method and system is disclosed for retrieving geo-location of media content with a hierarchical divide and conquer approach. The method and system utilizes the hierarchical divide and conquer approach to identify the geographic location that is associated with a set of tags of the media content. The method and system pre-selects one or several geographic areas of interest of higher expected relevance and performs a deeper analysis inside the selected areas to return coordinates that are most likely related to input tags.

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 13% of the total text.

A Method and System for Retrieving Geo-Location of Media Content with a Hierarchical Divide and Conquer Approach

Abstract

A method and system is disclosed for retrieving geo-location of media content with a hierarchical divide and conquer approach.  The method and system utilizes the hierarchical divide and conquer approach to identify the geographic location that is associated with a set of tags of the media content.  The method and system pre-selects one or several geographic areas of interest of higher expected relevance and performs a deeper analysis inside the selected areas to return coordinates that are most likely related to input tags.

Description

Disclosed is a method and system for retrieving geo-location of media content with a hierarchical divide and conquer approach.

The method and system estimates confidence scores for location predictions for the media content.  The method and system combines evidence provided by different hierarchical layers in a model using Dempster and Shafer’s theory of evidence [1,2].  These confidence scores can then be used to automatically accept or reject geotags under consideration.

The method and system provides an efficient and effective geotagging system without assuming any prior knowledge about a city or a country the media content was captured.  Moreover, the method and system does not require any external information such as gazetteers.  The key contribution is the tag analysis technique introduced to extract clues about a location.  The method and system identifies the tags that are likely to convey some geographical information and discards the tags that are deemed irrelevant.

After filtering out noisy tags in a pre-processing step, the method and system proposes two different tag processing techniques, both scalable and not limited to tags that are location names.  The first technique implements a text matching rule and serves as a baseline.  The second approach is a radically different tag analysis technique.  Based on a divide and conquer strategy, the relationship between tags and coordinates is analyzed to better reflect the informativeness of tags with respect to the location.

A new joint tag or location technique is one of the main contributions of the method and system.  Moreover, the method and system shows the usefulness of different strategies borrowed from other contexts.  In addition, the method and system estimates confidence of the predictions made by the hierarchical approach.  As different levels of granularity are available to use in the predictions of the geotags, there is a trade-off between accuracy and informativeness.  In particular, the method and system proposes an approach to deal with probabilistic information at different levels of granularity in a natural way.  Moreover, a corresponding theory dictates how evidence coming from different sources, in this case models of areas at different granularity levels can be combined.

The method and system pr...