Dieses Dokuwiki verwendet ein von Anymorphic Webdesign erstelltes Thema.

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

visual_analytics_of_classified_ads [2014/11/28 08:59]
melancon created
visual_analytics_of_classified_ads [2014/11/28 09:30] (current)
melancon
Line 1: Line 1:
 ====== Master'​s degree project ====== ​ ====== Master'​s degree project ====== ​
  
-===== Visual ​Search Engine ​=====+===== Visual ​Analytics of Classified Ads =====
  
 Application deadline: (a soon as possible) Application deadline: (a soon as possible)
  
-Host organization ​: [[http://​www.labri.fr|CNRS LaBRI UMR 5800]], Models & Algorithms for Bio-informatics and Visualization team+Host organizations ​: 
 + 
 +  * [[http://​www.labri.fr|CNRS LaBRI UMR 5800]], Models & Algorithms for Bio-informatics and Visualization team 
 +  * [[http://​comptrasec.u-bordeaux.fr|Centre de droit comparé du travail et de la sécurité sociale (COMPTRASEC)]]
  
 ==== Want to know more? ==== ==== Want to know more? ====
Line 22: Line 25:
 ==== Project abstract ==== ==== Project abstract ====
  
-Search engines usually provide ​textual ​and ranked list of pages corresponding ​to a user query most usually expressed as series ​of keywords. This search scenario suffers from various difficulties ​and efforts have been devoted ​to the design ​of interfaces supporting intelligent search, expanding queries and/or classifying query results.+Classified ads hide turbulent activity to try and capture customers. Although ads originally are posted by individual users to hopefully find an interested reader, sellers organize into networks and implement ​planned strategy. By looking at ads, we may hope to compare their content, build network ​of cross-reference ​and hopefully read these strategies ultimately leading ​to the identification ​of the underlying organizations.
  
-Because queries are answered by providing ​series of web pagesthe answer ​offered ​to the user may be thought of as a set of documents each being indexed by words that relate to the initial query.+All these cross-referenced ingredients naturally builds into layered network mixing different semantic entities such as places (towns)telephone numbers, services or goods that are being offered, pictures. Ads may compare because sellers use the exact same picture for different goods, a same phone number for a set of ads may indicate ​the presence of an organization (instead of individual sellers), etc.
  
-Renoust, Melançon and Viaud had proposed a methodology to analyze document corpus based on how keywords interact through documents, looking at things from the perspective of multiplex networks. +The goal of this Master'​s degree project is to first implement ​piece of work leveraging ​the content ​of the ads into multilayer network. The layered network model needs to be both robust (so we can deploy and collect ads from different classified ads websites) and flexible so it can usefully support ​analysis. ​The second phase of the project is to try and build visualizations of these networks (most probably sub-networks obtained by extracting useful views), using multiple ​network ​views and allowing ​to simultaneously ​query/filter operations.
- +
-The goal of this Master'​s degree project is to design ​layer on top of search engines organizing ​the result ​of a query according ​to the analysis ​of this underlying multiplex networkOther avenues are worth exploring, trying ​to use this multiplex ​network to feed query expansion or specialization,​ for instance.+
  
 === Expected profile === === Expected profile ===
Line 39: Line 40:
 === Expected work === === Expected work ===
  
-The project is an opportunity to extend Renoust et al.'s approach to multiplex networks in relation with searching the web. A presentation layer should be made compatible with existing text-based interfaces; other more elaborate interfaces may also be explored. Benchmark tests should show the relevancy of the approach in terms of precision and recall ​(for query expansion, for instance), or ultimately in terms of user satisfaction. +A small of the project will require to build/tune a web scraper. ​The core part of the project is an opportunity to extend Renoust et al.'s approach to multiplex networks in relation with searching the web. We expect ​the necessity to redefine/​redesign usual network metrics ​(centralitiesby adapting them to these layered networks
visual_analytics_of_classified_ads.txt · Last modified: 2014/11/28 09:30 by melancon
Trace:
Dieses Dokuwiki verwendet ein von Anymorphic Webdesign erstelltes Thema.
www.chimeric.de Valid CSS Driven by DokuWiki do yourself a favour and use a real browser - get firefox!! Recent changes RSS feed Valid XHTML 1.0