Native-code Rule Execution for efficient inference on large Web Graphs

From Master Projects
Revision as of 13:13, 3 December 2015 by Jui200 (talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

About Native-code Rule Execution for efficient inference on large Web Graphs

  • This project has not yet been fulfilled.
  • This project fits in the following Bachelor programs: {{#arraymap:|, |xXx|bachelorproject within::xXx|,}}
  • This project fits in the following masterareas: {{#arraymap:Computer Science and Communication, High Performance Distributed Computing, Internet and Web Technology, Knowledge Technology and Intelligent Internet Applications, Parallel and Distributed Computer Systems|, |xXx|project within::xXx|,}}


On the Web, search is moving from a traditional keyword search to a more entity-centric search. To support this new type of search, Web engines construct "Knowledge Graphs" -- graphs where nodes represent entities and edges identify semantic relations between them. See, for instance, Google's Knowledge Graph project, Microsoft's Satori, or Yahoo's Knowledge Graph.

An important research topic consists of performing inference on these graphs to enhance searches with more answers. One type of such inference is performed by if-then rules that can be extracted from ontologies. Applying if-then rules is challenging on Web graphs with billion edges and hundred of millions of nodes. In order to improve the performance, this project proposes to define new techniques to translate the declarative if-then rules into highly optimized C++ native code. This optimization should look into the actual structure of the rules, and produce code that is tailored for each specific rule. This project is embedded within the department's VLog project, which is currently one of the leading inference engines for large Knowledge Graphs.