Context-Aware Internet Sharma Chakravarthy UT Arlington December 19, 2008 1
Info. Mosaic We are working on a project termed Info. Mosaic -- to integrate information from heterogeneous sources One of the motivations is to extend the querying capability to internet instead of being satisfied with search Today, we know how to search sites individually; but cannot do much when it comes to combining data/information from multiple sources (except in limited/customized ways) The above requires lots of context information -- as we shall see 2
Motivation Find – Castles near London reachable by train in 2 -3 hours - Decision Making Process - Manually Integrate Results to arrive at a decision Schedules Train Schedules Castle Results Search: Castles near London 3
Motivation Example – 1 : Find – 3 -Bedroom House in Austin, TX within 2 miles of an “exemplary” school and within 5 miles of Y highway and priced under Z dollar Example – 2 : Find – A list of openings for Software Engineers in Companies having their stock price over ‘X’ dollars listed in Nasdaq Example – 3 : Find – Prices of CDs or Records of the 1998 Grammy award winner for Folk Category The GOAL is to integrate information from a small number of heterogeneous sources in different domains 4
Challenges Query Specification/Refinement/feedback – adaptive capability Query Planning & Optimization Query Reformulation/matching to sources Data Extraction Data, Schema / Ontology Integration Result ranking Result presentation/Visualization Inconsistency Management /confidence Security & Privacy Handling hidden web Source and semantics Discovery Generalization to arbitrary domains 5
Search and Querying Search Some work Querying Structured data Unstructured data 6
Query/search specification To specify the query, “Retrieve castles near London that are reachable by train in less than 2 hours” Input {castle, train, London} or {train, from, London, to, castle} or {train, castle, location, city, London, duration, 2 hours} or {castle, reachable, train, from, London, 2 hours, or, less than} Use context – profiles, semantics of usage, feedback and any other information to infer the above query or close to a meaningful query! For instance, the above inputs can also mean: 7
Query/search specification Input {castle, train, London} or {train, from, London, to, castle} or {train, castle, location, city, London, duration, 2 hours} or {castle, reachable, train, from, London, 2 hours, or, less than} Retrieve Castles near London that are reachable by Train Retrieve Hotels near London that are Castles and can be reached by a Train Retrieve Books whose title contain the words `Castle' or `Train' written by an author whose name is `London' 8
Useful context information Domain taxonomies Attribute associations with concepts in the taxonomy Types of attributes operators and their classification – spatial, temporal, other, … Dictionary ranking of the meaning of words User feedback on the usage of words or combinations Resolving ambiguities from the user Various types of source semantics Automatic or semi-automatic Discovery of this information is a separate (hard) problem. 9
The Info. Mosaic Framework Ranking XQueries Data Integration Spatial Data Repository Final Result-set Ranking XML Repository Data Store Query Results Dictionary Users Operators Domain Knowledge Metadata Ontology Source Semantics I/O Attributes Schemas Query Execution and Data Extraction Knowledge-base y Quer Web Pages Feed back Query User Refinement Query/Inten Module Internet Statistics User Query Planner and Optimizer Query Plan XML Data t Query Interfaces Spatial DB Web Services Hidden Web Data …… 10
Current Status Working on Query Specification Query Planning and Execution Identifying components of the KB Identifying source semantics Discovery of some of the above either automatically or semi-automatically 11
Thank you! 12
Скачать презентацию Context-Aware Internet Sharma Chakravarthy UT Arlington December 19
9af195d22b1eddad1c30d12835c5ccc5.ppt