EUREKA: Dialogue-based IR Antoine Raux IR Seminar and Lab 11-743 Fall 2003 Mid-term Presentation.

Download Report

Transcript EUREKA: Dialogue-based IR Antoine Raux IR Seminar and Lab 11-743 Fall 2003 Mid-term Presentation.

EUREKA: Dialogue-based IR
Antoine Raux
IR Seminar and Lab 11-743 Fall 2003
Mid-term Presentation
Goal
 Build a light-weight, “natural language”
dialogue-based interface to refine queries
and select relevant documents.
Overview
 Backend: Vivisimo (using web scraper)
 Dialogue Management: RavenClaw
(successor of CMU Communicator)
 Language Understanding: Light Open
Vocabulary Parser, created for this project.
Very simple so far (first word=command,
other words=arguments)
 NLG/TTS: template-based & Festival
Vivisimo
 Available clustering meta-search engine
www.vivisimo.com
 Hand-written Perl web scraper
(hope Vivisimo doesn’t change their page
design by the end of the semester…)
RavenClaw
 Semi-declarative language (C++ macros)
 Hierarchical agent architecture:
EUREKA
Greet User
Submit
Query
Prompt
Query
Get
Cluster List
New Query
Get
Doc List
Open
Cluster
Inform of
Results
Close
…
Cluster
LOV Parser
 Problem: traditional NL parsers require a
dictionary of all valid words  not applicable
to open domain IR
 Solution (implemented in C++):
– fix a small number of one-word commands
(new_query, open, list_clusters)
– parse each line as “[command] [arguments]” or
“[command]” or “[arguments]”
NLG/TTS
 Template-based Language Generation
(e.g. “I found <n_doc> documents.”)
 General purpose Festival voice for TTS
NB: Browsing through lists is not efficient
with speech, even for lists of clusters
Already Implemented
 Working prototype
 Commands:
– new_query
– list_clusters, list_documents
– open, close (cluster)
– more, back (list of clusters/documents)
Demo
Future Work
 More commands: summaries, query
modification (with +, -, “”, ~…)
 Based on literature (e.g. Belkin), design
search strategies
 Design/conduct evaluation experiment
To be continued…
Possible Actions by User





Make a new query
Refine/reformulate a query
Select a cluster
Ask for a document summary
Ask for a cluster summary
Possible Actions by the System





Give a list of doc titles
Give a list of clusters
Give the summary of a document
Give the summary of a cluster
Propose a query refinement/reformulation