Dialogue and Information Retrieval

Download Report

Transcript Dialogue and Information Retrieval

Dialogue and Information Retrieval
Dialogs on Dialogs
March all the way through April 2003
Intersections between
Dialog Systems and IR

Current work




Call Routing
Question Answering
Why so little?
What else? Let’s brainstorm!
Call Routing


Task: given a NL expression of a problem,
classify (route) it in one of several
categories
Examples



AT&T: How May I Help You
British Telecom
Jennifer Chu-Carroll
Call Routing (2)



It’s a classification problem!
Salience (co-ocurrence) based
approaches (AT&T)
IR-like approaches (J. Chu-Carroll)


Treat user requests as “documents”
Use VSM and cosine similarity to classify
The IR in Call Routing


Regard the problem as text classification
Do standard IR work:




LSA
LDA
Centroid vs. KNN approaches
Results? Classification perf?
The Dialog in Call Routing

Disambiguation


Follow-up dialog




Easy to do based on the VSM IR approach
HMIHY: frame-based follow-up dialogs
Q: Is Call Routing dialog management?
Q: Or is it more like understanding?
Q: Why typical understanding/DM
approaches fail in HMIHY-type domains?
Question Answering


Task: answer to a question in Natural
Language from a database of documents
in Natural Language.
Examples:


http://www.ai.mit.edu/projects/infolab/
http://www.ask.com
IR in Question Answering

Everywhere:




Document indexing
Retrieval
…
What is different from traditional IR?


Some parsing/understanding of questions and
documents
Some language generation (?)
Dialog in QA

Refining the question:



Clarification dialogue
Decide which question to ask
Only for very restricted domains  uses
fixed frames (Rutgers: HITIQA)
Why so little?

Different issues:



IR = lots of unstructured data, no NLP
Dialog = structured data, lots of NLP
Main problems:


Structural mismatch
NLU mismatch
But HUGE potential!

Voice-only random access to large
amounts of information (“Voice IR”):





technical manuals of “in-the-field” devices
(e.g. NASA)
tutorial systems
phone-based Google (e.g. legal information…)
GUI+ for IR
Learn dialog stuff from data (LM, NLG,
parsing…)