Casebased Question Classification

1
Case-based Question Classification

• John Bäckstrand jbd01003_at_student.mdh.se
• Master Thesis
• Supervisor Peter Sävström
• Examinator Peter Funk

2
Overview

• This presentation will
• Introduce some background concepts
• Why a CBR approach was choosen
• What work was done
• Summarize the results
• Further research?

3
Introduction Question Answering

• Question Answering is bunch of methods for
  automatically answering questions, in different
  ways
• Pure NLP Actually understand the question
• Statistical Google et al
• Template-based Who is X, where X is a person

4
Introduction Question Answering

• Pure NLP approach
• Parse the question into a syntax tree
• Match this against some kind of understanding
  system such as stored syntax trees, and the
  method to answer the question
• Very hard method!
• Works only in small domains.

5
Introduction Question Answering

• Statistical QA
• Search a number of documents for relevant
  sentences to return as answers.
• This is often done in multiple stages
• First retrieve important documents
• Then retrieve important passages
• Then try to extract the correct answer from a
  passage

6
Introduction Question Answering

• Template based QA
• Develop a collection of question templates that
  user input will have to match
• Pros
• Easy to implement
• Cons
• Bad performance
• Very narrow scope
• Tedious to develop templates

7
Introduction Question Answering

• A QA system want the answer type
• Does the user want
• a person
• a quantity
• a geographical location
• a kind of fruit
• a color
• And so on
• This can be tricky to know

8
Introduction Question Answering

• Tricky questions?
• What is the location of Norway location
• Where is Norway located location
• Where is society headed societal state
• Where is society today societal state
• Where is Norway today location?
• Where is Peter headed location?
• (Where is Peter headed, with his life) personal
  state
• (Where is Peter headed, in the car) location
• Where is Norway location
• How do I go to Norway description

9
Introduction Question Answering

• Understanding questions
• understanding language
• hard

10
Introduction NLP

• U

11
Question Answering

• U