Identifying and Processing Questions

1
Identifying and Processing Questions

• Summer School on QA
• Kasetsart University, Oct. 2009

2
Introduction

• Processing questions is one of the major aspects
  of Question Answering studies
• Correct identification of questions and of their
  constitutive elements is necessary in order to
  propose correct and useful answers
• Dedicated annotation systems are needed in order
  to make corpora easily accessible, and eventually
  proceed towards automatic recognition of
  questions and automatic production of answers
• Outline of the tutorial
• Preliminary aspects of questions
• Types of questions
• Focus of questions
• Question annotation

3
1. Preliminary aspects of questions

• Two definitions of question
• "A sentence worded or expressed so as to elicit
  information" (Oxford English Dictionary)
• "A sentence, phrase or word which asks for
  information, reply or response" (English
  Wiktionary)
• The purpose of a question is to obtain
  information, using different linguistic devices
• There are therefore several ways to ask a
  question

4
1. Preliminary aspects of questions (2)

• Syntactic constructions
• Interrogative word (insertion of DO)
  subject/verb inversion wh-interrogatives
• Ex When does the train for Paris leave?
• (Insertion of DO) subject/verb inversion
  yes/no interrogatives
• Ex Does the train for Paris leave at 10 a.m.?
• Punctuation only
• Non-interrogative sentence with a question mark
  (written context)
• The train for Paris leaves at 10 a.m.?
• Intonation
• Non-interrogative sentence with an interrogative
  intonation contour (oral context)
• "The train for Paris leaves at 10 a.m."
• Inference
• Especially in key word searches, a string of
  words can imply the existence of an inherent
  question
• Train for Paris timetable

5
1. Preliminary aspects of questions (3)

• The choice depends on several parameters
• Oral vs. Written
• Informal vs. Formal
• Type of information elicited (unknown
  information, confirmation, etc.)
• Devices vary according to languages
• Syntactic construction of interrogatives
• Absence/presence of punctuation
• Different intonation contours
• Consequences for QA analysis
• Need for proper identification of questions and
  of the type of information elicited
• Need to be able to process questions which are
  not in the form of interrogatives

6
2. Types of questions

• The type of a question is defined according to
  the type of answer which is expected
• ex Q. What is the capital of Thailand?
• A. Bangkok.
• The answer is a fact, so the
  question eliciting this answer is of the fact
  type
• ex Q. What happens during rainy season?
• A. Roads become flooded.
• The answer is an event, so the question
  eliciting this answer is of the event type

7
2. Types of questions (2)
(elaborated in collaboration with the NAIST lab
at Kasetsart University)
8
2. Types of questions (3)

• How can we determine the type of a question?
• by looking at the interrogative word(s)
• ex How, Who, Where, How long...?
• by looking at predicative forms
• ex What happens if...?, make pizza, stages
  of, the cheapest hotel in...

9
2. Types of questions (4)

• Exercises can you recognize the type of the
  following questions?
• What is the capital city of France?
• What are the main rice diseases in Thailand?
• How can I go from Kasetsart Uni. to the train
  station?
• How long does it take to go from Bangkok to
  Kanchanaburi?
• What is question-answering?
• What are the names of the last three U.S.
  Presidents?
• What are the stages of rice growth?
• Why is Jim Thompson famous?
• What happens when rice is infected with a
  disease?
• What is the most beautiful temple to see in
  Bangkok?

10
2. Types of questions (5)

• Answers

FACT
EVALUATION
PROCEDURE
FACT
DEFINITION
SET OF FACTS
SEQ. OF EVENTS (?)
EVENT (?)
SET OF EVENTS (?)
EVALUATION
11
3. Focus of questions

• The focus of a question is basically what the
  question is about.
• The focus usually takes the form of
• a Noun Phrase (NP) for questions of the fact
  type
• ex What is the capital city of France?
• a proposition for Yes/No questions and questions
  of the event types
• ex Does Koh Chang get crowded in August?
• a Verb Phrase (VP) for questions of the
  procedure and evaluation types
• ex How can I change a motherboard?

12
3. Focus of questions (2)

• Some hints on focus structure
• NPs head modifiers
• ex Where will the 2016 Olympic Games be?
• What is the name of the first university
  built in Bangkok?
• VPs predicate-argument structure modifiers
• ex Who won the elections in Japan last month?

13
3. Focus of questions (3)

• Exercises can you recognize the type and focus
  of the following questions?
• What is the name of the Thai national airline?
• How do I write a dissertation proposal?
• Did Barack Obama win an election in 2008?
• What are the names of the four Beatles?
• What is a wapiti?

14
3. Focus of questions (3)

• Answers

FACT
PROCEDURE
EVENT
SET OF FACTS
DEFINITION
15
4. Question annotation

• Questions are tagged using XML formalism enriched
  with attributes
• Several steps
• Tag questions for focus and type
• ltquestionTYPE, focusFOCUSgt
• NB FOCUS corresponds to the string of
  words identified as the focus (see ex. later)
• Tag NPs and PPs using attributes inspired from
  thematic roles (cf. previous presentation) VPs
  are annotated as predicates

16
4. Question annotation (2)

• Attributes for the annotation of NPs and PPs

(elaborated in collaboration with the NAIST lab
at Kasetsart University)
17
4. Question annotation (3)

• Examples

18
4. Question annotation (4)

• Exercises can you annotate the following
  questions using the schema we have introduced?
• How do I change a tyre?
• What are the border countries of Thailand?
• Why did World War One start in 1914?

19
4. Question annotation (5)

• Answers

20
Summary

• When processing questions, you should be able
  to...
• Identify the devices used to ask a question
  (syntactic construction, punctuation, intonation,
  implied question...)
• Identify the type of the question by looking at
  its elements and imagining the type of answer
  elicited
• Identify the focus of the question, i.e. what the
  question is about, as well as the syntactic
  nature of the focus (Noun Phrase, Verb Phrase,
  Proposition...)
• Annotate questions with a simple XML formalism,
  using thematic roles to annotate NPs and VPs
  (agent, theme, localization...), and identifying
  predicates.

21
Thank you!