Question Answering

1

Question Answering

Marti HearstNovember 14, 2005

2

Question Answering

OutlineIntroduction to QAA typical full-fledged QA systemA very simple system, in response to thisAn intermediate approachIncorporating a reasoning systemMachine Learning of mappingsOther question types (e.g., biography, definitions)

3

A of Search Types

What is the typical height of a giraffe?

What are some good ideas for landscaping my client’s yard?

What are some promising untried treatments for Raynaud’s disease?Text Data Mining

Browse and Build

Question/Answer

4Adapted from slide by Surdeanu and Pasca

Document RetrievalUsers submit queries corresponding to their information needs.System returns (voluminous) list of full-length documents.It is the responsibility of the users to find information of interest within the returned documents.

Open-Domain Question Answering (QA)Users ask questions in natural language.

What is the highest volcano in Europe?System returns list of short answers.

… Under Mount Etna, the highest volcano in Europe, perches the fabulous town …A real use for NLP

Beyond Document Retrieval

5

Questions and Answers

What is the height of a typical giraffe?

The result can be a simple answer, extracted from existing web pages.Can specify with keywords or a natural language query

– However, most web search engines are not set up to handle questions properly.

– Get different results using a question vs. keywords

6

7

8

9


The Problem of Question Answering

What is the nationality of Pope John Paul II? … stabilize the country with its help, the Catholic hierarchy stoutly held out

for pluralism, in large part at the urging of Polish-born Pope John Paul II. When the Pope emphatically defended the Solidarity trade union during a 1987 tour of the…

Natural language question, not keyword

queries

Short text fragment, not URL

list

11Adapted from slides by Manning, Harabagiu, Kushmeric, and ISI

Question Answering from text

With massive collections of full-text documents, simply finding relevant documents is of limited use: we want answers QA: give the user a (short) answer to their question, perhaps supported by evidence. An alternative to standard IR

The first problem area in IR where NLP is really making a difference.


People want to ask questions…

Examples from AltaVista query logwho invented surf music?how to make stink bombswhere are the snowdens of yesteryear?which english translation of the bible is used in official catholic liturgies?how to do clayarthow to copy psxhow tall is the sears tower?

Examples from Excite query log (12/1999)how can i find someone in texaswhere can i find information on puritan religion?what are the 7 wonders of the worldhow can i eliminate stressWhat vacuum cleaner does Consumers Guide recommend


A Brief (Academic) HistoryIn some sense question answering is not a new research area Question answering systems can be found in many areas of NLP research, including:

Natural language database systems– A lot of early NLP work on these

Problem-solving systems– STUDENT (Winograd ’77)– LUNAR (Woods & Kaplan ’77)

Spoken dialog systems– Currently very active and commercially relevant

The focus is now on open-domain QA is newFirst modern system: MURAX (Kupiec, SIGIR’93):

– Trivial Pursuit questions – Encyclopedia answers

FAQFinder (Burke et al. ’97)TREC QA competition (NIST, 1999–present)


AskJeeves

AskJeeves is probably most hyped example of “Question answering”How it used to work:

Do pattern matching to match a question to their own knowledge base of questionsIf a match is found, returns a human-curated answer to that known questionIf that fails, it falls back to regular web search(Seems to be more of a meta-search engine now)

A potentially interesting middle ground, but a fairly weak shadow of real QA


Question Answering at TREC

Question answering competition at TREC consists of answering a set of 500 fact-based questions, e.g.,

“When was Mozart born?”.

Has really pushed the field forward.The document set

Newswire textual documents from LA Times, San Jose Mercury News, Wall Street Journal, NY Times etcetera: over 1M documents now.Well-formed lexically, syntactically and semantically (were reviewed by professional editors).

The questionsHundreds of new questions every year, the total is ~2400

TaskInitially extract at most 5 answers: long (250B) and short (50B).Now extract only one exact answer.Several other sub-tasks added later: definition, list, biography.

16

Sample TREC questions

1. Who is the author of the book, "The Iron Lady: A Biography of Margaret Thatcher"?

2. What was the monetary value of the Nobel Peace Prize in 1989?

3. What does the Peugeot company manufacture?

4. How much did Mercury spend on advertising in 1993?

5. What is the name of the managing director of Apricot Computer?

6. Why did David Koresh ask the FBI for a word processor?

7. What is the name of the rare neurological disease with symptoms such as: involuntary movements (tics), swearing, and incoherent vocalizations (grunts, shouts, etc.)?

17

TREC Scoring

For the first three years systems were allowed to return 5 ranked answer snippets (50/250 bytes) to each question.

Mean Reciprocal Rank Scoring (MRR):– Each question assigned the reciprocal rank of the first correct

answer. If correct answer at position k, the score is 1/k. 1, 0.5, 0.33, 0.25, 0.2, 0 for 1, 2, 3, 4, 5, 6+ position

Mainly Named Entity answers (person, place, date, …)

From 2002 on, the systems are only allowed to return a single exact answer and the notion of confidence has been introduced.

18Adapted from slides by Manning, Harabagiu, Kusmerick, ISI

Top Performing SystemsIn 2003, the best performing systems at TREC can answer approximately 60-70% of the questionsApproaches and successes have varied a fair deal

Knowledge-rich approaches, using a vast array of NLP techniques stole the show in 2000-2003

– Notably Harabagiu, Moldovan et al. ( SMU/UTD/LCC )

Statistical systems starting to catch upAskMSR system stressed how much could be achieved by very simple methods with enough text (and now various copycats)People are experimenting with machine learning methods

Middle ground is to use large collection of surface matching patterns (ISI)

19

Example QA System

This system contains many components used by other systems, but more complex in some waysMost work completed in 2001; there have been advances by this group and others since then.Next slides based mainly on:

Paşca and Harabagiu, High-Performance Question Answering from Large Text Collections, SIGIR’01.Paşca and Harabagiu, Answer Mining from Online Documents, ACL’01.Harabagiu, Paşca, Maiorano: Experiments with Open-Domain Textual Question Answering. COLING’00


QA Block Architecture

QuestionProcessing

QuestionProcessing

PassageRetrieval

PassageRetrieval

AnswerExtraction

AnswerExtraction

WordNet

NER

Parser

WordNet

NER

ParserDocumentRetrieval

DocumentRetrieval

Keywords Passages

Question Semantics

Captures the semantics of the questionSelects keywords for PR

Extracts and ranks passagesusing surface-text techniques

Extracts and ranks answersusing NL techniques

Q A


Question Processing Flow

Q Questionparsing

Questionparsing

Construction of thequestion representation

Construction of thequestion representation

Answer type detectionAnswer type detection

Keyword selectionKeyword selection

Questionsemantic representation

AT category

Keywords


Question Stems and Answer Types

Question Question stem Answer type

Q555: What was the name of Titanic’s captain? What Person

Q654: What U.S. Government agency registers trademarks?

What Organization

Q162: What is the capital of Kosovo? What City

Q661: How much does one ton of cement cost? How much Quantity

Other question stems: Who, Which, Name, How hot...Other answer types: Country, Number, Product...

Identify the semantic category of expected answers


Detecting the Expected Answer Type

In some cases, the question stem is sufficient to indicate the answer type (AT)

Why REASONWhen DATE

In many cases, the question stem is ambiguousExamples

– What was the name of Titanic’s captain ?– What U.S. Government agency registers trademarks?– What is the capital of Kosovo?

Solution: select additional question concepts (AT words) that help disambiguate the expected answer typeExamples

– captain– agency– capital

24

Answer Type TaxonomyEncodes 8707 English concepts to help recognize expected answer typeMapping to parts of Wordnet done by hand

Can connect to Noun, Adj, and/or Verb subhierarchies


Answer Type Detection Algorithm

Select the answer type word from the question representation.

Select the word(s) connected to the question. Some “content-free” words are skipped (e.g. “name”).From the previous set select the word with the highest connectivity in the question representation.

Map the AT word in a previously built AT hierarchyThe AT hierarchy is based on WordNet, with some concepts associated with semantic categories, e.g. “writer” PERSON.

Select the AT(s) from the first hypernym(s) associated with a semantic category.


Keyword Selection

Answer Type indicates what the question is looking for, but provides insufficient context to locate the answer in very large document collectionLexical terms (keywords) from the question, possibly expanded with lexical/semantic variations provide the required context.


Lexical Term Extraction

Questions approximated by sets of unrelated words (lexical terms)Similar to bag-of-word IR models

Question (from TREC QA track) Lexical terms

Q002: What was the monetary value of the Nobel Peace Prize in 1989?

monetary, value, Nobel, Peace, Prize

Q003: What does the Peugeot company manufacture?

Peugeot, company, manufacture

Q004: How much did Mercury spend on advertising in 1993?

Mercury, spend, advertising, 1993

Q005: What is the name of the managing director of Apricot Computer?

name, managing, director, Apricot, Computer


Keyword Selection Algorithm

1. Select all non-stopwords in quotations2. Select all NNP words in recognized named entities3. Select all complex nominals with their adjectival

modifiers4. Select all other complex nominals5. Select all nouns with adjectival modifiers6. Select all other nouns7. Select all verbs8. Select the AT word (which was skipped in all previous

steps)


Keyword Selection Examples

What researcher discovered the vaccine against Hepatitis-B?

Hepatitis-B, vaccine, discover, researcher

What is the name of the French oceanographer who owned Calypso?

Calypso, French, own, oceanographer

What U.S. government agency registers trademarks?U.S., government, trademarks, register, agency

What is the capital of Kosovo?Kosovo, capital


Passage Retrieval

QuestionProcessing

QuestionProcessing

PassageRetrieval

PassageRetrieval

AnswerExtraction

AnswerExtraction

WordNet

NER

Parser

WordNet

NER


DocumentRetrieval

Keywords Passages

Question Semantics




Q A


Passage Extraction Loop

Passage Extraction ComponentExtracts passages that contain all selected keywordsPassage size dynamicStart position dynamic

Passage quality and keyword adjustmentIn the first iteration use the first 6 keyword selection heuristicsIf the number of passages is lower than a threshold query is too strict drop a keywordIf the number of passages is higher than a threshold query is too relaxed add a keyword


Passage Retrieval Architecture

Passage ExtractionPassage Extraction

Passage Quality

Keyword Adjustment

Keyword Adjustment

Passage Scoring

Passage Scoring

Passage Ordering

Passage Ordering

Keywords No

Passages

Yes

Documents

DocumentRetrieval

DocumentRetrieval

RankedPassages


Passage ScoringPassages are scored based on keyword windowsFor example, if a question has a set of keywords: {k1, k2, k3, k4}, and in a passage k1 and k2 are matched twice, k3 is matched once, and k4 is not matched, the following windows are built:

k1 k2 k3k2 k1

Window 1

k1 k2 k3k2 k1

Window 2

k1 k2 k3k2 k1

Window 3

k1 k2 k3k2 k1

Window 4


Passage Scoring

Passage ordering is performed using a radix sort that involves three scores:

SameWordSequenceScore (largest)– Computes the number of words from the question that

are recognized in the same sequence in the window

DistanceScore (largest)– The number of words that separate the most distant

keywords in the window

MissingKeywordScore (smallest)– The number of unmatched keywords in the window


Answer Extraction

QuestionProcessing

QuestionProcessing

PassageRetrieval

PassageRetrieval

AnswerExtraction

AnswerExtraction

WordNet

NER

Parser

WordNet

NER


DocumentRetrieval

Keywords Passages

Question Semantics




Q A


Ranking Candidate Answers

Answer type: Person Text passage:

“Among them was Christa McAuliffe, the first private citizen to fly in space. Karen Allen, best known for her starring role in “Raiders of the Lost Ark”, plays McAuliffe. Brian Kerwin is featured as shuttle pilot Mike Smith...”

Best candidate answer: Christa McAuliffe

Q066: Name the first private citizen to fly in space.


Features for Answer Ranking

relNMW number of question terms matched in the answer passage

relSP number of question terms matched in the same phrase as the candidate answerrelSS number of question terms matched in the same sentence as the candidate answerrelFP flag set to 1 if the candidate answer is followed by a punctuation signrelOCTW number of question terms matched, separated from the candidate answer by at most three words and one commarelSWS number of terms occurring in the same order in the answer passage as in the questionrelDTW average distance from candidate answer to question term matches

SIGIR ‘01


Answer Ranking based on Machine Learning

Relative relevance score computed for each pair of candidates (answer windows)relPAIR = wSWS relSWS + wFP relFP

+ wOCTW relOCTW + wSP relSP + wSS relSS

+ wNMW relNMW + wDTW relDTW + threshold

If relPAIR positive, then first candidate from pair is more relevant

Perceptron model used to learn the weightsScores in the 50% MRR for short answers, in the 60% MRR for long answers


Can we make this simpler?

One reason systems became so complex is that they have to pick out one sentence within a small collection

The answer is likely to be stated in a hard-to-recognize manner.Alternative Idea:

What happens with a much larger collection? The web is so huge that you’re likely to see the answer stated in a form similar to the question

Goal: make the simplest possible QA system by exploiting this redundancy in the web

Use this as a baseline against which to compare more elaborate systems.The next slides based on:

– Web Question Answering: Is More Always Better? Dumais, Banko, Brill, Lin, Ng, SIGIR’02

– An Analysis of the AskMSR Question-Answering System, Brill, Dumais, and Banko, EMNLP’02.


AskMSR System Architecture

1 2

3

45


Step 1: Rewrite the questions

Intuition: The user’s question is often syntactically quite close to sentences that contain the answer

Where is the Louvre Museum located? The Louvre Museum is located in Paris

Who created the character of Scrooge?

Charles Dickens created the character of Scrooge.


Query rewritingClassify question into seven categories

Who is/was/are/were…?When is/did/will/are/were …?Where is/are/were …?

a. Hand-crafted category-specific transformation rulese.g.: For where questions, move ‘is’ to all possible locations

Look to the right of the query terms for the answer.

“Where is the Louvre Museum located?” “is the Louvre Museum located” “the is Louvre Museum located” “the Louvre is Museum located” “the Louvre Museum is located” “the Louvre Museum located is”

b. Expected answer “Datatype” (eg, Date, Person, Location, …)When was the French Revolution? DATE

Nonsense,but ok. It’sonly a fewmore queriesto the search engine.


Query Rewriting - weighting

Some query rewrites are more reliable than others.

+“the Louvre Museum is located”

Where is the Louvre Museum located?

Weight 5if a match,probably right

+Louvre +Museum +located

Weight 1Lots of non-answerscould come back too


Step 2: Query search engine

Send all rewrites to a Web search engineRetrieve top N answers (100-200)For speed, rely just on search engine’s “snippets”, not the full text of the actual document


Step 3: Gathering N-Grams

Enumerate all N-grams (N=1,2,3) in all retrieved snippetsWeight of an n-gram: occurrence count, each weighted by “reliability” (weight) of rewrite rule that fetched the document

Example: “Who created the character of Scrooge?”

Dickens 117Christmas Carol 78Charles Dickens 75Disney 72Carl Banks 54A Christmas 41Christmas Carol 45Uncle 31


Step 4: Filtering N-Grams

Each question type is associated with one or more “data-type filters” = regular expressionWhen…Where…What …Who …

Boost score of n-grams that match regexpLower score of n-grams that don’t match regexpDetails omitted from paper….

Date

Location

Person


Step 5: Tiling the Answers

Dickens

Charles Dickens

Mr Charles

Scores

20

15

10

merged, discardold n-grams

Mr Charles DickensScore 45

N-Gramstile highest-scoring n-gram

N-Grams

Repeat, until no more overlap


Results

Standard TREC contest test-bed (TREC 2001):~1M documents; 900 questions

Technique doesn’t do too well (though would have placed in top 9 of ~30 participants)

– MRR: strict: .34– MRR: lenient: .43– 9th place

52

ResultsFrom EMNLP’02 paper

MMR of .577; answers 61% correctlyWould be near the top of TREC-9 runs

Breakdown of feature contribution:


Issues

Works best/only for “Trivial Pursuit”-style fact-based questionsLimited/brittle repertoire of

question categoriesanswer data types/filtersquery rewriting rules


Intermediate Approach:Surface pattern discovery

Based on:Ravichandran, D. and Hovy E.H. Learning Surface Text Patterns for a Question Answering System, ACL’02Hovy, et al., Question Answering in Webclopedia, TREC-9, 2000.

Use of Characteristic Phrases"When was <person> born”

Typical answers– "Mozart was born in 1756.”– "Gandhi (1869-1948)...”

Suggests regular expressions to help locate correct answer– "<NAME> was born in <BIRTHDATE>”– "<NAME> ( <BIRTHDATE>-”


Use Pattern Learning

Examples:“The great composer Mozart (1756-1791) achieved fame at a young age”“Mozart (1756-1791) was a genius”“The whole world would always be indebted to the great music of Mozart (1756-1791)”

Longest matching substring for all 3 sentences is "Mozart (1756-1791)”

Suffix tree would extract "Mozart (1756-1791)" as an output, with score of 3

Reminiscent of IE pattern learning


Pattern Learning (cont.)

Repeat with different examples of same question type

“Gandhi 1869”, “Newton 1642”, etc.

Some patterns learned for BIRTHDATEa. born in <ANSWER>, <NAME>b. <NAME> was born on <ANSWER> , c. <NAME> ( <ANSWER> -d. <NAME> ( <ANSWER> - )


QA Typology from ISITypology of typical question forms—94 nodes (47 leaf nodes)Analyzed 17,384 questions (from answers.com)

(THING ((AGENT (NAME (FEMALE-FIRST-NAME (EVE MARY ...)) (MALE-FIRST-NAME (LAWRENCE SAM ...)))) (COMPANY-NAME (BOEING AMERICAN-EXPRESS)) JESUS ROMANOFF ...) (ANIMAL-HUMAN (ANIMAL (WOODCHUCK YAK ...)) PERSON) (ORGANIZATION (SQUADRON DICTATORSHIP ...)) (GROUP-OF-PEOPLE (POSSE CHOIR ...)) (STATE-DISTRICT (TIROL MISSISSIPPI ...)) (CITY (ULAN-BATOR VIENNA ...)) (COUNTRY (SULTANATE ZIMBABWE ...)))) (PLACE (STATE-DISTRICT (CITY COUNTRY...)) (GEOLOGICAL-FORMATION (STAR CANYON...)) AIRPORT COLLEGE CAPITOL ...) (ABSTRACT (LANGUAGE (LETTER-CHARACTER (A B ...))) (QUANTITY (NUMERICAL-QUANTITY INFORMATION-QUANTITY MASS-QUANTITY MONETARY-QUANTITY TEMPORAL-QUANTITY ENERGY-QUANTITY TEMPERATURE-QUANTITY ILLUMINATION-QUANTITY

(SPATIAL-QUANTITY (VOLUME-QUANTITY AREA-QUANTITY DISTANCE-QUANTITY))

... PERCENTAGE))) (UNIT ((INFORMATION-UNIT (BIT BYTE ... EXABYTE)) (MASS-UNIT (OUNCE ...)) (ENERGY-UNIT (BTU ...)) (CURRENCY-UNIT (ZLOTY PESO ...)) (TEMPORAL-UNIT (ATTOSECOND ... MILLENIUM)) (TEMPERATURE-UNIT (FAHRENHEIT KELVIN CELCIUS)) (ILLUMINATION-UNIT (LUX CANDELA)) (SPATIAL-UNIT ((VOLUME-UNIT (DECILITER ...)) (DISTANCE-UNIT (NANOMETER ...)))) (AREA-UNIT (ACRE)) ... PERCENT)) (TANGIBLE-OBJECT ((FOOD (HUMAN-FOOD (FISH CHEESE ...))) (SUBSTANCE ((LIQUID (LEMONADE GASOLINE BLOOD ...)) (SOLID-SUBSTANCE (MARBLE PAPER ...)) (GAS-FORM-SUBSTANCE (GAS AIR)) ...)) (INSTRUMENT (DRUM DRILL (WEAPON (ARM GUN)) ...) (BODY-PART (ARM HEART ...)) (MUSICAL-INSTRUMENT (PIANO))) ... *GARMENT *PLANT DISEASE)


Experiments

6 different question typesfrom Webclopedia QA Typology

– BIRTHDATE– LOCATION– INVENTOR– DISCOVERER– DEFINITION– WHY-FAMOUS


Experiments: pattern precision

BIRTHDATE:1.0 <NAME> ( <ANSWER> - )0.85 <NAME> was born on <ANSWER>,0.6 <NAME> was born in <ANSWER>0.59 <NAME> was born <ANSWER>0.53 <ANSWER> <NAME> was born0.50 - <NAME> ( <ANSWER>0.36 <NAME> ( <ANSWER> -

INVENTOR1.0 <ANSWER> invents <NAME>1.0 the <NAME> was invented by <ANSWER>1.0 <ANSWER> invented the <NAME> in


Experiments (cont.)

DISCOVERER1.0 when <ANSWER> discovered <NAME>1.0 <ANSWER>'s discovery of <NAME>0.9 <NAME> was discovered by <ANSWER> in

DEFINITION1.0 <NAME> and related <ANSWER>1.0 form of <ANSWER>, <NAME>0.94 as <NAME>, <ANSWER> and


Experiments (cont.)

WHY-FAMOUS1.0 <ANSWER> <NAME> called1.0 laureate <ANSWER> <NAME>0.71 <NAME> is the <ANSWER> of

LOCATION1.0 <ANSWER>'s <NAME>1.0 regional : <ANSWER> : <NAME>0.92 near <NAME> in <ANSWER>

Depending on question type, get high MRR (0.6–0.9), with higher results from use of Web than TREC QA collection


Shortcomings & Extensions

Need for POS &/or semantic types– "Where are the Rocky Mountains?”– "Denver's new airport, topped with white fiberglass

cones in imitation of the Rocky Mountains in the background , continues to lie empty”

– <NAME> in <ANSWER>

NE tagger &/or ontology could enable system to determine "background" is not a location


Shortcomings... (cont.)

Long distance dependencies"Where is London?”

"London, which has one of the busiest airports in the world, lies on the banks of the river Thames”

would require pattern like:<QUESTION>, (<any_word>)*, lies on <ANSWER>

Abundance & variety of Web data helps system to find an instance of patterns w/o losing answers to long distance dependencies


Shortcomings... (cont.)

System currently has only one anchor wordDoesn't work for Q types requiring multiple words from question to be in answer

– "In which county does the city of Long Beach lie?”– "Long Beach is situated in Los Angeles County”– required pattern:

<Q_TERM_1> is situated in <ANSWER> <Q_TERM_2>

Does not use case– "What is a micron?”– "...a spokesman for Micron, a maker of semiconductors,

said SIMMs are..."

If Micron had been capitalized in question, would be a perfect answer

65Adapted from slide by Harabagiu and Narayanan

The results of the past 5 TREC evaluations of QA systems indicate that current state-of-the-art QA is determined by the recognition of Named EntitiesIn TREC 2003 the LCC QA system extracted 289 correct answers for factoid questionsThe Name Entity Recognizer was responsible for 234 of them

QUANTITY 55 ORGANIZATION 15 PRICE 3

NUMBER 45 AUTHORED WORK 11 SCIENCE NAME 2

DATE 35 PRODUCT 11 ACRONYM 1

PERSON 31 CONTINENT 5 ADDRESS 1

COUNTRY 21 PROVINCE 5 ALPHABET 1

OTHER LOCATIONS 19 QUOTE 5 URI 1

CITY 19 UNIVERSITY 3

The Importance of NER


The Special Case of Names

1934: What is the play “West Side Story” based on?Answer: Romeo and Juliet

1976: What is the motto for the Boy Scouts?Answer: Be prepared.

1982: What movie won the Academy Award for best picture in 1989?Answer: Driving Miss Daisy

2080: What peace treaty ended WWI?Answer: Versailles

2102: What American landmark stands on Liberty Island?Answer: Statue of Liberty

Questions asking for names of authored works

67Adapted from slide by Shauna Eggers

Problems

NE assumes all answers are named entitiesOversimplifies the generative power of language!What about: “What kind of flowers did Van Gogh paint?”

Does not account well for morphological, lexical, and semantic alternations

Question terms may not exactly match answer terms; connections between alternations of Q and A terms often not documented in flat dictionaryExample: “When was Berlin’s Brandenburger Tor erected?” no guarantee to match builtRecall suffers


LCC Approach:WordNet to the rescue!WordNet can be used to inform all three steps of the Q/A process1. Answer-type recognition (Answer Type Taxonomy)2. Passage Retrieval (“specificity” constraints)3. Answer extraction (recognition of keyword alternations)

Using WN’s lexico-semantic info: Examples“What kind of flowers did Van Gogh paint?”

– Answer-type recognition: need to know (a) answer is a kind of flower, and (b) sense of the word flower

– WordNet encodes 470 hyponyms of flower sense #1, flowers as plants

– Nouns from retrieved passages can be searched against these hyponyms

“When was Berlin’s Brandenburger Tor erected?”– Semantic alternation: erect is a hyponym of sense #1 of

build

69

WN for Answer Type RecognitionEncodes 8707 English concepts to help recognize expected answer typeMapping to parts of Wordnet done by hand

Can connect to Noun, Adj, and/or Verb subhierarchies


WN in Passage Retrieval

Identify relevant passages from textExtract keywords from the question, andPass them to the retrieval module

“Specificity” – filtering question concepts/keywords

Focuses search, improves performance and precisionQuestion keywords can be omitted from the search if they are too generalSpecificity calculated by counting the hyponyms of a given keyword in WordNet

– Count ignores proper names and same-headed concepts– Keyword is thrown out if count is above a given threshold

(currently 10)


WN in Answer ExtractionIf keywords alone cannot find an acceptable answer, look for alternations in WordNet!

Q196: Who wrote “Hamlet”?

Morphological Alternation: wrote written

Answer: before the young playwright has written Hamlet – and Shakespeare seizes the opportunity

Q136: Who is the queen of Holland?

Lexical Alternation: Holland Netherlands

Answer: Pricess Margrit, sister of Queen Beatrix of the Netherlands, was also present

Q196: What is the highest mountain in the world?

Semantic Alternation: mountain peak

Answer: first African country to send an expedition to Mount Everest, the world’s highest peak


EvaluationPaşca/Harabagiu (NAACL’01 Workshop) measured approach using TREC-8 and TREC-9 test collectionsWN contributions to Answer Type Recognition

Count number of questions for which acceptable answers were found; 3GB text collection, 893 questions

All What only

Flat dictionary (baseline) 227 (32%) 48 (13%)

A-type taxonomy (static) 445 (64%) 179 (50%)

A-type taxonomy (dynamic) 463 (67%) 196 (56%)

A-type taxonomy (dynamic + answer patterns) 533 (76%) 232 (65%)

Method # questions with correct answer type


Evaluation

WN contributions to Passage RetrievalImpact of keyword alternations

Impact of specificity knowledge

No alternations enabled 55.3% precision

Lexical alternations enabled 67.6%

Lexical + semantic alternations enabled 73.7%

Morphological expansions enabled 76.5%

TREC-8 TREC-9

Not included 133 (65%) 463 (67%)

Included 151 (76%) 515 (74%)

Specificity knowledge

# questions with correct answer infirst 5 documents returned

74

Going Beyond Word Matching

Use techniques from artificial intelligence to try to draw inferences from the meanings of the wordsThis is a highly unusual and ambitious approach.

Surprising it works at all!Requires huge amounts of hand-coded information

Uses notions of proofs and inference from logicAll birds fly. Robins are birds. Thus, robins fly.forall(X): bird(X) -> fly(x)forall(X,Y): student(X), enrolled(X,Y) -> school(Y)


Inference via a Logic Prover

The LCC system attempts inference to justify an answerIts inference engine is a kind of funny middle ground between logic and pattern matchingBut quite effective: 30% improvement

Q: When was the internal combustion engine invented?A: The first internal-combustion engine was built in 1867.

invent -> create_mentally -> create -> build


COGEXWorld knowledge from:

WordNet glosses converted to logic forms in the eXtended WordNet (XWN) project Lexical chains

– game:n#3 HYPERNYM recreation:n#1 HYPONYM sport:n#1

– Argentine:a#1 GLOSS Argentina:n#1

NLP axioms to handle complex NPs, coordinations, appositions, equivalence classes for prepositions etceteraNamed-entity recognizer

– John Galt HUMAN

A relaxation mechanism is used to iteratively uncouple predicates, remove terms from LFs. The proofs are penalized based on the amount of relaxation involved.


Logic Inference Example

“How hot does the inside of an active volcano get?” get(TEMPERATURE, inside(volcano(active))) “lava fragments belched out of the mountain were as hot as 300 degrees Fahrenheit” fragments(lava, TEMPERATURE(degrees(300)), belched(out, mountain)) volcano ISA mountain lava ISPARTOF volcano -> lava inside volcano fragments of lava HAVEPROPERTIESOF lava

The needed semantic information is in WordNet definitions, and was successfully translated into a form that was used for rough ‘proofs’


Axiom Creation

XWN AxiomsA major source of world knowledge is a general purpose knowledge base of more than 50,000 parsed and disambiguated WordNet glosses that are transformed into logical form for use during the course of a proof.

Gloss: Kill is to cause to die

Logical Form: kill_VB_1(e1,x1,x2) -> cause_VB_1(e1,x1,x3) & to_TO(e1,e2) & die_VB_1(e2,x2,x4)


Lexical ChainsLexical Chains

Lexical chains provide an improved source of world knowledge by supplying the Logic Prover with much needed axioms to link question keywords with answer concepts.

Question: How were biological agents acquired by bin Laden?

Answer: On 8 July 1998 , the Italian newspaper Corriere della Serra indicated that members of The World Front for Fighting Jews and Crusaders , which was founded by Bin Laden , purchased three chemical and biological_agent production facilities in

Lexical Chain: ( v - buy#1, purchase#1 ) HYPERNYM ( v - get#1, acquire#1 )


Axiom Selection Lexical chains and the XWN knowledge base work together to select and

generate the axioms needed for a successful proof when all the keywords in the questions are not found in the answer.

Question: How did Adolf Hitler die?

Answer: … Adolf Hitler committed suicide …

The following Lexical Chain is detected:( n - suicide#1, self-destruction#1, self-annihilation#1 ) GLOSS ( v - kill#1 ) GLOSS ( v - die#1, decease#1, perish#1, go#17, exit#3, pass_away#1, expire#2, pass#25 ) 2

The following axioms are loaded into the Prover: exists x2 all e1 x1 (suicide_nn(x1) -> act_nn(x1) & of_in(x1,e1) &

kill_vb(e1,x2,x2)).

exists x3 x4 all e2 x1 x2 (kill_vb(e2,x1,x2) -> cause_vb_2(e1,x1,x3) & to_to(e1,e2) & die_vb(e2,x2,x4)).

81

LCC System Refecences

The previous set of slides drew information from these sources:

The Informative Role of WordNet in Open-Domain Question Answering, Pasca and Harabagiu, WordNet and Other Lexical Resources, NAACL 2001 WorkshopPasca and Harabagiu, High Performance Question/Answering, SIGIR’01Moldovan, Clark, Harabagiu, Maiorano: COGEX: A Logic Prover for Question Answering. HLT-NAACL 2003Moldovan, Pasca, Harabagiu, and Surdeanu: Performance issues and error analysis in an open-domain question answering system. ACM Trans. Inf. Syst. 21(2): 133-154 (2003) Harabagiu and Maiorano, Abductive Processes for Answer Justification, AAAI Spring Symposium on Mining Answers from Texts and Knowledge Bases, 2002

82

Using Machine Learning in QA

The following slides are based on:Ramakrishnan, Chakrabarti, Paranjpe, Bhattacharyya, Is Question Answering an Acquired Skill? WWW’04

83

Learning Answer Type Mapping

Idea: use machine learning techniques to automatically determine answer types and query terms from questions.Two types of answer types:

Surface patterns– Infinite set, so can’t be covered by a lexicon– DATES NUMBERS PERSON NAMES LOCATIONS– “at DD:DD” “in the ‘DDs” “in DDDD” “Xx+ said”– Can also associate with synset [date#n#7]

WordNet synsets– Consider: “name an animal that sleeps upright”– Answer: “horse”

Ramakrishnan et al., Is Question Answering an Acquired Skill? WWW’04

84Ramakrishnan et al., Is Question Answering an Acquired Skill? WWW’04

Determining Answer Types

The hard ones are “what” and “which” questions.Two useful heuristics:

If the head of the NP appearing before the auxiliary or main verb is not a wh-word, mark this as an a-type clueOtherwise, the head of the NP appearing after the auxiliary/main verb is an atype clue.

85

Learning Answer Types

Given a QA pair (q, a):– (“name an animal that sleeps upright”, “horse”)

(1a) See which atype(s) “horse” can map to(1b) Look up the hypernyms of “horse” -> S(2a) Record the k words to the right of the q-word(2b) For each of these k words, look up their synsets

– An, animal, that

(2c) Increment the counts for those synsets that also appear in SDo significance testing

Compare synset frequencies against a background setRetain only those that are significantly associated with the question word more so than in general (chi-square)


86

Learning Answer Tyeps


87

Learning to Choose Query Terms

Which words from the question to use in the query?A tradeoff between precision and recall.Example:

“Tokyo is the capital of which country?”Want to use “Tokyo” verbatimProbably “capital” as wellBut maybe not “country”; maybe “nation” or maybe this word won’t appear in the retrieved passage at all.

– Also, “country” corresponds to the answer type, so probably we don’t want to require it to be in the answer text.


88

Learning to Choose Query Terms

Features:POS assigned to word and immediate neighborsStarts with uppercase letterIs a stopwordIDF scoreIs an answer-type for this questionAmbiguity indicators:

– # of possible WordNet senses (NumSense)– # of other WordNet synsets that describe this sense

E.g., for “buck”: stag, deer, doe (NumLemma)

Learner:J48 decision tree worked best


89

Learning to Choose Query TermsResults

WordNet ambiguity indicators were very helpful– Raised accuracy from 71-73% to 80%

Atype flag improved accuracy from 1-3%

90

Learning to Score Passages

Given a question, and answer, & a passage (q, a, r)Assign +1 if r contains aAssign –1 otherwise

Features:Do selected terms s from q appear in r?Does r have an answer zone a that does not s?Are the distances between tokens in a and s small?Does a have a strong WordNet similarity with q’s answer type?

Learner:Use logistic regression, since it produces a ranking rather than a hard classification into +1 or –1

– Produces a continuous estimate between 0 and 1


91


Results:F-scores are low (.33 - .56)However, reranking greatly improves the rank of the corresponding passages.Eliminates many non-answers, pushing better passages towards the top.

92Ramakrishnan et al., Is Question Answering an Acquired Skill? WWW’04


93

Computing WordNet Similarity

Path-based similarity measures are not all that good in WordNet

3 hops from entity to artifact3 hops from mammal to elephant

An alternative: Given a target synset t and an answer synset a:

– Measure the overlap of nodes on the path from t to all noun roots and from a to all noun roots

– Algorithm for computing similarity of t to a: If t is not a hypernym of a: assign 0 Else collect the set of hypernym synsets of t and a Call them Ht and Ha Compute the Jaccard overlap

» |Ht Intersect Ha| / |Ht Union Ha|


94

Computing WordNet SimilarityAlgorithm for computing similarity of t to a:

– |Ht Intersect Ha| / |Ht Union Ha|

chordate

vertebrate

mammal

elephant

placental mammalproboscidean

animal

living thing

organism

object

entity|Ht Intersect Ha|

|Ht Union Ha|

Ht = mammal, Ha = elephant7/10 = .7

Ht = animal, Ha = elephant5/10 = .5

Ht = animal, Ha = mammal4/7 = .57

Ht = mammal, Ha = fox7/11 = .63



System Extension: Definition Questions

Definition questions ask about the definition or description of a concept:

Who is John Galt?What is anorexia nervosa?

Many “information nuggets” are acceptable answersWho is George W. Bush?

– … George W. Bush, the 43rd President of the United States… – George W. Bush defeated Democratic incumbent

Ann Richards to become the 46th Governor of the State of Texas…

ScoringAny information nugget is acceptablePrecision score over all information nuggets


Definition Detection with Pattern Matching

What <be> a <QP> ?Who <be> <QP> ? example: “Who is Zebulon Pike?”<QP>, the <AP><QP> (a <AP>)<AP HumanConcept> <QP> example: “explorer Zebulon Pike”

Question patterns

Answer patterns

Q386: What is anorexia nervosa?

cause of anorexia nervosa, an eating disorder...

Q358: What is a meerkat? the meerkat, a type of mongoose, thrives in...

Q340: Who is Zebulon Pike? in 1806, explorer Zebulon Pike sighted the...


Answer Detection with Concept Expansion

Enhancement for Definition questionsIdentify terms that are semantically related to the phrase to defineUse WordNet hypernyms (more general concepts)

Question WordNet hypernym Detected answer candidate

What is a shaman? {priest, non-Christian priest}

Mathews is the priest or shaman

What is a nematode?

{worm} nematodes, tiny worms in soil

What is anise? {herb, herbaceous plant}

anise, rhubarb and other herbs


Online QA Examples

Examples (none work very well)AnswerBus:

– http://www.answerbus.com

Ionaut: – http://www.ionaut.com:8400/

LCC: – http://www.languagecomputer.com/demos/question

_answering/index.html