Departamento de Lenguajes y Sistemas Informáticos

30
Departamento de Lenguajes y Sistemas Informáticos Spoken Document Retrieval experiments with IR-n system Fernando Llopis Pascual Patricio Martínez-Barco

description

Spoken Document Retrieval experiments with IR-n system. Fernando Llopis Pascual Patricio Martínez-Barco. Departamento de Lenguajes y Sistemas Informáticos. Index. IR-n System. Adapting IR-n System to SDR task. Evaluation. Conclusions and future work. Index. IR-n System. - PowerPoint PPT Presentation

Transcript of Departamento de Lenguajes y Sistemas Informáticos

Page 1: Departamento de  Lenguajes y Sistemas Informáticos

Departamento de Lenguajes y Sistemas Informáticos

Spoken Document Retrieval experiments with IR-n system

Fernando Llopis PascualPatricio Martínez-Barco

Page 2: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 3: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 4: Departamento de  Lenguajes y Sistemas Informáticos

Use short fragments of documents instead of whole documents to evaluate the relevance or similarity

These fragments are called passages Each document is divided into passages

before calculating the relevance

IR-n System

Passage Retrieval Systems

Page 5: Departamento de  Lenguajes y Sistemas Informáticos

Why IR-n system use the sentence to define the passages ?

A sentence expresses an idea in the document There are algorithms to obtain each sentence with a

great precision Sentences are full units allowing to show an

understandable information by users or provide this information to a subsequent system

IR-n System

Passage concept

Page 6: Departamento de  Lenguajes y Sistemas Informáticos

General Custer was Civil War Union Major soldier. One of the most famous and controversial figures in United States Military history. Graduated last in his West Point Class (June 1861). Spent first part of the Civil War as a courier and staff officer. Promoted from Captain to Brigadier General of Volunteers just prior to the Battle of Gettysburg, and was given command of the Michigan "Wolverines" Cavalary brigade.

He helped defeat General Stuart's attempt to make a cavalry strike behind Union lines on the 3rd Day of the Battle (July 3, 1863), thus markedly contributing to the Army of the Potomac's victory (a large monument to his Brigade now stands in the East Cavalry Field in Gettysburg). Participated in nearly every cavalry action in Virginia from that point until the end of the war, always performing boldly, most often brilliantly, and always seeking publicity for himself and his actions. Ended the war as a Major General of Volunteers and a Brevet Major General in the Regular Army.

Upon Army reorganization in 1886, he was appointed Lieutenant Colonel of the soon to be renown 7th United States Cavalry. Fought in the various actions against the Western Indians, often with a singular brutality (exemplified by his wiping out of a Cheyenne village on the Washita in November 1868).

His exploits on the Plains were romanticized by Eastern Unites States newspapermen, and he was elevated to legendary status in his time. The death of his friend, Lucarelli change his life.

1 – Obtains sentences from the document

2 – Defines passages according to a fixed number of sentences

IR-n System

Passage concept

IR-n system defines the passages in the following way

SENTENCE 1

SENTENCE 2

SENTENCE 3

SENTENCE 4

SENTENCE 5

SENTENCE 6

SENTENCE 7

SENTENCE 8

SENTENCE 9

SENTENCE 10

SENTENCE 11

SENTENCE 12

SENTENCE 13

SENTENCE 14

SENTENCE 15

Passage 1

Passage 2

Passage 3

Page 7: Departamento de  Lenguajes y Sistemas Informáticos

Every passage has the same number of sentences

This number depends on The collection of documents Size of the query

IR-n System

Passage concept

Page 8: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 9: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 10: Departamento de  Lenguajes y Sistemas Informáticos

As appointed by Dahlback (1997): Spoken input is often incomplete and incorrect Contains interruptions and repairs Sentences occur only very occasionally

Conclusion: Sentence concept is not valid in spoken input

Therefore new basic units for dialogue models must be proposed:

Utterances instead of sentences Turns instead of paragraphs

Adapting IR-n system to SDR task

Spoken input

Page 11: Departamento de  Lenguajes y Sistemas Informáticos

Utterance: sequency of words chained by a speaker between two

pauses.

Turn: set of utterances that a speaker can express between

two speaker changes (dialogues) set of utterances that a speaker expresses about the

same subject (monologues)

(each section of TREC SDR collection is going to be considered as a turn)

Adapting IR-n system to SDR task

Definitions

Page 12: Departamento de  Lenguajes y Sistemas Informáticos

The lack of punctuation marks impedes the recognition of utterance boundaries

Utterances boundaries must be estimated detecting longest pauses

Some turns have not semantic content “Morning C.N.N. headline news I’m Sachi Koto”

Some turns are interrupted due to Overlaps Speaker mistakes Repetitions Modifications of previous information

Noise incorporate by Automatic transcriptors

Adapting IR-n system to SDR task

SDR problems

Page 13: Departamento de  Lenguajes y Sistemas Informáticos

Adapting IR-n system to SDR task

IR-n problems

The lack of sentences to define passages must be solved with the use of utterances

An utterance splitter was developed Overlapping passage technique was used to minimize

fails of utterance splitting

Noise inputs How the system supports them must be tested

Page 14: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 15: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 16: Departamento de  Lenguajes y Sistemas Informáticos

The main goal of this experiment is to know the robustness of IR-n system:

How a system based on passages (therefore based on sentences) can be adapted to utterances

How the system supports noise

Evaluation

Evaluation goal

Page 17: Departamento de  Lenguajes y Sistemas Informáticos

Discovering the minimum time between words to consider a new utterance

…………………..<Word S_time=156.010 E_time=156.140> TO </Word>

<Word S_time=156.140 E_time=156.600> THWART </Word>

<Word S_time=156.600 E_time=156.830> THEIR </Word>

<Word S_time=156.830 E_time=157.480> ABILITY </Word>

<Word S_time=157.510 E_time=157.810> TO </Word>

<Word S_time=157.840 E_time=158.330> ACQUIRE </Word>

<Word S_time=158.330 E_time=158.450> AND </Word>

<Word S_time=158.450 E_time=158.890> DEVELOP </Word>

<Word S_time=158.920 E_time=159.350> WEAPONS </Word>

…………………..

Evaluation

Training focus

Page 18: Departamento de  Lenguajes y Sistemas Informáticos

Discovering the minimum time between words to consider a new utterance

…………………..<Word S_time=156.010 E_time=156.140> TO </Word>

<Word S_time=156.140 E_time=156.600> THWART </Word>

<Word S_time=156.600 E_time=156.830> THEIR </Word>

<Word S_time=156.830 E_time=157.480> ABILITY </Word>

<Word S_time=157.510 E_time=157.810> TO </Word>

<Word S_time=157.840 E_time=158.330> ACQUIRE </Word>

<Word S_time=158.330 E_time=158.450> AND </Word>

<Word S_time=158.450 E_time=158.890> DEVELOP </Word>

<Word S_time=158.920 E_time=159.350> WEAPONS </Word>

…………………..

Evaluation

Training focus

That is not a new utterance

Page 19: Departamento de  Lenguajes y Sistemas Informáticos

Discovering the minimum time between words to consider a new utterance

…………………..<Word S_time=215.130 E_time=215.330> BUT </Word><Word S_time=215.350 E_time=215.470> FOR </Word><Word S_time=215.470 E_time=215.610> THE </Word><Word S_time=215.610 E_time=215.900> BAY'S </Word><Word S_time=215.900 E_time=216.190> CHIEF </Word><Word S_time=217.680 E_time=218.270> I </Word><Word S_time=219.780 E_time=219.950> WHAT </Word><Word S_time=220.010 E_time=220.160> WOULD </Word><Word S_time=220.160 E_time=220.340> THEY </Word><Word S_time=220.340 E_time=220.910> ACHIEVED </Word>…………………..

Evaluation

Training focus

Page 20: Departamento de  Lenguajes y Sistemas Informáticos

Discovering the minimum time between words to consider a new utterance

…………………..<Word S_time=215.130 E_time=215.330> BUT </Word><Word S_time=215.350 E_time=215.470> FOR </Word><Word S_time=215.470 E_time=215.610> THE </Word><Word S_time=215.610 E_time=215.900> BAY'S </Word><Word S_time=215.900 E_time=216.190> CHIEF </Word><Word S_time=217.680 E_time=218.270> I </Word><Word S_time=219.780 E_time=219.950> WHAT </Word><Word S_time=220.010 E_time=220.160> WOULD </Word><Word S_time=220.160 E_time=220.340> THEY </Word><Word S_time=220.340 E_time=220.910> ACHIEVED </Word>…………………..

Evaluation

Training focus

That is a new utterance

Page 21: Departamento de  Lenguajes y Sistemas Informáticos

Discovering the better size for passages

Evaluation

Training focus

UTTERANCE 1

UTTERANCE 2

UTTERANCE 3

UTTERANCE 4

UTTERANCE 5

UTTERANCE 6

UTTERANCE 7

UTTERANCE 8

UTTERANCE 9

UTTERANCE 10

UTTERANCE 11

UTTERANCE 12

UTTERANCE 13

UTTERANCE 14

UTTERANCE 15

UTTERANCE 1

UTTERANCE 2

UTTERANCE 3

UTTERANCE 4

UTTERANCE 5

UTTERANCE 6

UTTERANCE 7

UTTERANCE 8

UTTERANCE 9

UTTERANCE 10

UTTERANCE 11

UTTERANCE 12

UTTERANCE 13

UTTERANCE 14

UTTERANCE 15

UTTERANCE 1

UTTERANCE 2

UTTERANCE 3

UTTERANCE 4

UTTERANCE 5

UTTERANCE 6

UTTERANCE 7

UTTERANCE 8

UTTERANCE 9

UTTERANCE 10

UTTERANCE 11

UTTERANCE 12

UTTERANCE 13

UTTERANCE 14

UTTERANCE 15

Page 22: Departamento de  Lenguajes y Sistemas Informáticos

Training corpus : TREC SDR-8 collection (according to the track specification)

Parameters to be evaluated: Number of utterances / passage = (from 1 to 9)

Pause size considered for utterance split = (0.1, 0.2, 0.3 sec.)

Models: With query expansion Without query expansion

Evaluation

Training

Page 23: Departamento de  Lenguajes y Sistemas Informáticos

Evaluation

Training

Best AvgP

Best pause estimation

0.4620

0.2

Best size of passage 5

Training results

Best model WITH

Page 24: Departamento de  Lenguajes y Sistemas Informáticos

Evaluation

Monolingual test

Organization

ITC-irst

IR-n Alicante

AvgP

0,3944

0,3637

Exeter 0,3824

1

3

2

JHU/APL 0,31844

IR-n Alicante 0,35633

Monolingual results

Test corpus : TREC SDR-9 collection Parameters :

Number of utterances / passage = 5 Pause size considered for utterance split = 0.2 seconds

Model : with query expansion

Page 25: Departamento de  Lenguajes y Sistemas Informáticos

French queries were translated into English using machine translation:

Power translator Free translator Babel fish

Evaluation

Bilingual test (French-English)

Page 26: Departamento de  Lenguajes y Sistemas Informáticos

Organization

ITC-irst

IR-n Alicante

AvgP

0,3064

0,2825Exeter

0,3032

1

3

2

JHU/APL 0,19044

IR-n Alicante 0,28462

Bilingual results

Evaluation

Bilingual (French-English)

Test corpus : TREC SDR-9 collection Parameters :

Number of utterances / passage = 5 Pause size considered for utterance split = 0.2 seconds

Model : with query expansion

Page 27: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 28: Departamento de  Lenguajes y Sistemas Informáticos

Index

Adapting IR-n System to SDR task

Evaluation

IR-n System

Conclusions and future work

Page 29: Departamento de  Lenguajes y Sistemas Informáticos

Conclusions: IR-n System is robust when working in SDR task (+) IR-n System performance must be increased (-)

Future work: Reduce noise produced by repetitions – modifications Remove turns without semantic content Evaluate and improve our utterance splitter

Conclusions and future work

Page 30: Departamento de  Lenguajes y Sistemas Informáticos

Departamento de Lenguajes y Sistemas Informáticos

Spoken Document Retrieval experiments with IR-n system

Fernando Llopis PascualPatricio Martínez-Barco