ajeb first seminar

40
Ajeb Arabic Question Answering System

Upload: khaled-sayed

Post on 11-Jun-2015

490 views

Category:

Technology


0 download

TRANSCRIPT

Page 1: Ajeb First Seminar

AjebArabic Question Answering System

Page 2: Ajeb First Seminar

سج

Page 3: Ajeb First Seminar

Project Members

Eid Mosad El-Sayed CS Khaled Ahmed Sayed CS Sarah Abdel Monem Ismail CS

Supervised by: Prof. Dr. Mostafa Aref Dr. Ibrahim Fathy TA. Mohamed Hamdy سج

Page 4: Ajeb First Seminar

Agenda

Introduction:• Motivations.

• Problem Definition.

• Objective.

• Challenges.

Survey Summary:• Name Entity Recognition(NER).

• Previous Work.

Ajeb Architecture. Tools. Project Time Plan. References. سج

Page 5: Ajeb First Seminar

Introduction

Page 6: Ajeb First Seminar

Motivations

Introduction

Page 7: Ajeb First Seminar

Motivation

سج Ongoing update & progress in Arabic Natural

Language Processing (ANLP).

Page 8: Ajeb First Seminar

Motivation

سج There is still little research in the Arabic

language.

Page 9: Ajeb First Seminar

Motivation

سج Arabic is our mother language, so we prefer

the Arabic language over any other one.

Page 10: Ajeb First Seminar

Motivation

سج The Arabic language was ranked as the fifth

most important language in the world with 300 million speakers.

EnglishFrançaise

DeutschEspañol

Page 11: Ajeb First Seminar

Motivation

سج Arabic is the language of the holy Quran.

Page 12: Ajeb First Seminar

Problem Definition

Introduction

Page 13: Ajeb First Seminar

Problem Definition

سج The amount of available Arabic information

is becoming very huge.

Page 14: Ajeb First Seminar

Problem Definition

سج Search engines are not able to provide an

exact answer.

Page 15: Ajeb First Seminar

Problem Definition

سج Lack of time to find a short and precise

answer among the variety of available documents.

Page 16: Ajeb First Seminar

Objective

Introduction

Page 17: Ajeb First Seminar

Objective

سج Obtaining a brief and concise answer

for Arabic factoid questions extracted from internet corpus.

سج

Page 18: Ajeb First Seminar

Challenges

Introduction

Page 19: Ajeb First Seminar

Challenges

سج Arabic is highly inflectional and derivational,

which makes morphological analysis a very complex task.

فعلفعول

مفعول فاعل

فعال

Page 20: Ajeb First Seminar

Challenges

سج The absence of diacritics (which represent

most vowels) in the written text creates ambiguity.

Page 21: Ajeb First Seminar

Challenges

سج

The writing direction is from right-to-left and some of the characters change their shapes based on their location in the word.

Page 22: Ajeb First Seminar

Survey Summary

Page 23: Ajeb First Seminar

Common Methodology

سجDocuments

Name Entity Recognition (NER)

Page 24: Ajeb First Seminar

Name Entity Recognition (NER)

Survey Summary

Page 25: Ajeb First Seminar

Name Entity Recognition (NER)

سج

Classify elements in text into predefined categories:• Persons.

• Organizations.

• Locations.

• Dates.

• Quantities, monetary values and percentages.

Page 26: Ajeb First Seminar

Name Entity Recognition (NER)

سج

Importance:• Since the Arabic language is very hard to

understand by the computer, NER is been used to make the QA system semi-understand.

Page 27: Ajeb First Seminar

Name Entity Recognition (NER)

سج

What is the problem of recognizing names in the Arabic language?• Non-vocalization.

• Lack of capitalization.

• Delimitation problems.

Page 28: Ajeb First Seminar

Previous Work

Survey Summary

Page 29: Ajeb First Seminar

Previous Work - QARAB

سج

QARAB finds answers under the following assumptions: The answer exists in a collection of Arabic newspaper

text extracted from the Al-Raya newspaper. All supporting information for the answer lies in one

document . The answer is a short passage.

Page 30: Ajeb First Seminar

Ajeb Architecture

Page 31: Ajeb First Seminar

Ajeb Architecture

سج

Page 32: Ajeb First Seminar

Ajeb Architecture

.Question Analysis (1 سج

Page 33: Ajeb First Seminar

Ajeb Architecture

.Passage Retrieval (2 سج

Page 34: Ajeb First Seminar

Ajeb Architecture

.Answer Extraction (3 سج

Page 35: Ajeb First Seminar

Tools

Page 36: Ajeb First Seminar

Tools

سج

Page 37: Ajeb First Seminar

Project Time Plan

Page 38: Ajeb First Seminar

Project Time Plan

سج

Page 39: Ajeb First Seminar

References

سج

[1]Hammo, B., H.Abu-Salem, S.Lytinen and M.Evens, 2002. QARAB: A question answering system to support the Arabic language. Proceedings of the 40th Association for Computational Linguistics on Computational Approaches to Semetic Languages, ACL’02 University of Pennsylvania, PA, USA, 55-65.

[2]Paolo Rosso, Yassine Benajiba, and Abdelouahid Lyhyaoui, "Towards an Arabic Question Answering System" 2007.

[3]Wissal Brini, Mariem Ellouze, Omar Trigui(ANLP Research Group), Slim Mesfar, Lamia Hadrich Belguith, Paolo Rosso, "Factoid and definitional Arabic Question Answering system", 2008.

[4]Saleem Abuleil, "Extracting Names From Arabic Text For Question-Answering Systems", 2003.

Page 40: Ajeb First Seminar

Thankshttp://www.ajeb-aqas.blogspot.com