Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with...

45
Ontology-based fuzzy event extraction agent for Chinese e-news summarization Expert Systems with Applicati ons Volume: 25, Issue: 3, October, 2003, pp. 431-447 Ya-pei Lin ( 林林林 )

Transcript of Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with...

Page 1: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

Ontology-based fuzzy event extraction agent for Chinese e-

news summarization

Expert Systems with Applications

Volume: 25, Issue: 3, October, 2003, pp. 431-447

Ya-pei Lin( 林雅珮 )

Page 2: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

2

Introduction

• Ontology collection of – key concepts– interrelationships collectively

• User and system can communicate with each other by the shared and common understanding of a domain

Page 3: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

3

Motivation

• Prohibited an easy access to the right information.

• Spend a lot of time manually sifting out useful or relevant information.

Page 4: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

4

Goal

• Summarization is to take from the extracted content

• Present the most important to the user in a condensed form.

Page 5: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

5

Solution

• Ontology-based Fuzzy Event Extraction (OFEE) agent

• The OFEE agent consists – Retrieval Agent (RA)– Document Processing Agent (DPA) – Fuzzy Inference Agent (FIA)

Page 6: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

6

• AAA

Page 7: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

7

• AAA

Page 8: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

8

• AAA

Page 9: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

9

• AAA

Page 10: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

10

• AAA

Page 11: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

11

• AAA

Page 12: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

12

Fuzzy Model

• L–R type fuzzy number

m: 指 x 的平均值α: 指 x 的左散度β: 指 x 的右散度

α β

m

Page 13: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

13

• AAA

Page 14: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

14

Fuzzy Inference Agent

• Input linguistic layer• Input term layer• Rule layer• Output term layer• Output linguistic layer

Page 15: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

15

Define of term

• Part-of-speech (POS)• Term Word (TW)• Term Frequency (TF)

Page 16: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

16

• 111

Page 17: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

17

Input linguistic layer

• The input vectors are the term set retrieved from Chinese e-news document and domain ontology.

Page 18: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

18

Page 19: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

19

Input term layer(1/9)

• Three input fuzzy variables – Term POS similarity– Term Word (TW) similarity – Term Frequency (TF) similarity

Page 20: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

20

Input term layer(2/9)

• POS similarity – utilize the length of the path

Page 21: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

21

Input term layer (3/9)

• The path length of the tagging tree is bounded in the interval [0,6].

1 2 3 4 5 6

Page 22: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

22

Input term layer (4/9)

• TW similarity– compute the number of the same

words that different term pairs

Page 23: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

23

Input term layer (5/9)

• The bound of the number of the same word for any Chinese term pair is [0,6].

1 2 3 4 5 6

Page 24: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

24

Input term layer (6/9)

• TF similarity – for every two Chinese terms located

in the retrieved e-news document and e-news domain ontology

Page 25: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

25

Input term layer (7/9)

• The universe of discourse for TF similarity interval is [0,1].

0.2 0.3 0.5 0.7 0.8 1

Page 26: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

26

Input term layer (8/9)

transferred

Page 27: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

27

Input term layer (9/9)

Page 28: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

28

Page 29: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

29

Rule layer (1/2)

• The rule layer is used to perform precondition matching of fuzzy logic rules.

Page 30: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

30

Rule layer (2/2)

• Hence, each rule node of the rule layer should perform the fuzzy AND operation.

Page 31: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

31

Page 32: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

32

Output term layer (1/2)

• The output term layer performs the fuzzy OR operation to integrate the fired rules which have the same consequence.

• The fuzzy variable defined in the output layer is terms relation strength (TRS).

Page 33: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

33

Output term layer (1/2)

• TRS fuzzy set

Page 34: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

34

Page 35: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

35

Output linguistic layer

• output linguistic layer – Defuzzification process to get the TRS of t

he Chinese term pair.– Center Of Area (COA) method

Page 36: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

36

Event Ontology Filter (1/2)

Page 37: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

37

Event Ontology Filter (2/2)

• EOF is proposed for getting the extracted-event ontology.

• The EOF utilizes the computing results of FIA and the e-news ontology to extract the e-news event.

Page 38: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

38

Summarization Agent (1/2)

Page 39: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

39

Summarization Agent (2/2)

• Document summarization• Chinese e-news summary

generated • Stored into the repository

Page 40: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

40

Example (1/4)

RA

Page 41: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

41

Example (2/4)

DPA

Page 42: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

42

Example (3/4)

FIA & EOF

Page 43: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

43

Example (4/4)

SA

Page 44: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

44

Conclusions

• In this article, we propose an OFEE agent for Chinese e-news summarization.

• Summarization is most important to the user in a condensed form.

• Topic-focused summary is more suitable for full-text searching, browsing

Page 45: Ontology-based fuzzy event extraction agent for Chinese e- news summarization Expert Systems with Applications Volume: 25, Issue: 3, October, 2003, pp.

Thanks for your listening.

Q & A