LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge
description
Transcript of LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge
![Page 1: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/1.jpg)
LINDEN : Linking Named Entities with Knowledge Base
via Semantic Knowledge
Date : 2013/03/25Resource : WWW 2012Advisor : Dr. Jia-Ling KohSpeaker : Wei Chang
![Page 2: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/2.jpg)
2
Outline• Introduction• Approach• Experiment• Conclusion
![Page 3: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/3.jpg)
3
A Real World Entitywith
Different Name
• New York City
• Big Apple
![Page 4: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/4.jpg)
4
Different Entities with the Same Name
Michael Jordan
![Page 5: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/5.jpg)
5
Knowledge Bases• e.g. Yago, DBpedia
![Page 6: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/6.jpg)
6
QuestionMichael Jordan won his first NBA championship in 1991.
Michael Jordan(Person)
m : entity mention e : an entity in Knowledge Base
![Page 7: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/7.jpg)
7
LINDAN framework• Assumption : that the named entity
recognition process has been completed
• Goal : linking the detected named entity mention with the knowledge base
• Tool : Yago, Wikipedia-Miner
![Page 8: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/8.jpg)
8
Outline• Introduction• Approach• Experiment• Conclusion
![Page 9: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/9.jpg)
9
LINDEN Framework
Candidate Entity Generation
Name Entity Disambiguation
Unlinkable Mention Prediction
d
E0
Scorem(e)
d : a document to be processedE0 : All candidate entitiesScorem(e) : Score of entity
![Page 10: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/10.jpg)
10
Candidate Entity Generation
1. Build a dictionary from Wikipedia
2. Lookup the dictionary
• Entity pages• Redirect pages• Disambiguation
pages• Hyperlinks
![Page 11: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/11.jpg)
11
Entity pages
Michael Jordan Michael Jordan(footballer)
![Page 12: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/12.jpg)
12
Redirect pages
![Page 13: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/13.jpg)
13
Disambiguation pages
![Page 14: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/14.jpg)
14
Hyperlinks
![Page 15: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/15.jpg)
15
Look up the dictionary
Count is the number of links which point to the entity.
![Page 16: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/16.jpg)
16
Link Probability
e.g. LP(Michael I. Jordan | m) = 10/(65+10+7+3)
P.S. The candidate entities with very low link probability will be discarded .
![Page 17: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/17.jpg)
17
LINDEN Framework
Candidate Entity Generation
Name Entity Disambiguation
Unlinkable Mention Prediction
d
E0
Scorem(e)
d : a document to be processedE0 : All candidate entitiesScorem(e) : Score of entity
![Page 18: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/18.jpg)
18
Name Entity Disambiguation
Steps :1. Semantic Network Construction2. Semantic Associativity3. Semantic Similarity4. Global Coherence5. Candidates Ranking
![Page 19: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/19.jpg)
19
Candidates Ranking• Feature vector :
• Score :
SVM• a set of labeled documents as
training data• Feature Vector
![Page 20: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/20.jpg)
20
Name Entity Disambiguation
Steps :1. Semantic Network Construction2. Semantic Associativity3. Semantic Similarity4. Global Coherence5. Candidates Ranking ✔
![Page 21: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/21.jpg)
21
Semantic Network Construction
• Tool : Yago, Wikipedia-Miner
![Page 22: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/22.jpg)
22
Example of Semantic Network Construction
![Page 23: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/23.jpg)
23
Steps1. Find candidate entities by the dictionary2. Use Wikipedia-Miner to find the context
concept.3. Find other Wikipedia articles.4. Use Yago to find the taxonomy relations.
![Page 24: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/24.jpg)
24
Name Entity Disambiguation
Steps :1. Semantic Network Construction ✔2. Semantic Associativity3. Semantic Similarity4. Global Coherence5. Candidates Ranking ✔
![Page 25: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/25.jpg)
25
Semantic Associativity
E1 and E2 are the sets of Wikipedia concepts that link to e1 and e2
![Page 26: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/26.jpg)
26
Examples of SmtAssSmtAss(Michael J. Jordan, Chicago Bulls) =
SmtAss(National Basketball Association, Chicago Bulls) =
![Page 27: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/27.jpg)
27
Name Entity Disambiguation
Steps :1. Semantic Network Construction ✔2. Semantic Associativity ✔3. Semantic Similarity4. Global Coherence5. Candidates Ranking ✔
![Page 28: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/28.jpg)
28
Semantic Similarity (1)Given two Wikipedia concepts e1 and e2, we assume thesets of their super classes are Φe1and Φe2 , respectively.
C0
C1 C2 Ca Cb
![Page 29: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/29.jpg)
29
Semantic Similarity (2)
• P(C) is the probability that a randomly selected object belongs to the subtree with the root of C in the taxonomy.
• C0 is the root of the smallest subtree that contains both C1 and C2 in the taxonomy.
![Page 30: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/30.jpg)
30
Examples
• sim(C1, C2) =
• sim(C1, Cb) =
![Page 31: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/31.jpg)
31
Semantic Similarity (3)
the set of k context concepts in Γd which have the highest semantic similarity with entity e as Θk
![Page 32: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/32.jpg)
32
Name Entity Disambiguation
Steps :1. Semantic Network Construction ✔2. Semantic Associativity ✔3. Semantic Similarity ✔4. Global Coherence5. Candidates Ranking ✔
![Page 33: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/33.jpg)
33
Global Coherence
![Page 34: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/34.jpg)
34
LINDEN Framework
Candidate Entity
Generation
Name Entity Disambiguation
Unlinkable Mention
Prediction
d
E0
Scorem(e)
d : a document to be processedE0 : All candidate entitiesScorem(e) : Score of entity
Learn the threshold τ to validate the predicted entity
![Page 35: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/35.jpg)
35
Outline• Introduction• Approach• Experiment• Conclusion
![Page 36: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/36.jpg)
36
Experiment• Data Set : CZ, TAC-KBP2009 data• Using 10-fold cross validation
![Page 37: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/37.jpg)
37
CZ
![Page 38: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/38.jpg)
38
TAC-KBP2009
![Page 39: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/39.jpg)
39
Outline• Introduction• Approach• Experiment• Conclusion
![Page 40: LINDEN : Linking Named Entities with Knowledge Base via Semantic Knowledge](https://reader035.fdocuments.us/reader035/viewer/2022062315/568161cc550346895dd1b4bd/html5/thumbnails/40.jpg)
40
Conclusion• Entity linking is a very important task for many applications
such as Web people search, question answering and knowledge base population.
• This paper, propose LINDEN, a novel framework to link named entities in text with YAGO knowledge base.