Entity Linking via Joint Encoding of Types, Descriptions, and - - PowerPoint PPT Presentation

▶
entity linking via joint encoding of types descriptions
SMART_READER_LITE
LIVE PREVIEW

Entity Linking via Joint Encoding of Types, Descriptions, and - - PowerPoint PPT Presentation

Aug 04, 2023 3.13k likes •3.39k views

Entity Linking via Joint Encoding of Types, Descriptions, and Context Author: Nitish Gupta, Sameer Singh, Dan Roth Source: EMNLP17 Speaker: Ya-Wen Yu Date: 2018/7/31 1 Outline Introduction Method Experiment Conclusion

slide-1
SLIDE 1

Entity Linking via Joint Encoding of Types, Descriptions, and Context

Author: Nitish Gupta, Sameer Singh, Dan Roth Source: EMNLP’17 Speaker: Ya-Wen Yu Date: 2018/7/31

  • 1
slide-2
SLIDE 2

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion
  • 2
slide-3
SLIDE 3

Introduction

  • Challenge in Entity Linking

(mention)

(entity)

  • 3
slide-4
SLIDE 4

Introduction

  • Goal - How to correctly resolve the mentions
  • local context
  • textual descriptions
  • fine-grained type
  • 4
slide-5
SLIDE 5

Training Data

Wikipedia anchors as mentions, links as true entity Description: first 100 tokens of the entity’s Wikipedia page Type: 112 fine-grained entity-types from Freebase

  • 5
slide-6
SLIDE 6

Framework

  • 6
slide-7
SLIDE 7

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion
  • 7
slide-8
SLIDE 8

Method

Mention Context Encoder, C

Local-Context Encoder Document Context Encoder

  • 8
slide-9
SLIDE 9

Method

Mention Context Encoder, C

Context Objective

  • 9
slide-10
SLIDE 10

Method

Entity Description Encoder, D

  • 10
slide-11
SLIDE 11

Method

Fine-grained Type Encoding, E

Entity-Type Objective

  • 11
slide-12
SLIDE 12

Method

Type-Aware Context Representation, T

Mention-context representation aware of entity type information should be helpful

Learning Unified Entity Representation

  • 12
slide-13
SLIDE 13

Method

Entity Linking

Top-30 Candidate entities (Cm) from Cross-Wikis

  • 13
slide-14
SLIDE 14

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion
  • 14
slide-15
SLIDE 15

Experiment

Datasets

CoNLL-YAGO (Hof- fart et al., 2011), ACE 2004 (NIST, 2004; Rati- nov et al., 2011), ACE 2005 (NIST, 2005; Ben- tivogli et al., 2010), Wikipedia (Ratinov et al., 2011).

  • 15
slide-16
SLIDE 16

Experiment

Entity Linking Performance

  • 16
slide-17
SLIDE 17

Experiment

Entity Linking Performance

  • 17
slide-18
SLIDE 18

Experiment

Cold-Start Entities

  • Dataset- 3791 mentions of 1000 rare entities

from Wikipedia

  • New entities-Unknown during training. No

linked mentions

  • 18
slide-19
SLIDE 19

Experiment

  • 19
slide-20
SLIDE 20

Experiment

Typing Prediction

predict fine-grained types for mentions from context (using vm and vt).

  • 20
slide-21
SLIDE 21

Experiment

Example Predictions

  • 21
slide-22
SLIDE 22

Outline

  • Introduction
  • Method
  • Experiment
  • Conclusion
  • 22
slide-23
SLIDE 23

Conclusion

  • Jointly learn entity embeddings using multiple information

sources and leads to more accurate entity linking.

  • Unstructured: KB Description, Mention Contexts

(Wikipedia)

  • Structured: Fine-grained Entity Types (Freebase)
  • Robust to missing or incomplete information: Not all 


information sources needed for all entities

  • Cold-start entity linking: Accurate entity linking also to

entities that were never observed during training 


  • 23