Guy Divita Qing Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

33
1 Guy Divita Qing Zeng-Treitler Salt Lake City VA, University of Utah School of Medicine Pragmatic Interoperability

description

Pragmatic Interoperability. Guy Divita Qing Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine. Toy Story. We all agree one-off-solution is not good Many approaches to Interoperability. Internal Interoperabil ity. Framework Supporting Interoperability. - PowerPoint PPT Presentation

Transcript of Guy Divita Qing Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

Page 1: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

1

Guy DivitaQing Zeng-TreitlerSalt Lake City VA,

University of Utah School of Medicine

Pragmatic Interoperability

Page 2: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

2

Toy Story• We all agree one-off-solution is not good• Many approaches to Interoperability

Page 3: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

3

Internal Interoperability

Page 4: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

4

Framework Supporting Interoperability

Page 5: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

5

Within System Issues

Many Master Builders who do not share a plan!And plan changes…

Page 6: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

6

Within System Issues

Sometimes integration is more work than building from scrtach

Page 7: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

7

Pragmatic Interoperability

Page 8: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

8

Overview

Page 9: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

9

Terminology

Page 10: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

10

Messaging Protocol

Page 11: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

11

Ontological Alignment

Page 12: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

12

Pragmatics

Page 13: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

13

Pragmatics (2)

Page 14: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

14

Feature Alignment Semantics

Page 15: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

15

Feature Alignment Semantics (2)

Page 16: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

16

Expressibility

Page 17: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

17

Expressibility (2)

Page 18: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

18

Expressibility(3)

Page 19: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

19

Pragmatics (again)

Page 20: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

20

Interoperability Efforts within v3NLP

Page 21: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

21

• From survey of existing NLP systems

• Common ontological referent points

• Combines CDA, Penn Treebank labels

• New labels where needed – Document structure (table,

figure, snippet, slot value …

• Rendered into • UIMA Type Descriptors• Protégée Ontology

Common Set of Labels

Page 22: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

22

Interoperability Efforts within v3NLP

Page 23: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

23

Special Considerations

Page 24: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

24

Modeling Negation

Page 25: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

25

Asymmetric Interoperability

Page 26: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

26

Barriers to Interoperability

Page 27: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

27

Solution

Limit the number of moving parts

+

Some planning

Simple protocol

Simple semantics

Page 28: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

28

Pragmatic InteroperabilityPrinciples

Page 29: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

29

• Programs, projects, ideas, utilities, services and shared resources

• workflow aids, annotation editors, evaluation tools, lexicon generation aids, ontology development tools, dashboards

• Repo’s, wiki’s, website(s)• Evolving standards, guidelines, best practices

NLP Ecosystem Elements

Page 30: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

30

• Big goals, grand challenges, many little dance steps

• Engendering developer and user communities• Engendering collaborative environments• Governing body to influence direction• ……

NLP Ecosystem Elements (continued)

Page 31: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

31

One-off code for specific

tasks (GSpell)

Open source software

developed on common

framework (HITEx)

Interoperable and user friendly

systems on common

framework (V3NLP)

NLP ecosystem/Marketplace

Clinical NLP Development Trend

Page 32: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

32

EMMASophia

cTAKESVoogo

ARC

UFIT

Common Labels

Sublanguage Model

UIMA-GATE converter

WEKA

UIMA

U Compare

MetaMap

RapTAT

Mallot

… Toward a NLP Ecosystem

Dave

Page 33: Guy  Divita Qing  Zeng- Trei tler Salt Lake City VA, University of Utah School of Medicine

33

POS Tagger

Shallow NP Parser

Full Parser

Section Tokenizer

Document Classifier

De-Identification

Semantic Classification

Concept Identification

Annotation ToolsCorpus Annotation

Analysis

Negation

Relationship Identification

Temporal Relations

Local Terminology Development Tool

Co-reference Resolution

Word Sense Disambiguation

Document / Section / Sentence/ Phrase

Information Retrieval indexes

Concept: value Identification

Table Tokenization

Figure/Caption Identification

Document Summarization

Multi-Document Summarization –

Novelty Detection

Anaphora Resolution

Text Normalization (stemming)

Spelling Suggestion

Text Simplification

Theme Detection

Concordance

Term Identification

Template Detection

Text-To-Structured Data

GATE

UIMA

Knowtator

NLP Standards

An Ecosystem that includes