Visualization of Relational Text Information

1
Visualization of Relational Text Information

• for Biomedical Knowledge Discovery

James W. Cooper IBM T J Watson Research
Center Hawthorne, NY
2
Overview

• Prior work
• Java based text mining
• Computation of unnamed relations
• Graphical display of relations

Text
Text
Text
Text
Text
Text
Text
Text
Text
3
Relations between terms

• Noun phrase co-occurrence statistics Roark,
  Charniak
• Choose seed words and look for terms near them.
  Brin Gravano, Agichtein
• Repeat
• Biomedical domain
• Blaschke used dictionary of common verbs
• Pustejovsky found inhibit relations
• Stevens, Palakal, Mostafa
• Detected abstract-wide co-occurrence using
  dictionary of genes and useful verbs.

4
Graphical Displays

• Biolayout protein similarity
• ProtInAct interactive system using yFiles
• Zhang interactive 3D system
• Jenssen gene network
• Leroy GeneScene

5
BioLayout Enright and Ouzounis
Five related protein families and their
corresponding relationships.
Spheres represent proteins and lines represent
protein similarities.
6
ProInAct- Spencer and Bennett
Proteins clustered by functional interaction
7
Zhang-Protein interaction mapping
8
Jenssen A literature network
Lines connect genes that have co-occurred in 1 or
more papers.
9
Leroy GeneScene
10
What would we like to do?

• Find scientifically meaningful connections
  between important terms.
• Such as Swansons Reynauds disease fish oil
  connection.
• Allow exploration of relations by user.
• Filter the relations by ontology or term types
• Perform path analysis
• Let the user vary the graphical display.

11
Data we analyzed

• Two sets of patent data
• 584 patents on Viagra and phosphodiesterase
  inhibitors.
• 1514 patents on quinolones (like Cipro)
• Recognized major technical terms in each patent.
• Filtered organic chemical nomenclature.

12
The Talent text mining system

• Text Analysis and Language Engineering Tools
• Finds multiword noun phrases
• Does shallow parse
• Can extract NPs and VGs
• As well as all other sentence parts

13
The JTalent Library

• Java class library with JNI interface
• To Talent DLL
• Creates database load files of terms
• Paragraph
• Sentence
• Offset
• Term type (NP, VG)

14
TalentShow Demo
15
The KSS Library

• Java class library of functions for
• Accessing a database (DB2, Access)
• Manipulating a search engine
• Manipulating tables of information created by
  JTalent.

16
Database Tables

• Documents
• Title, author, URL, ID
• TermDocs
• Term
• Paragraph
• Sentence
• Offset
• Type
• Dictionary of terms, types and IDs
• Such as MeSH

17
Computing term information

• Compute unique terms from Termdocs
• Compute frequency
• Compute salience
• Based on frequency
• Number of docs they appear in more than once

18
Compute term relations

• Named relations based on abbreviation expansions.
• Unnamed relations based on proximity, with weight
  based on how frequently they occur near each
  other.
• Mutual information weight

19
Tuning Computed relations

• Select only terms above a salience threshold.
• Only relations in which one or both are members
  of an ontology.
• Store relations in a database table for rapid
  access
• Term weight term

20
Original System

• Visual client
• SOAP server
• Queries database to get relations
• Round trip for each new query
• Instead, we export the data for the user to
  visualize as they wish.

21
Exporting relations

• Save relations and ontology information in xml
  file.
• ltrelationgt
• lttermgt
• ltiqgt78lt/iqgt
• ltsourcegtMeSHlt/sourcegt
• ltrelationDocumentsgt
• ltdocgt 34lt/doc
• lt/termgt
• lttermgt lt/termgt
• lt/relationgt
• This XML file is a portable version of the
  computed relations that we can then use with any
  number of viewers.

22
A Graphical Relations Viewer

• Creates a Java Relations object for each relation
  it reads from the XML file.
• Inserts them into a Trie structure based on lower
  cased first term.
• If there is already a Relation at that point, it
  adds them to a Vector for that term.
• Creates an alphabetical list of all terms in a
  2nd Trie.

23
Using the Viewer

• When you enter part of a term, it shows all terms
  starting with that fragment in the left list box.
• When you click on a term, it shows all its
  relations in the right list box.

24
Lexical Navigation

• Displays relations between terms graphically and
  allows you to explore them without formulating a
  specific query.

25
Possible enhancements

• Show only terms belonging to an ontology.
• Show only higher IQ terms
• Show the documents the relations occur in.
• Show the ontology reference.
• Show computed paths
• Show more kinds of named relations.
• Inhibits, expresses

26
Evaluations of Information Visualization

• Few, if any, graphical displays have been
  evaluated thus far for effectiveness.
• Usability studies are hard to construct and carry
  out.
• Intuition seems to show
• that exploration may result in discoveries.
• Relations more than one step apart seem best
  displayed graphically.
• Remains to be shown that such visualizations are
  actually useful.

27
Differences in Intent

• Displays may represent information your system
  has discovered.
• Gene protein relations
• Or they may represent data from which the user
  may discover new information.
• New 2nd or 3rd order relationships
• These are rather different applications of
  visualization technology

28
Summary

• Java-based text mining system
• Database of terms and positions
• Computation of relations
• Export as XML
• Graphical relations viewer
• The value of such visual interfaces has not yet
  been established.

29
Acknowledgements

• Bhavani Iyer XML export
• Eric Brown DictMatcher hash code
• Daniel Tunkelang graphical layout
• Bob Mack paper suggestions