Text Mining for the Social Sciences

1
Text Mining for the Social Sciences

• Brian Rea and Sophia Ananiadou
• National Text Mining Centre
• www.nactem.ac.uk
• University of Manchester

2
Outline

• What is Text Mining
• Techniques and Solutions
• How can the Social Sciences Benefit
• Case Study ASSERT Project
• Future Opportunities
• Conclusions

3
What is Text Mining

• Text mining discovers and extracts information
  hidden in unstructured texts.
• It aids to construct hypotheses based upon
  associations between the extracted information
• Due to this it can often discover things
  overlooked by human readers

4
What Text Mining is not

• It is not Google!
• Text mining is not based upon understanding of
  document content.
• Instead it predicts the most likely meaning of a
  fragment of text based upon the language models.
• Text mining will generally not pick up on
  sarcasm, irony or other subtleties of language
  usage.
• Text mining tools must be tuned before use on
  different text types, styles or languages.

5
NaCTeM Our Vision

• to harness the synergy from service provision and
  user needs within biomedicine, allied to
  development and research within text mining
• to provide quality service underpinned from
  proven text mining tools / techniques enabled via
  MIMAS for biosciences in first instance
• to consolidate existing knowledge, activities,
  and models and transfer them widely to other
  areas, e.g. humanities / social sciences /
  clinical, etc.

6
Gathering Information

• Information retrieval
• Gather, select, filter, documents that may prove
  useful
• Information extraction
• Partial, shallow, deep language analysis
• Find relevant entities, facts about entities
• Resources ontologies, lexicons, terminologies,
  thesauri, grammars, annotated corpora
• Tools document converters, sentence detectors,
  tokenisers, taggers, chunkers, parsers, NE
  recognisers, semantic analysers

7
Information Retrieval

• The base technology is similar to that used in
  common search engines.
• This is greatly enhanced by text mining
• Away from common bag of words approaches
• By indexing concepts, entities and facts
• Using pre-processing to improve efficiency
• We use this approach iteratively with other tools
  to minimise intensive runtime processing

8
Terminology and Named-Entities

• Extraction of named entities (names of
  organisations, people, locations), technical
  terms for any domain
• Discovery of concepts allows semantic annotation
  of documents
• Improves information access by going beyond index
  terms
• Links same entities via co-reference (anaphora)
• Enables semantic querying
• Construction of concept networks from text
• Allows clustering, classification of documents
• Visualisation of knowledge maps

9
Terminology Management

• Term clustering (linking semantically similar
  terms) and term classification (assigning terms
  to classes from a pre-defined classification
  scheme)
• Possible applications
• Metadata creation
• Topic detection
• Conceptual indexing (with facts, events)
• Document clustering classification

10
Information Extraction

• Extraction of relationships (events and facts)
  for knowledge discovery
• Information extraction, more sophisticated
  annotation of texts
• Beyond named entities facts, events
• Fact THE QUEEN and the Royal Family remain at
  Osborne
• Event THE DUKE, attended by the Hon. E.C. Yorke,
  honoured the Lyceum with his presence on Saturday
  evening, to witness the performance of The
  Bells.

11
Information Extraction Annotation

• Annotated texts for semantic IR and IE
• Example The GENIA annotation editor
• Mesh term Human, Blood Cells, and Transcription
  Factors.
• Annotation POS, named entity, parse tree
• Any text may be annotated using named entities,
  facts, events
• Consider searching for document about things that
  kill cancer

12
GRID-based Development
70 million seconds, that is, about 2 years
13
Using the GRID

• Challenge
• Analysis of very large data sets
• Combining distributed resources
• Solution
• Large PC clusters
• National infrastructure (NDS)
• Grid enabled software (UIMA)
• Experiments
• The entire MEDLINE was parsed in 8 days

14
How the Social Sciences Can Benefit

• Improved browsing and searching of domain
  specific and general literature.
• Assisted analysis of text-based qualitative data
• Automated extraction of text-based quantitative
  data
• Discovery of key concepts and relationships in
  literature, ideal for virtual learning
  environments
• Discovery of trends and change over time

15
Case Study ASSERT Project

• Automatic Summarisation for Systematic
  Reviews using Text Mining
• Engaging the user community EPPI
• (Evidence for Policy and Practice Information and
  Co-ordinating Centre)
• Document clustering/classification (see demo)
• Information extraction
• Summarisation
• Visualisation

16
Systematic Reviews

• First, extensive searches are carried out in
  order to locate as much relevant research as
  possible according to a query.
• Then the mass of data retrieved by this process
  is screened until only the most relevant and
  reliable literature remains to form the focus of
  the review.
• Finally, the literature is synthesised and
  summary reports are written to inform policy and
  practice by helping users of the research to make
  evidence-informed decisions.

17
Search Solutions

• A combination of Web Crawl and Information
  Retrieval systems to allow for iteratively deeper
  searches.
• Terminology Management to discover key concepts
  for later stages and to improve search criteria
• Clustering Techniques to categorise and collate
  documents on similar subtopics
• Visualisation to allow for improved usability and
  access to documents

18
Screening Solutions

• As above for retrieval, visualisation and access
  to subtopics
• Named Entity Recognition
• Fact and Event Extraction to provide key details
  to the reviewer
• Summarisation techniques to identify significant
  sections of each document

19
Synthesis Solutions

• Multi-Document Summarisation techniques to assist
  with comprehension of the subtopics.
• Fact and Event databases to assist in examination
  and linking of evidence
• Evidence retrieval and reference through existing
  systems to assist in report generation

20
Demonstration

• ASSERT

21
Community Call

• Additional funding by JISC to organise and
  support a community call for the social sciences.
• 360K to be shared between at least 2 projects
  to support expansion of ASSERT and related tools
  for other social science research.
• Specific focus on RoI from assistive technologies
  and significant benefit to existing research.

22
Looking into the future

• NCeSS usability case studies
• Close links with industry AZ, IBM, Xerox
• Involvement with UKPMC
• Pilot project with the BBC
• Not only science
• history, archaeology, humanities, business
• Improving scalability through parallelisation and
  GRID computing
• Further inclusion of Data Mining techniques
• New and extended services for full text analysis

23
BBC pilot project

• Analyse, structure and visualise BBC news online,
  according to a users query using advanced text
  mining techniques
• Concept discovery and retrieval
• interface allows a user to enter a query across
  the document collection and automatically
  calculate a list of concepts specific to the
  query and ranked by perceived importance.
• Creation of user oriented knowledge maps
• Based on clusters of articles and their automatic
  concept categorisation.

24
Sentiment Analysis

• Information from user survey is an essential
  resource for obtaining opinions
• Fixed style questionnaires make it hard to obtain
  free opinion whilst open questions are often hard
  to analyse consistently.

Model ABC BAD Lose battery GOOD style is
nifty MISC. Back panel is heated
Cellular phone ABC loses battery easily, but the
style is very nifty. Back panel is sometimes
heated up.
25
Epidemiology (NIBHI)

• Combines techniques from
• Systematic reviews
• Information extraction
• Knowledge management and inference
• Evidence-informed hypothesis generation
• Filtering to specific location or time period for
  detailed analysis

26
Conclusions

• Text Mining has a proven track record in the
  sciences
• The numerous techniques are fully extendable to
  social science challenges
• The ASSERT project highlights some of the key
  benefits of this assistive technology
• Usability of tools is a key challenge, one that
  is solvable with input from potential end-users
• GRID computing is allowing for novel solution to
  previously impossible challenges

27
How to contact us

• Visit the Text Mining Centre Website at
• http//www.nactem.ac.uk
• brian.rea_at_manchester.ac.uk
• sophia.ananiadou_at_manchester.ac.uk