Discourse Level Software

1
Discourse Level Software

• Current Statusand Future Directions

Nov. 16, 2004 Lars Huttar (lars_huttar_at_sil.org) Kn
owledge Management Services
2
Abstract (I)

• Discourse analysis (DA, a.k.a. textlinguistics)
  is a task frequently cited as needing
  computer-assisted tools.
• Some tools are currently available for certain
  tasks, but as yet, no user-ready applications
  specifically for the discourse charting commonly
  used on the field.

3
Abstract (II)

• This presentation will review a few of the
  existing tools most pertinent to DA on the field,
  and software that is planned or under
  development.
• I will also mention the conceptual model for
  constituent charting described in my thesis,
  which uses XML encoding of text and analysis,
  from which a chart is rendered via XSL.

4
Overview

• The need for discourse analysis software
• Whats already out there?
• Whats coming down the pike?

5
Need for Discourse Software

• The task
• Help the user produce charts, diagrams, and
  summaries of texts in such a way as to facilitate
  discovery of discourse patterns and to expedite
  testing of hypotheses.

6
Major features desired

• Manual annotation of features incl. coherence and
  prominence
• Search text, IT, and annotations
• Chart/summary of results, hyperlinked to data
• Accessible to MTTs/OTTs
• Geoffrey Hunt
• Kent Spielmann

• Import (interlinear) text
• Segment and move pieces into chart columns
• Mark genre(s)
• Configurable auto-highlighting, e.g. color by
  POS.
• Toggle highlighting of certain features

7
Example constituent chart
8
Current Practice

• Pencil paper
• MS Word
• MS Excel
• A few bravesouls useother tools

9
The Right Tools?
Specialized tools could make it quicker and
easier!
10
How to Address the Need?

• Use existing software
• SIL FieldWorks DA tool(s)
• Extend existing tools?

11
Whats already here?

• MDA
• BART
• RSTTool
• MATE
• CiCaDA

12
Multilinear Discourse Analysis

• Generate statistics and diagrams relating to span
  analysis, topic continuity statistics, and other
  issues
• Input is an SFM marked up text (e.g. from Shoebox)

• In Beta 2
• More info phil.quick_at_sil.org

13
Biblical Analysis Research Tool

• BART has features supporting discourse analysis
  of biblical texts
• Comes with extensive built-in morphosyntax
  markup supports customizable tagging and complex
  queries.
• Only for biblical texts cant enter vernacular
  texts.

• Part of TW, or available from WordSearch Corp.
• www.sil.org/translation/bart.htm

14
RSTTool

• Lets user diagram relations between text
  chunks.
• Free download from http//www.wagsoft.com/RSTTOOL
• User can define own set of relations, schemas,
  etc. such as SSA or Longacres propositional
  relations.

• Can generate statistics based on the tree
  structures built by the user.
• File format is XML-based.
• Text can be edited even after struc-turing has
  begun.

15
MATE Workbench

• Tool to aid in the display, editing and querying
  of annotated speech corpora
• Encodes data in XML and displays via XSL-like
  stylesheets could be programmed to produce
  various displays.
• In early demo version (2001). Looks like it has
  potential, but I cant get it to runon my
  machine.
• http//mate.nis.sdu.dk/

16
CiCaDA

• Produce fairly feature-complete constituent
  charts from XML data using XSLT stylesheets.
• Encode text, column assignments, and chart
  configuration in XML chart is produced
  automatically.
• Open standards promote modification/ reuse of
  data.
• There is no application no user-friendly way
  to enter the XML data.

17
Helps available

• LinguaLinks Library has several items, including
• Analyzing Discourse a Manual of Basic Concepts
  Dooley Levinsohn (avail. on the web as well as
  in LLL). Very practical.

18
Do you know of others?

• Please let me know if you are aware of other
  useful discourse-level software tools!

19
Whats coming?

• TCC
• AGTK
• FieldWorks DA tools

20
TCC

• A tool for drawing syntax trees could also be
  used for discourse chunking and highlighting
• Looks very easy to use. Collapsible tree makes it
  easy to browse large text structures.
• Supports Latin-1 charset.
• Author taking feedback to make TCC more useful
  for SILs work.
• Still in beta. No release sched.
• Info http//ulrikp.org/

21
Annotation Graph ToolKit

• AGTK is a toolkit for annotating texts
• TreeTrans edit syntactic trees charting
  chunking possible
• InterTrans interlinearize text (very beta)

• Saves in an abstract XML format potential good
  basis for Lego solution
• Not ready for end users.

22
SIL FieldWorks DA Tool(s)

• FW DA software is still on the drawing board but
  is a high priority.
• Would leverage the huge benefits of all the work
  that has gone into FieldWorks!
• FW tools already support interlinear text, text
  annotations/tagging and highlighting.
• Preliminary work has begun on design of
  constituent charting features.
• Wish list for DA features exists but requirements
  not yet prioritized.Guidance team has not yet
  beenformed.

23
Conclusion

• There are some good tools already out there for
  certain tasks related to DA. Unfortunately they
  dont interoperate much, and there are no
  domain-aware applications for constituent
  charting.
• SIL FieldWorks tools, as they become available,
  should cover certain DA tasks well, such as
  constituent charting.

24
Questions?
Comments?