TAPoR Tools: Portal text analysis tools and other primitives

poster / demo / art installation
Authorship
  1. 1. Geoffrey Rockwell

    McMaster University, Communication Studies and Multimedia - University of Alberta, Philosophy and Humanities Computing - University of Alberta

  2. 2. Lian Yan

    McMaster University

  3. 3. Stéfan Sinclair

    McGill University, McMaster University, Department of Languages, Literatures & Cultures - University of Alberta

Work text
This plain text was ingested for the purpose of full-text search, not to preserve original formatting or readability. For the most complete copy, refer to the original conference program.


TAPoR Tools: Portal text analysis tools and other
primitives

Geoffrey
Rockwell

McMaster University
grockwel@mcmaster.ca

Lian
Yan

McMaster University
lyan@mcmaster.ca

Stéfan
Sinclair

University of Alberta
Stefan.Sinclair@ualberta.ca

2003

University of Georgia

Athens, Georgia

ACH/ALLC 2003

editor

Eric
Rochester

William
A.
Kretzschmar, Jr.

encoder

Sara
A.
Schmidt

This poster will demonstrate a collection of text processing tools designed
to work through a portal over the Web. The tools are designed to work on
plain text, html or xml encoded e-texts. They are easily used to search
electronic texts without the need to install software, preprocess the texts,
or master complex tools.

HOW DO TAPOR.TOOLS (T.TOOLS) WORK?
T.tools are written in Ruby, an object-oriented scripting language like Perl
and Python.Ruby is available for Macintosh, Unix and Windows at
the “Ruby Home Page”, URL: ,
Accessed Nov. 20, 2002. The T.tools are written so that they can
be run on the command line or as CGI programs off our portal. This means
that users of the tools need not install or maintain them, but if they wish,
advanced users can download and adapt them.
Using Web forms as an interface to the tools gives T.tools the capacity to be
easily adapted to hide complexity or to provide local adaptations. Simple
search and concordance forms can be created that can utilize xml markup
without having to change the tools. Small scale publishers of electronic
texts can provide T.tool Web forms that process their Web accessible e-texts
without installing the software. The forms simply pass the URL for the text
in a hidden field to the appropriate tool residing on our portal for
processing. (We will demonstrate the adaptation of these tools to support
the Hyperliste project, a collection of French medieval poetry online.)

WHAT ARE PORTAL TOOLS?
A portal is an entry point into a field.Katz, Richard N. and
Associates. Web Portals and Higher Education;
Technologies to Make IT Personal. San Francisco:
Jossey-Bass, 2002. In this case the T.tools are written to
provide simple text processing tools for TAPoR (Text Analysis Portal for
Research), a multi-institutional project which has Canada Foundation for
Innovation funding to create a portal for text analysis.Text
Analysis Portal for Research, URL: ,
Accessed Nov. 21, 2002. It should be noted that in this poster
presentation we will not be presenting on the portal as a whole, just a
specific set of tools designed to work through (or not) the portal that
is being implemented.They are designed to provide a suite of
simple text transformations that will eventually be managed by a portal
environment that additionally provides user and interface customization
tools. At present they can do the following:
List and count words in a text.
List and count elements in an xml text.
List attributes and values in an xml text.
Extract elements from an xml text.
Find patterns (words or phrases) in a text.
Find patterns in specific elements in a text.
Create a concordance of found patterns or elements.
Output results in either html for reading or xml for further
processing.

This combination of functions allows the user to query an xml text to find
words in specific parts of text or to extract selected parts by element name
and attribute value. Users can also, should they not know the structure of
the text, get a list of the elements or a list of words to search for. Users
have a choice of output from simple output in html to output in xml that
could be saved and processed locally.

WHAT TYPES OF USERS ARE ENVISAGED?
T.tools are designed to be used in three ways by users of differing levels of
expertise.
Introductory Users. A portal should be place one can
learn through playful discovery about a field like computer assisted text
analysis. T.tools are designed so that new users can try basic operations on
electronic texts without having to install software or texts. As the T.tools
do not preprocess texts they can be run on any text a new user can find on
the Web. This allows a new user to experiment with text analysis on texts
they know without much training. The tools are also designed so that they
can be explained and documented in different ways to make them accessible to
different communities.
Small E-text Publishers. While large e-text projects
have access to programmers and systems that allow them to adapt text
processing tools to their texts, many small projects cannot afford to do
more than make available their scholarly texts on the Web in html or xml/css
form. T.tools provides tools run on our server which can be passed a text
(actually a URL) for processing from a Web form set up by the publisher.
Thus small projects can adapt our forms to their needs and integrate them
into their sites.
Advanced Users. As the code is made available as
“open source” according to the definition at the Open Source Initiative,
advanced users can download it and adapt it to their research needs.”The Open Source Definition”, URL: ,
Accessed Nov. 20, 2002. T.tools have been built around a library
of object classes commonly used in text processing whose methods can be
called in new Ruby scripts, from the command line, or through IRB
(Interactive Ruby). Thus the advanced user can use T.tools as a text
processing language and then build new scripts to do things unanticipated by
the developers.

WHAT ARE THE PROBLEMS WITH THIS MODEL?
The major drawback to this model is that T.tools are slow by comparison to
other Web text tools like TACTWeb because they do not work with preprocessed
indexes.For more on TACTWeb see URL: , Accessed Nov. 20,
2002. TACTWeb is built on TACT which has a preindexing program MAKEBASE
which prepares the Text DataBase file (TDB) which is then used by TACT
and TACTWeb to quickly process queries. T.tools work best on
chapter to book length texts not on larger corpora. The processing
capabilities of the server used for the portal is also important. TAPoR has
been funded to install high-end servers for the portal that will partially
compensate for the cost of processing, but there is no substitution for more
efficient tools when dealing with large texts. This is an issue which
deserves further study. Further, T.tools is based on a “pipe-and-flow” model
common in the Unix world which may not scale to certain types of interactive
processing needed in the humanities, for example, situations where texts are
being enriched and studied simultaneously.

WHY A POSTER?
This project is presented as a poster as that will provide the most
convenient way to demonstrate the tools to potential users and redevelopers.
Through the venue of a poster session we can demonstrate how T.tools work on
the texts of the poster visitors. It will also allow us to engage humanists
with small text projects that might benefit from such tools. With individual
visitors we can adapt Web forms to show the usefulness of T.tools to their
projects. CD-ROMs of the code and documentation will be distributed to
advanced users.

If this content appears in violation of your intellectual property rights, or you see errors or omissions, please reach out to Scott B. Weingart to discuss removing or amending the materials.

Conference Info

In review

ACH/ALLC / ACH/ICCH / ALLC/EADH - 2003
"Web X: A Decade of the World Wide Web"

Hosted at University of Georgia

Athens, Georgia, United States

May 29, 2003 - June 2, 2003

83 works by 132 authors indexed

Affiliations need to be double-checked.

Conference website: http://web.archive.org/web/20071113184133/http://www.english.uga.edu/webx/

Series: ACH/ICCH (23), ALLC/EADH (30), ACH/ALLC (15)

Organizers: ACH, ALLC

Tags
  • Keywords: None
  • Language: English
  • Topics: None