Эйден (Serge Heiden). Франция Лион. École Normale Supérieure de Lyon: лаборатории ICAR
Textometry has been developed in France since the
70's through pioneering works in statistical analysis of lexical data (P.
Guiraud & C. Muller) and general data analysis (J-P Benzécri).
Textometry combines several computerized tools to
assist the systematic analysis of large bodies of digitized texts in any
language. Texts are organized in sets called corpora. They can come from
written language productions (books, news, letters...) or from transcriptions
of speech (political discourse, interview...). They are digitally encoded
according to some conventions like XML encoding according to the TEI guidelines
that can be used in the analysis. They contain words that are used to access
their content and style. Words are automatically segmented and tagged with
properties like a grammatical category or a lemma. Documentary tools use any
combination of those word properties to list their patterns and display in
which contexts they occur (frequency lists, concordance and edition display).
Statistical tools apply statistical models to the counting of word properties
to analyze their distribution across corpora (factorial analysis, clustering),
their over- or under-representation in some sub-corpus (specificity analysis)
or to analyze attraction between word patterns (cooccurrence analysis).
All statistical tools results are related to
documentary tools to give access to the detail of any textual event. As a
result, textometry provides a comprehensive set of highly synthetic or very
precise views of text contents.
TXM implements the textometry tools as an end user
desktop software for Windows, Mac or Linux. Its Graphical User Interface
provides common services for interacting with the software: menus to launch
commands, views to display corpora, text editions and tools results. At the
heart of TXM, the CQP search engine helps the user to list any word patterns
occurrences in a corpus to list, count and display them. The workshop
introduces sequentially to the following concepts and tools:
- whole lexicon and frequencies display;
- index and frequencies of some patterns expressed by
CQL expressions for the search engine;
- concordance display of CQL patterns;
- reading of text editions;
- progression analysis of some CQL patterns;
- sub-corpus building;
- specificity analysis;
- cooccurrence analysis.
Each command result, as a table of numbers or as a
graphic, can be exported to be analyzed or edited by another tool.
TXM version as a web portal software - to allow on
line access - is finally introduced.
To be able to be analyzed by TXM, a corpus must first
be imported into it.
TXM can import and analyze two main types of textual
- written texts: which can be the result of an OCR
process with the associated images of the original texts, and can have various
levels of encoding (from raw text as a base stream of characters to XML encoded
representations according to the TEI guidelines)
- speech transcriptions: in which the speech turns of
each participant is delimited and can be synchronized to the original audio or
The result of any import process into TXM is a new
corpus containing the elements of its corpus data model:
- a corpus contains several texts that can each have
several properties (called metadata), such as an author, date, title, genre,
- a text can contain several internal structures that
can each have several properties, such as a chapter, section, paragraph, speech
- a text contains words that can have several
properties, such as the graphical form, the grammatical category and lemma
The workshop introduces progressively to the following
concepts and tools:
- simple import from the system clipboard (text copied from any
application: word processor, browser, mail client, etc.);
- import of TXT raw text files and tuning the language for automatic
lemmatization of words;
- import of TXT raw text files and tuning texts metadata;
- import of XML files and tuning internal structures;
- import of XML files and tuning some specific words pre-encoding.
For each level of sources imported, the different elements of the corpus
data model available in TXM are verified and manipulated by tools to show the
trade off between the work of source texts encoding and what is necessary for
the analysis of a particular corpus using TXM
to textual data analysis with textometry»
часов лекционных занятий и практикумов, сроки:
19-21 ноября 2015 г.)
Лекция 1: Introduction to textual data analysis with textometry.
Практикум 1: TXM software
initiation hands-on workshop
Практикум 2: Textual corpora
preparation & importation into TXM software hands on workshop
1. Heiden, S.
(2010). The TXM Platform : Building Open-Source Textual Analysis Software
Compatible with the TEI Encoding Scheme. In K. I. Ryo Otoguro (Ed.), 24th
Pacific Asia Conference on Language, Information and Computation - PACLIC24 (p.
389-398). Institute for Digital Enhancement of Cognitive Development, Waseda
University, Sendai, Japan. <http://halshs.archives-ouvertes.fr/halshs-00549764/en>