Obiettivo
This project attempts to integrate work in the fields of:
automatic retrospective conversion of library catalogues;
OCR/ICR technology;
natural language processing.
This project attemps to define and implement tools for intelligent recognition, analysis and transformation of information contained in a variety of library type documents.
The project developped a toolbox which allows analysis of the field/subfield structure underlying bibliographic references, dictionary entries, indexes of scientific periodicals etc. An intelligent document recognition system has been developed to enhance existing OCR/ICR technology by better image pre-processing, segmentation and incremental feedback from analysis of the documents. A structured analysis procedure allows the breakdown of texts into shorter units until individual informational elements are made explicit. These elements are then transferred to SGML which enables a normalised structure model to be developed.
The toolbox has been tested on indexes and references in scientific periodicals, tables of contents and catalogue cards.
Impact and results:
The BIBLIOTECA toolbox will substantially decrease the cost, time and effort involved in creation and update of bibliographic databases, by the substitution of manual analysis and key-board entry with intelligent document reading, using scanning, OCR/ICR and artificial intelligence techniques.
The benefits and results of this project include:
creation of keyword indexes from table of contents and indexes in books;
creation of article databases from content pages in serials;
creation of citation indexes from bibliographic references;
investigation of the possibility of more advanced 'intelligent' systems for indexing and classification;
automatic transformation of card files into standard formats.
Deliverables
The main deliverable from this project is a system which can:
produce keyword indexes from tables of contents and book indexes;
create article databases from contents pages in serials;
generate citation indexes from bibliographic references;
provide bibliographic databases from printed bibliographical dictionaries;
automatically transform card files into standard formats.
Development software and certain specifications, lists and technical documents were categorised as restricted.
Deliverables in the public domain are:
Detailed and top level work plans;
Technical reports: text selection criteria, draft framework, field structure, integrity criteria;
Appraisal: strengths, weaknesses, costs, benefits, performance;
Testing, examples and results;
Project reports (including final).
Technical approach:
The project was organised into seven workpackages:
Corpus selection and analysis;
Intelligent document recognition, enhanced with pre-processing functions, high level character segmentation and incremental processing;
Automatic field compositional analysis - outputting field parsers for different document classes;
Context based error detection and correction;
Conversion to SGML and translation to bibliographic database, CD-ROM and MARC variant;
Training, testing and evaluation;
Dissemination of information and exploitation.
The project used a 'rapid prototyping' methodology for software development.
Key issues:
BIBLIOTECA integrated work in the fields of:
automatic retrospective conversion of library catalogues;
OCR/ICR technology;
natural language processing
to develop a toolbox for analysis of informal and formal field/subfield structures underlying indexes of periodicals, dictionaries, tables of contents and bibliographical references.
It then resolved the following issues:
Adaptation of OCR/ICR to librarian needs and upgrade of image pre-processing, segmentation and feedback from document analysis.
Devising a flexible system for analysis of semi-formatted documents, with contextual error detection, correction and feedback.
Breakdown of texts into explicit elements suitable for transformation into SGML (Standard Generalised Markup Language), which can facilitate conversion to other formats, such as MARC.
Documentation is available from the contact below and from http://www.csic.es/cbic/teca.htm.
Campo scientifico
- natural sciencescomputer and information sciencesartificial intelligence
- natural sciencescomputer and information sciencesdatabases
- natural sciencescomputer and information sciencesdata sciencenatural language processing
- natural sciencescomputer and information sciencessoftwaresoftware development
- humanitiesother humanitieslibrary sciencespublic libraries
Argomento(i)
Invito a presentare proposte
Data not availableMeccanismo di finanziamento
Data not availableCoordinatore
28040 Madrid
Spagna