International Journal of Document Analysis and Recognition
SPECIAL ISSUE CALL FOR PAPERS
Detection and Understanding of Tables and Forms for Document Processing Applications
In many document image-processing tasks, detection and understanding of tables and forms pose a huge challenge. Detection of tables and forms in electronic documents can be one small part of a series of tasks required for document processing, but often the accuracy of physical and logical layout extraction from these documents depends to a great extent on the accuracy of this solution available to the system. Because errors at this stage tend to get magnified as the subsequent processes are applied in sequence. Beside the obvious 'tables' and 'forms', other manifestations of these structures include 'lists' and 'table of contents' (TOC). Some common features of these structures include, but are not limited to:
– Tabular representation of information
– Coexistence of graphical, syntactical and semantic structure
– Absence of obvious sequential order
– Necessity to understand the whole structure for its interpretation
– Variety of basic construction principles
– Combinations of construction principles
What particular combination of features makes a table, or a form? What is the role of the user in defining these structures (e.g. in industry, "the customer often defines what a table is")? How is a table different from a list? How is a list different from a TOC? Does geometric alignment of content guarantee a robust structure? Or how much semantic relationship needs to be included within the content to qualify a region as such? What is the role of cross-disciplinary techniques used by Information Retrieval (IR) and Natural Language Processing (NLP) communities in understanding tables and forms? These also can be recursive, e.g. a table can contain images, lists and even tables. These are only some of the questions that need answering.
Some applications of this technology may include, but are not limited to, electronic document conversion from one format to another (e.g. PDF to HTML), invoice reading/processing, archiving, alternate views on small screen devices (e.g. handhelds and cell phones), audio readout for very small screen devices (e.g. a watch with internet connectivity), faster understanding of content, locating relevant content and/or finding answers within document repositories etc., to name a few.
High quality unpublished full length papers are sought on any aspect of this subject area, but the following will be particularly welcome:
o Papers which establish some theoretical underpinning
o Papers which demonstrate comparative evaluation of available or new techniques/systems
o Papers which introduce novel detection algorithms
o Papers which describe complete document processing systems with focus on the role of table or form detection and understanding
Please submit your manuscripts (full papers) electronically according to the instructions on the IJDAR web page (http://ijdar.cfar.umd.edu/). Any questions can be sent to the Guest Editors below with a CC to the editorial office (firstname.lastname@example.org). Electronic (email) correspondence is preferred.
Full Paper Submission: June 30th, 2004
Editors' Decisions: January 31st, 2005
Publication Date: March-April, 2005
For additional information, please visit http://www.bcltechnologies.com/rd/special_issue/ or http://www.dfki.uni-kl.de/~klein/IJDAR-special-issue.