The workshop will explore evaluation methodologies for multilingual OCR systems. By `multilingual' we mean to include systems that are capable of reading more than one language in the same document, as well as one-language-per-document systems that can be easily retargeted to new languages. We hope to bring together researchers from many countries to discuss these and related topics:

• What methodologies should be used to evaluate multilingual OCR systems?
  How do we compare accuracies across languages?
• What ground-truthed data sets are now available in various languages?
  What kind of datasets need to be collected?
  How is this to be achieved?
  What organizations might be willing to support such an the effort?
• What multilingual OCR evaluation tools and error visualization tools
  are available or should be developed?
• What OCR evaluation methods and metrics will be useful for OCR-based
  machine translation and cross-language information retrieval?
• What kind of multilingual operating systems, editors, codesets, and environments are appropriate for OCR?
• What is the market for multilingual OCR/IR technology?
• Open research problems, promising dissertation topics, etc.

This will be a one-day workshop for a maximum of 70 participants. Each participant will submit an extended abstract which will be distributed at the Workshop. All participants are expected to contribute to the discussions.
At the outset of the workshop, three volunteers will present brief, informal summaries of the

• methodologies,
• corpora, and
• tools mentioned in the submitted abstracts

After the workshop, the organizing committee will compile a Workshop Summary, based on the working group notes, and make it available on the Web. It is hoped that the workshop will stimulate cooperative follow-on activities that will accelerate the pace of research in multilingual document image analysis.

Each potential participant or group of participants should submit an extended abstract, electronically via E-mail (in plain ASCII), no later than April 30, 1999 to:

Tapas Kanungo
Center for Automation Research
University of Maryland
College Park, MD 20742
Email: mlocr@cfar.umd.edu

The abstract should include the name, address, telephone, fax, and email address of the author(s). It should ordinarily be limited to six printed pages including references (no figures, please). Longer submissions may be admitted in special cases, e.g. for catalogues of resources. Accepted abstracts will be distributed at the Workshop and posted on the Workshop website.

Postscript Version