Abstract
Optical Character Recognition (OCR) systems offer poor performance when used on handwritten text or printed text document collections with special characteristics, such as historical archives. Because of their flexibility, recognition engines capable of learning from examples are an appropriate solution for transcribing text in restricted domains or with unique characteristics. These engines can be advantageously integrated into interactive Computer Aided Transcription systems, even if they do not achieve performance comparable to that of ROC systems for modern typefaces. In this field, our group has more than 15 years of research experience in speech and writing technologies.
Scientific officer
Stakeholders
Applications
Integration of written interaction systems in devices.
Technical advantages
Facilitate accessibility to information services using computerized systems.
Benefits it provides
- Expand the accessibility of the services that a company may present and that require interaction with users.
- Expand digitized bibliographic collections.
Relevant experience
The Form Recognition and Artificial Intelligence (RFIA) research group has more than 15 years of experience in speech and writing technologies research. It develops its activities in the Department of Computer Systems and Informatics (DSIC) of the Polytechnic University of Valencia (UPV). From the theoretical point of view, the work of the group has focused on the development of inference algorithms for certain types of formal grammars from examples for machine learning of structural models. Some results on stochastic models have also been obtained by extending the previous results. From the practical point of view, the above techniques have been applied to Acoustic-Phonetic Decoding and language modeling in Continuous Speech Recognition systems. Recently, some of these techniques have also been successfully applied to the development of trainable Language Understanding and Translation systems in limited domain tasks.