Pattern Recognition and Human Language Technology

Research Center

Research

Activities and results in the main PRHLT areas can be found in Projects, Demos, Software, and Publications.

Multimodal Interaction

Technologies to deal with a recent paradigm shift in the design of Pattern Recognition, where the traditional concept of full-automation is being changed to systems where the decision process is conditioned by human feedback. Problems and applications considered within this area include: Relevance-based (image) Information Retrieval and Interactive-Predictive processing for Computer Assited Machine Translation, as well as for the Interactive Transcription of speech audio streams and text images. Read more

Machine Translation [Showcase]

Speech-to-speech translation or text-to-text translation for limited domains fall within these kind of projects. Finite-state and statistical transducers are used as the basis of the machine translation systems. These models can be learnt automatically from real examples of translation. Some applications included (but are not limited to) translation of technical reports, hotel services, etc. Read more

Handwritten Text Recognition [Showcase]

Both off-line (document images) and on-line HTR (tablet or e-pen signals) are considered. No prior character or word segmentation is needed. Technology, borrowed from Speech Recognition, relies on character Hidden Markov Models, Finite State word models, and syntactic N-Grams. After model training, for each given text line image, a holistic ("Viterbi") search provides both an optimal transcription and the corresponding word and character segmentations. Applications: Transcription of ancient and legacy documents, transcription of unconstrained handwritten text in survey forms, etc. Read more

Image Analysis

Identification of the objects in an image. Statistical and Syntactic Pattern recognition techniques are used. Applications: OCR and document analysis, medical diagnosis, biometric identification, image and video retrieval, classification of chromosomes, aids for the handicapped, manufacturing quality control, etc. Read more

Automatic Speech Recognition

The speech utterances are decoded into strings of words or into strings of semantic units. Finite-state grammars are used as the basis of such systems. These finite-state grammars are learnt automatically from real examples of utterances or text. Applications: telephone exchange services, device control by voice, information queries, etc. Read more

Natural Language Processing

We propose a principled solution to handle the cross-script term matching and spelling variation where the terms across the scripts are modelled jointly in a deep-learning architecture and can be compared in a low-dimensional abstract space. Read more