Machine Translation application PangeaMT: a reference use case

Pangeanic has been invited to showcase its technology as a guest speaker at a ministry-sponsored event, " Multimodal Interaction in Pattern Recognition and Computer Vision (MIPRCV)."

This 5-year project (2007-2012) within the more extensive scientific research program CONSOLIDER Ingenio 2010 proposes significant advances in intelligent multimodal interactive systems in the areas where Pattern Recognition and Computer Vision can be applied. Multimodal interactive systems can be applied to many technologies, like Multimodal interactive transcription of text images, Interactive and adaptive machine translation using a web-based architecture, and several other pattern recognition applications, which include medical scanning, detection of objects on the seabed and, in general, any error or non-pattern detection.  The project aims to add multimodal interaction with the user to the traditional techniques of pattern recognition and computer vision, leading to a synergy between machines and people that has not yet been achieved. One of the applications that provides more general interest in pattern recognition is language, as seen in the last article, and a clear proof of research-to-industry transfer is PangeaMT, Pangeanic's Moses-based DIY environment.

The application has been presented in many fora, including the EU's DGT. The event (in Spanish) will take place on 25th October at the Aula Màster, Campus Diagonal Nord. Edificio A3, Universitat Politècnica de Catalunya, to raise awareness of the project and its many commercially exploitable possibilities in many industrial sectors. The results are divided into three main sections:

• Audio, voice, and language, including machine translation, speech transcription, transcribing music, etc.
• Analysis and Image processing and analysis (transcription of handwritten text images, image retrieval from the web, medical image analysis, etc.)
• Video and robotics for surveillance, advanced driving assistance, robot interaction, people tracking, human action recognition, video retrieval, etc.

In the Demonstrations section of the MIPRCV website, you can find the prototypes that have been developed in the three research areas that the project has been divided into:

• Audio, speech, and language

• Image processing and analysis

• Video and robotics

The dissemination event will take place on the following October 25th in Barcelona. The event's program will be divided into five blocks: an introductory block, three blocks where the most significant results of each project area will be presented, and a final block featuring a round table where several companies from sectors related to MIPRCV will debate about how they see the future of the technologies developed in MIPRCV. As a company related to the project's theme (and one that has already collaborated with the PRHLT group), Pangeanic was invited to attend the event and participate in the round table.

Over the last few years, Multimodal human Interaction (MI) has become an essential field of increasing attention. Advanced man-machine interfaces with high cognitive capabilities are a hot research topic that aims to solve challenging application problems in the information technology society. The outstanding need for the development of interactive systems with cognitive capabilities is reflected, for instance, in the Spanish Research Programme for Information Technologies and in the Information and Communication Technologies Work Programme of the European Union, where these developments are included as one of the priority research challenges for short and middle term research in information systems. The idea of computer interactive systems stems from the early stages of computer science. In the seventies, Jarvis already foresaw a highly relevant role in areas such as Computer Vision (CV) and Pattern Recognition (PR) in the development of advanced man-machine interfaces.

Nevertheless, from those bright promises, only a tiny fraction of the enormous potential of PR and CV in MI has been exploited so far. Tapping this potential entails several research challenges and opportunities to adapt Computer Vision and Pattern Recognition approaches to interactive systems' dynamic and changing environments. The MIPRCV project aims to establish a research program to explore all these opportunities and challenges.

The ultimate goal is to show how existing PR and CV technologies can naturally evolve to help develop advanced multi-modal interactive systems that will realize the long-standing promises of a seamless synergy between persons and machines. The 81-strong team that makes up this research consortium includes highly qualified scientists and engineers. They come from seven research groups and ten different public research institutions. Among them are many internationally renowned researchers in Pattern Recognition, Machine Learning, Image Processing, Computer Vision, Speech and Language Processing, and Robotics.