•New
Share
The first prototype, which will be available in July, will be the first application that translates radio, TV and text between Spain's official languages and English
First multilingual browsing and translation system for audiovisual content
The UPC's Speech Processing Group (VEU) has created the first computer application that automatically translates radio, TV and text between Spain's official languages and English. The aim of the project, dubbed BUCEADOR, is to enhance automatic speech recognition, translation and speech synthesis systems and technology.
01/06/2011
Developed under the BUCEADOR project, led by the Speech Processing Group (VEU) of the Universitat Politècnica de Catalunya. BarcelonaTech (UPC), the new system enables users to browse and translate video and audio files and text documents in Spanish, Catalan, Basque, Galician and English. With the new system, a search for a video or audio file will yield a document that is either translated and synthesised or subtitled in another language, and searches for written texts will yield either another text or an audio file in the chosen language.
This is the first application that translates radio, TV and text documents between Spain's official languages and English. According to Asunción Moreno, coordinator of BUCEADOR, the system can for instance translate a television broadcast from Basque into Catalan in audio and text formats, and searches can be made in either of the languages and results will be available in the same languages.
The project is led by the UPC under the National R&D&I Plan, with the participation of the University of the Basque Country and the University of Vigo, with support from Barcelona Media. A new prototype will be available in July, at the mid-term of research, which is expected to come to an end in 2012.
The ultimate goal of the project is to improve automatic speech recognition, translation and speech synthesis systems and technology. Although the application is designed for the audiovisual industry, according to Moreno it can also be used to browse and translate videos, photographs and texts that are associated with a word into any language.
The UPC uses statistical translation systems because they allow for speech translation in a manner similar to Google Translate and the N-II automatic translator for Spanish, Catalan and English, which was also created by the VEU group. Unlike rule-based systems, statistical systems can translate strings of words, learn from previous translations and specialise in specific registers or areas. Because they are context-based, they would also be capable of discerning whether the Catalan word porter refers to a goalie or a caretaker, for instance.
The new system carries out the following procedure: it transcribes what is said (speech recognition), the resulting text is translated into another language and this text is converted back into speech. This process enables the system to detect pauses, deal with diffluent speech and give the correct intonation and meaning to the message.
This is the first application that translates radio, TV and text documents between Spain's official languages and English. According to Asunción Moreno, coordinator of BUCEADOR, the system can for instance translate a television broadcast from Basque into Catalan in audio and text formats, and searches can be made in either of the languages and results will be available in the same languages.
The project is led by the UPC under the National R&D&I Plan, with the participation of the University of the Basque Country and the University of Vigo, with support from Barcelona Media. A new prototype will be available in July, at the mid-term of research, which is expected to come to an end in 2012.
The ultimate goal of the project is to improve automatic speech recognition, translation and speech synthesis systems and technology. Although the application is designed for the audiovisual industry, according to Moreno it can also be used to browse and translate videos, photographs and texts that are associated with a word into any language.
Translation at the heart of the system
The BUCEADOR project goes a step further than AVIVAVOZ, an earlier automatic subtitling system for Galician, Catalan and Spanish that was run by the UPC, the University of Vigo and the University of the Basque Country. The UPC uses statistical translation systems because they allow for speech translation in a manner similar to Google Translate and the N-II automatic translator for Spanish, Catalan and English, which was also created by the VEU group. Unlike rule-based systems, statistical systems can translate strings of words, learn from previous translations and specialise in specific registers or areas. Because they are context-based, they would also be capable of discerning whether the Catalan word porter refers to a goalie or a caretaker, for instance.
The new system carries out the following procedure: it transcribes what is said (speech recognition), the resulting text is translated into another language and this text is converted back into speech. This process enables the system to detect pauses, deal with diffluent speech and give the correct intonation and meaning to the message.
Follow us on Twitter
