Round Table

  • Print

The interfaces of the speech: linguistic theory and language technology

  • PRESENTATION - PDF
  • Speaker: Professor Sara Candeias (IT / DEEC-UC)
  • Abstract: In the amazing world of today, which increasingly defines a society eager for information under the aegis of the technological revolution, speech and language technologies are of key importance and growing as it supports dialogue between man and machine and make information processing easier. But machines do not share the same abilities of humans: to be communicative in nature, to analyze and segment reality in order to convey it and interpret it linguistically, either by voice or by text. For the machines to 'talk' and 'hear', like humans, they must be taught, programmed, trained. It is necessary to simulate neural networks in machines that allow humans to choose the language units that let you build relevant messages and understand information. This presentation aims at showing how the theoretical knowledge provided by the language can be integrated in automatic processing of the Portuguese language, particularly in the aspect of speech processing. It is also to describe some experiences that result from the confluence of linguistic studies and engineering and electrical engineering and computers that have been experienced by the Signal Processing Lab of the Institute of Telecommunications (University of Coimbra - Portugal) targeted at equipping machines with the ability to process Portuguese. At the same time, a new sense of linguistic study will be recognized: the engineering of language or language processing technology.

Building Applications Using NLP and Resources for Speech Recognition and Synthesis Provided by Project FalaBrasil UFPA

  • PRESENTATION - PDF
  • Speaker: Professor Aldebaro Klautau (Federal University of Pará)
  • Abstract: In the last years there has been a proliferation of applications using speech recognition and synthesis techniques in conjunction with natural language processing (NLP). This is because the speech synthesis (SF) reached a level of quality that the synthesized voice is often perceived as human. And speech recognition (RF) has reasonable performance today, having become popular with applications like Apple's Siri, which uses Nuance's technology for RF distributed, ie, using "cloud computing". For example, the use of PLN, RF and SF to allow communication of speakers of different languages has become part of the routine. A scene now common in some countries: a person speaks in English with his/her "smartphone", the parameters representing the voice is sent to a server that uses RF to convert them to text, an NLP system performs automatic translation of English to Spanish, the text is converted into a speech signal through SF in Spanish and sent to the smartphone that requested the service. The reverse process enables the conversation. In this context, the objective of our participation is to briefly review the SF and RF technologies, focusing on developing applications that use them together with PLN. In order to provide a basis for discussion, we will seek to identify areas of joint work of engineers and linguists in developing solutions and research covering the areas mentioned. We will also present the lessons learned over the past eight years developing resources for the RF and SF in Brazilian Portuguese, aiming its free distribution and promotion of research and Brazilian industry in this area. By comparing the resources available within the project FalaBrasil and the state of the art for the English language, we aim to give a clear indication of the maturity of these resources and possibilities of its adoption together with PLN. Finally, we will indicate emerging research topics and future scenarios.

Assistive Technologies and Robotics Education in Computational Linguistics

  • Speaker: Luis Felipe Uebel, PhD (ASR LABS)
  • Abstract: Recognition and speech synthesis and machine translation are being used to develop applications that help people with special needs to relate with others with or without disabilities, as well as control equipment and receive information in a more friendly and natural way. The RSA Labs has developed assistive technologies in order to improve the quality of life of these people, especially with visual impairment and deafness / muteness. Solutions for the control of equipment, computers and software the company developed with the assistance of FAPESP and CNPq will be presented during the ELC 2012. Speech technologies are being employed by the company in the development of educational applications with the help of robotics. Studies show computational linguistics that can assist in setting educational content areas of mathematics, physics and science. The project was sponsored d by FINEP.

VOCALIZE: experiences and opportunities of a Brazilian technology company of speech and language

  • Speaker: Edmilson Morias, PhD and Jussara Melo Vieira, PhD (vocalize - Speech and Language Technology Solutions)
  • Abstract: In this roundtable, VOCALIZE (www.e-vocalize.com.br) will present its experience in developing solutions, products and technologies of speech and language; more specifically, the conversion of text to speech, automatic speech recognition, audio transcriptions, speech analytics and computational linguistics in Brazil. There will also be demonstrations of VOCALIZE technologies / solutions aimed at market-oriented customer service, digital marketing, digital TV, tablets / smartphones and accessibility. Since 2006, vocalize has generated opportunities for professional engineers, linguists, computer scientists, mathematicians and audiologists by working in interdisciplinary teams in the Brazilian market for speech technology and language - yet to be further explored, especially considering the performance of Brazilian companies.


Translation Disclaimer

The texts on this page were sent by the authors originally in Portuguese. Translations were done by volunteer translators. Though they were not meant to be official translations, reasonable efforts were made to provide accurate information. Still, portions may be incorrect or inaccurate. Authors have not been asked to validate the translated texts.