Centre d’innovació i Tecnologia BARCELONATECH


Language and Speech Technologies and Applications

The Center for Language and Speech Technologies and Applications (TALP) is specialized in the automatic processing of written and spoken natural language. Its aim is to help overcome language barriers by facilitating interaction between people, and between people and devices, and increasing the accessibility of multilingual information.

TALP UPC was founded in 1998, is comprised of a fifty-person team and is a member of the Catalan Cluster of Language Industries(CISTERLINGUA).

The Center works in the following research and development areas:

Automatic speech recognition

Transforms voice into text, which enables machines to understand what a person says and to interpret voice commands.

Text mining

Uses knowledge contained in texts for various purposes: to recover information, to extract information (to structure the unstructured knowledge within a text, to extract facts, events or relevant characteristics that are described in a text), to summarize automatically (to summarize one or more documents, and thus make it easier to access and understand them), to answer questions (to find the answer to a specific question in a set of documents).

Voice and audio recognition

Identifies the person who is speaking by characteristics of their voice (voice biometrics); identifies the language or dialect that is being spoken; detects the emotional state of the speaker; distinguishes whether a sound is speech, music, applause, background noise or another kind of signal; filters relevant acoustic signals in noisy environments.

Linguistic processor

Analyses the linguistics of texts, which is essential in any application that requires a certain degree of language comprehension. Linguistic processors carry out: morphological analysis, morphological and syntactic disambiguation, detection of groups and syntagms, construction of a syntactic tree, semantic analysis (disambiguation of meanings, roles and co-reference).

Speech synthesis

Converts text into speech, to generate intelligible, natural prosody that is appropriate for each application.

Human-machine dialogue systems

Can converse with a human for a specific purpose (to provide the required information, to make a sale or place an order), using coherent structures.

Automatic translation

Translates a text or speech from one language to another, to help people to access multilingual information; automatically subtitles audiovisual material.

TALP UPC is one of the main contributors to the computing cluster of the UPC’s Department of Software (LSI). Since 2004, it has had a smart room, which enables speech to be studied in a more natural context. It also has a multimedia laboratory in the Department of Signal Theory and Communications (TSC), equipped with an audio and TV recording studio.

TALP’s technology has led to two spin-offs: Verbio Technologies (founded in 1999) and HERTASECURITY in 2009).

The sectors in which these technologies can be applied, include the following:

Written and spoken communication mediaOnline opinion gathering and recommendation of contents or productsOnline reputation analysisVirtual assistants (language learning; online purchases; automobile, medical and tourism sector, among others)Assistance for disabled people with cognitive, speech or hearing problemsText mining in different domains (for example, medical and financial)Translation assistance servicesCustomer careCall centersSecurity and acoustic monitoringEntertainment industry: including dubbing films and subtitling

http://www.talp.upc.edu/ | info.cit@upc.edu | +34 93 405 46 90 | +34 93 405 44 19

  • Second voice

    TALP UPC develops various speech technologies to apply in the medical sector. The aim is to provide a voice synthesizer for patients whose voice was totally or partially lost after surgery that affected their vocal chords. The project is developed as part of the INNPACTO program in conjunction with the universities of Vigo and the Basque Country. TALP UPC’s work focuses on synthesizing the voice of each person by applying connectionist learning models (e.g. neural networks) to the automatic recognition of the speech of people whose voices have been distorted as a result of surgery.

    Voice 1 - Voice 2

  • Technology for translating patents

    The Center for Language and Speech Technologies and Applications (TALP UPC) has participated in the development of a set of tools to produce real-time, high quality translations of texts into different languages. On the basis of hybridization between syntactic-semantic grammars and statistical machine translation, TALP UPC has created a tool for the automated translation of patents in three languages (English, French and German). This technology has been developed as part of the MOLTO European project.

  • New Digital Observatory

    Researchers at the Center for Language and Speech Technologies and Applications (TALP UPC) have taken part in a pilot project with the Spanish National Reference Centre for the Application of ICT Based on Open Source (CENATIC) and the Ministry of Industry, Energy and Tourism’s Secretary of State for Telecommunications and Information Society (SETSI), in the framework of the "Innovation Support Platform" that these organizations are developing. The aim of the platform is to enable the institutions to take the pulse of industrial sectors; specifically the ICT sector in the current pilot project. To achieve this, thousands of sector-related documents (patents, competitions, calls for applications and contracts, among others) are monitored automatically to create a digital observatory of companies’ activities that can detect new products and innovative initiatives. TALP UPC has adapted its text analysis tools to organize the information in the texts under analysis. As a result, texts can be processed using document classification or datamining engines.

  • Speech synthesiser for patients with dysarthria