Related projects



Eneko Agirre


Knowledge mining is emerging as the enabling technology for new forms of information access and multilingual information access (MLIA, which encompasses both terms), as it combines the last advances in text mining, knowledge acquisition, natural language processing and semantic interpretation. Question answering, information access based on entities, cross-lingual information access, and navigation via crossdocument relations are examples of new applications that are being adopted both by start-ups and consolidated companies such as Google, Yahoo and Microsoft. KNOW2 will emulate and improve current MLIA systems with research to enable the construction of an integrated environment allowing the cost-effective deployment of vertical information access portals for specific domains.

The KNOW project (TIN2006-15049-C03) already enhanced Cross-Lingual Information Retrieval and Question Answering technology with improved concept-based Natural Language Processing technologies. KNOW2 plans to move from general domains to specific domains as a strategy to obtain better performance, and the incorporation of text-mining and collaborative interfaces. In fact the main research objective consists in advancing the state-of-the-art in the integration of text-capture, semantic interpretation, non-standard text treatment (blogs, e-mails, oral transcriptions) and inference and logic reasoning with semantic-based MLIA methods. Given the current state-of-the-art in those areas, we plan to develop intuitive collaborative interfaces which will allow communities of users to improve the systems, including multilingual communities involving Basque, Catalan, English and Spanish.

Regarding the expertise and human resources gathered in this project, rather than just piling upc experts from loosely related areas, we have selected on purpose a coordinated groups of researchers from four groups that together form a virtual research laboratory that gathers the necessary critical mass. KNOW2 is formed by an interdisciplinary group including computer-science expertise on natural language processing and industrial applications, and linguistic expertise on the target languages.

The advances of KNOW2 will be demonstrated by quality publications on top-ranking conferences and journals, as well as demonstrators and prototypes on domains such us environment, European parliament and/or geographic texts, including public portals dedicated to popular science ( and BasqueResearch, part of AlphaGalileo) owned by Elhuyar, which is a KNOW2 partner. The fact that we apply our state-of-the-art research to real scenarios, and the adoption of the last representation standards and free software licenses will facilitate the technology transfer of the developed technology to industrial environments. The large number of EPOs in this proposal, and their level of commitment, already shows the interest that this proposal raises.



HAP/LAP master thesis (Noelia Migueles, 2016-06-27) 2017-Jun-27

Talk: Computational explorations of creative language (C. Strapparava, 2017-07-07) 2017-Jun-26

PhD Thesis: Automatic Scansion of Poetry (M. Agirrezabal, 2015/06/19) 2017-Jun-15

Neural Machine Translation. Open workshop with Kyunghyun Cho (2017-05-29) 2017-May-16

Read more IXA    

TALP Talk: Generative adversarial networks (GAN) applied to Speech Enhancement

Neural Machine Translation

Deep Dive in Deep Learning with TensorFlow

Mapping Unseen Words to Task-Trained Embedding Spaces

Read more UPC    

GRIAL: Grup de Recerca Interuniversitari en Aplicacions Lingüístiques

Read more UB    

Send mail to the webmaster to comment on these pages                                                      manage