LUMII

LUMII > Research directions > Machine learning and Computational Linguistics

Machine learning and Computational Linguistics

Research in computational linguistics and machine learning is conducted at the Artificial Intelligence Laboratory in the following directions:

speech recognition and text-to-speech synthesis for Latvian,
syntactic and semantic parsing for multilingual information extraction and summarisation,
multilingual natural language generation,
controlled natural language for knowledge representation (multilingual authoring, verbalisation and querying of ontologies),machine translation (statistical, neural, rule-based)m,
development of annotated Latvian language resources for natural language processing (NLP): lexicons, speech corpora, text corpora (incl. treebanks and sembanks),
computational lexicography and corpus linguistics,
deep neural networks for classical algorithm learning,
creation of Latvian language support resources.

Research in computational linguistics and machine learning is conducted at the Artificial Intelligence Laboratory in the following directions:

speech recognition and text-to-speech synthesis for Latvian,
syntactic and semantic parsing for multilingual information extraction and summarisation,
multilingual natural language generation,
controlled natural language for knowledge representation (multilingual authoring, verbalisation and querying of ontologies),machine translation (statistical, neural, rule-based)m,
development of annotated Latvian language resources for natural language processing (NLP): lexicons, speech corpora, text corpora (incl. treebanks and sembanks),
computational lexicography and corpus linguistics,
deep neural networks for classical algorithm learning.

We have recently implemented multiple research and market-oriented projects on large-scale text analytics and speech recognition in cooperation with National Information Agency LETA, in order to handle Big Data problems arising in the media monitoring business. This collaboration has allowed LETA to join a large EU Horizon 2020 project on Scalable Understanding of Multilingual MediA (SUMMA) together with University of Edinburgh, UCL, University of Sheffield, BBC, Deutsche Welle and others.

In parallel, we have developed a text processing system for National Library of Latvia, dealing with part-of-speech tagging and named-entity recognition in a corpus of billions of words from OCR-scanned texts. We also collaborate with Riga Stradins University, developing a platform for monitoring user-generated content in news portal communities, and for analysing transcripts of parliament debates.

Our researchers successfully participate in international competitions, achieving the top result in the task on Abstract Meaning Representation (AMR) parsing at the International Workshop on Semantic Evaluation (SemEval-2016).

Leading researchers: Normunds Grūzītis, Guntis Bārzdiņš, Inguna Skadiņa, Ilze Auziņa, Pēteris Paikens, Kārlis Freivalds, Artūrs Sproģis, Andrejs Spektors.