Centar za jezična istraživanja UNIRI Vas sa zadovoljstvom poziva na predavanje “Opening up dictionaries for natural language understanding” koje će održati Simon Krek, istraživač na Institutu “Jožef Stefan” u Laboratoriju za umjetnu inteligenciju i voditelj Centra za jezične resurse i tehnologije na Sveučilištu u Ljubljani.
Simon Krek je renomirani stručnjak u područjima leksikografije i leksikogramatike, korpusnog jezikoslovlja, obrade prirodnog jezika, infrastrukture jezičnih tehnologija i računalno potpomognutog učenja i poučavanja jezika.
U svom predavanju će se osvrnuti na važnost otvaranja leksikografskih resursa za postizanje cilja razumijevanja prirodnog jezika u kontekstu razvoja i primjene umjetne inteligencije u lingvistici.
Predavanje će se održati u četvrtak, 6. travnja 2023., od 12:00 do 13:00 sati, u predavaonici F101, Filozofski fakultet, Sveučilišna avenija 4, Rijeka u organizaciji Centra za jezična istraživanja.
Pridružite nam se i saznajte više o tome kako moderna leksikografija može pomoći u postizanju cilja razumijevanja prirodnog jezika u umjetnoj inteligenciji.
Naslov: Opening up dictionaries for natural language understanding
Sažetak: Lexicographers have been quite cautious about claiming that representations of meaning in their dictionaries are absolute. Unfortunately, one usually finds such warnings only in prefaces, dictionary front matter, commentaries and other more obscure works, and they are in most cases ignored. However, it is important to take these warnings seriously, especially when lexicographic resources are used as absolute truths for semantic tasks in Natural Language Processing.
Natural language understanding (NLU) is one of the most important goals in Artificial Intelligence. In the race to achieve the NLU goal, the fact that knowledge is essentially defined by humans sometimes tends to be forgotten. In this sense, understanding what constitutes shared meaning in a language community, or between different language communities, is still relevant. As a consequence, modern lexicography needs to adapt and provide adequate data (also) for NLU tasks. I will try to provide some answers to the following questions: are we actually measuring anything meaningful in NLU benchmarks (e.g. WiC in SuperGLUE)? Can modern lexicography help with this?
Lexicographers have been quite cautious about claiming that representations of meaning in their dictionaries are absolute. Unfortunately, one usually finds such warnings only in prefaces, dictionary front matter, commentaries and other more obscure works, and they are in most cases ignored. However, it is important to take these warnings seriously, especially when lexicographic resources are used as absolute truths for semantic tasks in Natural Language Processing.
Natural language understanding (NLU) is one of the most important goals in Artificial Intelligence. In the race to achieve the NLU goal, the fact that knowledge is essentially defined by humans sometimes tends to be forgotten. In this sense, understanding what constitutes shared meaning in a language community, or between different language communities, is still relevant. As a consequence, modern lexicography needs to adapt and provide adequate data (also) for NLU tasks. I will try to provide some answers to the following questions: are we actually measuring anything meaningful in NLU benchmarks (e.g. WiC in SuperGLUE)? Can modern lexicography help with this?