The lecture "Learning Terminological Concept Systems from Multilingual Texts" addresses the first successes of the Text2TCS project in extracting terminological concept systems.
Abstract:
Terminological inconsistency represents one major source of misunderstanding in specialized communication. One vital measure to counteract such inconsistency is the creation of a terminological concept system (TCS) that represents concepts, their terms and interrelations. A multilingual TCS can ensure that different parties in a communication, such as medical, political, and news teams in times of crisis, consistently refer to phenomena by utilizing the same words. For instance, "COVID-19 isSpread airborne" represents a highly informative relation, especially when equipped with terms in several languages. Several approaches to extract terms from text have been proposed, however, few also consider representing interrelations between concepts and terms. In this talk, I will present ongoing research within the project Extracting Terminological Concept Systems from Natural Language Text (Text2TCS) to improve multilingual term and relation extraction in domain-specific contexts. To this end, we currently rely on pre-trained language models, in particular XLM-R, as well as innovative uses of Neural Machine Translation (NMT) models, which I will present alongside additional experiments we have conducted. Resulting solutions of the project will be integrated into the European Language Grid (ELG) until summer this year.
Join Zoom Meeting
fau.zoom.us/j/92343482845
Meeting ID: 923 4348 2845
Passcode: 5556969
One tap mobile
+496971049922,,92343482845#,,,,*5556969# Germany
+493056795800,,92343482845#,,,,*5556969# Germany
Dial by your location
+49 69 7104 9922 Germany
+49 30 5679 5800 Germany
+49 69 3807 9883 Germany
+49 695 050 2596 Germany
Meeting ID: 923 4348 2845
Passcode: 5556969
Find your local number: fau.zoom.us/u/cbWb6R0j1m
Join by SIP
92343482845@fr.zmeu.us
Join by H.323
213.19.144.110 (Amsterdam Netherlands)
213.244.140.110 (Germany)
Meeting ID: 923 4348 2845
Passcode: 5556969
