Interlanguage Vector Space: An AI-driven Approach to Replace Electronic Dictionaries

Track: Multilingual AI | TA5 | MICRO TALK |
Wednesday, January 27, 2021, 1:30pm – 2:00pm
Held in: Auditorium
Presenter:
Rafał Jaworski - XTM International

The interlanguage vector space is a technology based on extensive machine learning using neural networks and algebraic transformations. Its purpose is to compute the probability of a target language word being the translation of a given source word. It works on the level of word embeddings trained separately for over 200 languages. These embeddings — mathematical vectors — are then transformed into a single vector space, which make it possible to compare them with algebraic functions. The interlanguage vector space is able to compare cross-language semantic relations between words using numbers, vectors and applied mathematics.

Takeaways: Attendees will learn how a latest advance in neural network-driven machine learning can augment and impact the translation process; how interlanguage vector space is used to calculate the probability of a target language word being the translation of a given source language word; how this is a significant impact in the world of natural language processing AI; and how it is and will impact daily translation activity.