Samsung electronics co., ltd. (20240321262). MULTILINGUAL DOMAIN DETECTION USING ONE LANGUAGE RESOURCE simplified abstract
Contents
MULTILINGUAL DOMAIN DETECTION USING ONE LANGUAGE RESOURCE
Organization Name
Inventor(s)
Tapas Kanungo of Redmond WA (US)
Stephen Walsh of Sunnyvale CA (US)
Preeti Saraswat of Santa Clara CA (US)
Yurii Lozhnevsky of Sunnyvale CA (US)
Nehal Bengre Juraska of Cupertino CA (US)
Qingxiaoyang Zhu of Woodland CA (US)
MULTILINGUAL DOMAIN DETECTION USING ONE LANGUAGE RESOURCE - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240321262 titled 'MULTILINGUAL DOMAIN DETECTION USING ONE LANGUAGE RESOURCE
The abstract describes a method for generating a multilingual training corpus and fine-tuning a multilingual language model using the corpus.
- The method involves creating a training corpus with labeled utterances in multiple languages, including translations of at least one utterance.
- It also includes the process of refining a multilingual language model using the generated corpus.
Potential Applications: This technology can be applied in the development of multilingual natural language processing systems, machine translation, and speech recognition software.
Problems Solved: This method addresses the challenge of training language models in multiple languages simultaneously, improving the accuracy and performance of multilingual systems.
Benefits: The use of a multilingual training corpus and language model can enhance the effectiveness of communication across different languages, leading to more accurate and efficient language processing applications.
Commercial Applications: This technology can be utilized by companies developing multilingual chatbots, virtual assistants, and customer service platforms to improve language understanding and interaction with users.
Questions about Multilingual Training Corpus and Language Model: 1. How does the method ensure the accuracy of translations in the multilingual training corpus? 2. What are the potential limitations of using a multilingual language model for natural language processing tasks?
Original Abstract Submitted
a method includes generating, using at least one processing device of an electronic device, a multilingual training corpus including labeled utterances in multiple languages including a first language and a second language. the multilingual training corpus includes at least one utterance that has been translated into the multiple languages. the method also includes fine-tuning, using the at least one processing device, a multilingual language model using the multilingual training corpus.
