Resource2Vec: Linked Data distributed representations for term discovery in automatic speech recognition
作者：
Highlights：
• We present Resource2Vec, an embedding that compiles Linked Data resources.
• Resource2Vec proves very suitable for Term Discovery in Automatic Speech Recognition.
• Our Term Discovery strategy discovers Out-of-vocabulary terms in Linked Data corpora.
• A quantitative analysis of Resource2Vec helps to tune our Term Discovery strategy.
• With Resource2Vec, our strategy leads to a significant reduction of word error rates.
摘要
•We present Resource2Vec, an embedding that compiles Linked Data resources.•Resource2Vec proves very suitable for Term Discovery in Automatic Speech Recognition.•Our Term Discovery strategy discovers Out-of-vocabulary terms in Linked Data corpora.•A quantitative analysis of Resource2Vec helps to tune our Term Discovery strategy.•With Resource2Vec, our strategy leads to a significant reduction of word error rates.
论文关键词：Resource2Vec embedding,Linked Data,Term discovery,Out-of-vocabulary words,Automatic speech recognition
论文评审过程：Received 20 February 2018, Revised 25 May 2018, Accepted 16 June 2018, Available online 18 June 2018, Version of Record 30 June 2018.
论文官网地址：https://doi.org/10.1016/j.eswa.2018.06.039
