Deep learning model for automated assessment of lexical stress of non-native english speakers

Abstract

In this paper, we present a novel system to practice lexical stress in L2 English learning with Amazon Alexa home assistant. The language learning for non-native English speakers mostly focuses on practicing correct grammar, extending language vocabulary, and improving pronunciation. The system proposed enables a person to practice lexical stress skills at home by having conversations with Alexa assistant. The system assesses student's abilities to enunciate words with a correct lexical stress and automatically selects the next words to practice. After a series of exercises, the system informs the student on the improvement. The main scientific contribution of the work presented is a deep learning model for automated assessment of lexical stress of non-native English speakers. The model is based on a transfer learning technique. First, we train the model to predict the location of a lexical stress on a syllable level using a large corpus of native English speech. Then, we tune the model with a limited amount of a non-native speech. A corpus of non-native English speech obtained from Polish speakers is incorporated into the training and testing of the model. It is shown that the system enables to create a vocabulary for a particular speaker interactively.In this paper, we present a novel system to practice lexical stress in L2 English learning with Amazon Alexa home assistant. The language learning for non-native English speakers mostly focuses on practicing correct grammar, extending language vocabulary, and improving pronunciation. The system proposed enables a person to practice lexical stress skills at home by having conversations with Alexa assistant. The system assesses student's abilities to enunciate words with a correct lexical stress and automatically selects the next words to practice. After a series of exercises, the system informs the student on the improvement. The main scientific contribution of the work presented is a deep learning model for automated assessment of lexical stress of non-native English speakers. The model is based on a transfer learning technique. First, we train the model to predict the location of a lexical stress on a syllable level using a large corpus of native English speech. Then, we tune the model with a limited amo...