Publications

The NTNU System at the Interspeech 2020 Non-Native Children’s Speech ASR Challenge

Interspeech 2020

• Technical Report of the Interspeech 2020 Non-Native Children’s Speech ASR Challenge supported by the SIG-CHILD group of ISCA.
• Achieved Second place in the competition.
• We built our ASR system on top of CNN-TDNNF-based acoustic models and tried various data augmentation strategies.

Download here

An Effective End-to-End Modeling Approach for Mispronunciation Detection

Interspeech 2020

• Presented a novel use of hybrid CTC-Attention approach to the Mispronunciation Detection task.
• Performed input augmentation with text prompt information to make E2E model more tailored for the Mispronunciation Detection task.

Download here

An Effective Contextual Language Modeling Framework for Speech Summarization with Augmented Features

EUSIPCO 2020

• Adopt and contextualize the state-of-the-art BERT-based model for speech summarization.
• Explore the incorporation of confidence scores to see if such an attempt could help alleviate the negative effects caused by imperfect ASR.
• Augmented the sentence embeddings obtained from BERT with extra structural and linguistic features to boost the performance.

Download here

A Hierarchical Encoding Framework for Text Readability Prediction

ROCLING 2019

• Presented a hierarchical framework to encode both the semantic or syntactic structures for better readability classification.
• Combined extra use of traditional handcrafted feature as side information to further boost the model’s performance.
• Awarded Runner-up of Best Poster Presentation Award.