Dialpad scientists to present their findings on improving punctuation for transcriptions
Dialpad Inc., the industry leader in AI-powered communication and collaboration, today announced its paper “Improving Punctuation Restoration for Speech Transcripts via External Data” will be presented at the 2021 Conference on Empirical Methods in Natural Language Processing (EMNLP). The conference takes place November 7-11, 2021 and Dialpad Applied Scientist, Xue-Yong Fu, will be showcasing the paper’s findings in front of 3,000 AI researchers and scientists.
The AI team at Dialpad is focused on developing and improving Dialpad’s proprietary Voice Intelligence (Vi™) engine which delivers real-time business optimization, from call coaching and automated note-taking to sentiment tracking and transcription analysis. Over the past several years, Dialpad has improved the accuracy of Vi and after analyzing almost two billion minutes of business voice data and benchmarking its ASR technology against competitors, Dialpad’s transcription model consistently comes out ahead when measuring accuracy. Beyond the unique challenges of transcribing real-life human-to-human business conversations, Dialpad’s AI Research team is tasked with extrapolating meaning and sentiment from conversations to improve transcript readability and identify key moments for customers.
“In addition to enhancing Dialpad’s Vi technology, our AI team is also focused on pushing the voice technology industry forward by externalizing key findings through blogs, published research papers and academic conferences,” said Simon Corston-Oliver, Senior Manager, Machine Learning, Dialpad. “Dialpad is committed to making voice technology more accessible, accurate and ethical. We believe that sharing our research and having the opportunity to learn from others working on similar problems through opportunities like EMNLP is critical to accelerating innovation across the fields of computational linguistics and natural language processing.”
At EMNLP 2021, Dialpad will be presenting its paper “Improving Punctuation Restoration for Speech Transcripts via External Data” during the Workshop on Noisy User-generated Text (W-NUT) on November 11 at 17:55 GMT. Written and researched by Xue-Yong Fu, Cheng Chen, Md Tahmid Rahman Laskar, Shashi Bhushan and Simon Corston-Oliver, the paper describes a new approach to make AI-generated transcripts more readable through proper punctuation.
“While today’s automatic speech recognition (ASR) systems are highly accurate when transcribing human speech, these systems do not typically take punctuation into account. Without proper punctuation, readers may find it difficult to understand longer transcriptions,” said Xue-Yong Fu, Applied Scientist, Dialpad. “So, we’ve been working to develop a new way to tackle the punctuation restoration problem and are excited to share our approach with the EMNLP community.”
- Download the paper “Improving Punctuation Restoration for Speech Transcripts via External Data”
- Learn why punctuation matters in AI-generated transcripts
- See here for open roles on Dialpad’s AI team
- Read more from Dialpad’s AI team