Track: VoiceTech |
| Text Summarization using Transformer-based Language Models |
| In this project, we focus on developing a model, based on pre-trained open-source transformer-based language models (LM), to process a customer’s phone call or voicemail transcript and generate a summary. While LMs such as GPT-3 can generate a reasonably accurate summary of the transcript of a phone call or voicemail, integrating GPT-3 into a commercial service raises cost and privacy concerns. So, this project aims to experiment with a few open source LMs for text summarization, and finetune a subset of them with our own dataset, then evaluate the performance using commonly adopted metrics such as ROGUE score, so that the best performing model for our dataset can be obtained for potential use in analyzing call transcripts. |
|
| Presentation Video |
| Presentation Notes |
| Ahmed_RTC_Text-Summarization-using-Transformer-based-Language-Models.pdf |
