[Adapt] [Seminar] BERTScore: Evaluating Text Generation With Bert

[朱煜烨]

Hi Adapters,

This Wednesday I'll introduce you with an ICLR 2020 paper "BERTScore: Evaluating Text Generation With Bert". This paper proposes an automatic evaluation metric for text generation, which is based on pretrained BERT contextual embeddings. In my speech, I will talk about the shortcomings of some traditional metrics. For some task generation tasks, it is crucial to find a resonable evaluation metric. Here are the links for the paper and github repository:

Paper Link: https://arxiv.org/abs/1904.09675

Github Link: https://github.com/Tiiiger/bert_score

Time: Wed 4:00pm

Venue: SEIEE 3-414

Best regards,

Yuye