Question about ELMo embedding caching design

CogComp / zoe

Zero-Shot Open Entity Typing as Type-Compatible Grounding, EMNLP'18.

43 stars 5 forks source link

Question about ELMo embedding caching design #23

Closed hitercs closed 5 years ago

hitercs commented 5 years ago

Hi,

I have a question about your ELMo embedding caching design. When querying caching ELMo vectors of standard dataset like FIGER, you query the vector using the mention string only, regardless of the context. https://github.com/CogComp/zoe/blob/cf0924b18d144497c440573b7313cfbb4c564fe2/zoe_utils.py#L207 Do you based on the assumption that standard dataset you used don't have two mentions with exactly the same surface form?

Slash0BZ commented 5 years ago

Yes, that is a flawed assumption in the experiment, thanks for pointing it out. However, the concatenated mention surfaces rarely repeat in the datasets, as they are named entities and nominals. We re-did the experiments, and the numbers were affected by around 1%.

hitercs commented 5 years ago

OK. Got it. Thanks.

Slash0BZ commented 5 years ago

No problem. Please let me know if you have other questions.