Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
Use BERT to compute sentence similarity
See original GitHub issueI want to compute similarity between two sentences (sentA and sentB).
I have encoded each sentence using script i.e. load_and_extract.py. so now embedding matrix of sentA and sentB has shape (1,512,768). After that i am thinking to add fully connected layer to compute the similarity between two sentences.
Note: I am using base model (with 12 hidden layers)
Question: Is this right approach to use BERT for sentence similarity? Furthermore, I have also seen some people are using MaskedGlobalMaxPool1D after hidden layers to encode the sentences. Do I have to take embeddings after applying MaskedGlobalMaxPool1D? Why there is need of MaskedGlobalMaxPool1D ?
Thanks in advance.
Issue Analytics
- State:
- Created 5 years ago
- Comments:6 (1 by maintainers)
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
See #7 and #19.
I know bert-as-a-service and I’ve asked how to encode works: https://github.com/hanxiao/bert-as-service/issues/384
I would like to know how to extract the ‘sentence embeddings’ myself. I’ll check the ELMo like embeddings as well. Thanks.