Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Details on Hardware requirement and duration for training #4

Open
sathiyabalu89 opened this issue Dec 3, 2020 · 0 comments
Open

Details on Hardware requirement and duration for training #4

sathiyabalu89 opened this issue Dec 3, 2020 · 0 comments

Comments

@sathiyabalu89
Copy link

sathiyabalu89 commented Dec 3, 2020

First of all thank you for making the embedding available. I have few queries with respect to hardware requirements and data preparation.

  1. What is the hardware configuration? Would like to know the following details     
    a. GPU used and number of GPUs used     
    b. GPU RAM     
    c. Number of CPU cores and CPU RAM   

  2. What is the duration taken to train the chempatent embedding?

  3. How do I prepare the training data if I have many multi word token in domain like chemistry. For example:

       1. Original Sentences: "This is a multi word chemical component 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide. \n This is another sentence."
       
       Here "3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide" is a single token. There are multiple words inside the token which are white space separated. This would lead to the above token to be split as 3 tokens: ['3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl','tetrazolium', 'bromide'].
       
       How can I avoid this? Can I give the input training data in the following format to avoid this?
       
       Training data(1) : List of tokens for each sentences. So the training text file will have list of list tokens.
       
       [['This', 'is', 'a', 'multi', 'word', 'chemical', 'component', '3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide'], ['This', 'is', 'another', 'sentence.']]
       
       Training data(2): Here I have concatenated the multi keyword token by '|' symbol.
       "This is a multi word chemical component 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl|tetrazolium|bromide. \n This is another sentence." Then I tweak the ELMO code to handle the | symbol and retain them as a single token.
       
       Please guide on the best way to prepare the training data.
    
  4. In your paper "Improving Chemical Named Entity Recognition in Patents
    with Contextualized Word Embeddings" OpenNLP is used for sentence detection and Oscar4 for word tokenization right?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant