NVIDIA / sentiment-discovery

Unsupervised Language Modeling at scale for robust sentiment classification
Other
1.06k stars 202 forks source link

Minimal sentiment classification code #51

Closed harsham05 closed 5 years ago

harsham05 commented 5 years ago

I'm trying to expose the pre-trained models over REST, such that an HTTP POST request with a list of strings returns a response with their corresponding sentiment scores. The command line args can be replaced with their defaults.

I noticed that configure_data.py is being used to split the input data into train, test & validation batches & is also calling methods from data_utils module to prepare the data for training

https://github.com/NVIDIA/sentiment-discovery/blob/7f5ab28918a6fc29318a30f557b9454f0f5cc26a/classifier.py#L49

But, is there a simpler way to pass in a list of strings to the classify method?

https://github.com/NVIDIA/sentiment-discovery/blob/7f5ab28918a6fc29318a30f557b9454f0f5cc26a/classifier.py#L72

tanmaypandey7 commented 4 years ago

Hi Harsham, were you able to find a way to do it?