4
$\begingroup$

I have trained a stance detection model using SVMs. Wanted to know how can I deploy this as a chrome extensions. I do understand that the question is a bit broad but any links, suggestions etc. will be appreciated:)

$\endgroup$
5
  • $\begingroup$ Well if it's about how to develop a chrome extension there are plenty of resources: google.com/search?q=develop+chrome+extension. Otherwise you should give more detail the desired setup: is the model trained before and deployed for testing, or does the deployment involves training, and in this case does this happen on the client side or server side...? $\endgroup$ – Erwan Sep 26 '19 at 12:53
  • $\begingroup$ Hi! No, it's not about developing a chrome extension but more of how to deploy a machine learning model as a chrome extension. The model is already trained. I am not sure whether will run on the client or server-side. $\endgroup$ – Shawn Sep 26 '19 at 17:19
  • $\begingroup$ Also, I am a high school student working on this project. I am not sure this is the right place but I was looking for someone knowledgeable to guide me on this. Deploying machine learning projects as a chrome extension is kind of a fringe area and not too many people work on it. If possible, please do share a medium through which I can personally conact you $\endgroup$ – Shawn Sep 26 '19 at 17:22
  • $\begingroup$ You have a better chance to get useful advice here on DataScienceSE (or stackoverflow.com about the programming side) than asking me personally: I know a fair bit about ML, but almost nothing about chrome extensions or server-client communication. When you ask a question you should try to make it as specific as possible, because when it's vague people are not sure what to tell you so nobody answers. $\endgroup$ – Erwan Sep 26 '19 at 19:05
  • $\begingroup$ Oh! Will keep in mind while asking more question $\endgroup$ – Shawn Sep 27 '19 at 11:18
1
$\begingroup$

Since what you want to do is to apply a pre-trained model and use the predictions on the client side, you have two options:

  1. Client-side application of the model: the extension would have to download the model file and store it, then it would need to compute the predictions from this model. This would probably require the SVM library used for the training to be available on the client side, this might be an issue.
  2. Server-side application of the model: the extension would have to upload the data to the server, then the server computes the predictions using the model, and sends the predictions back to the client.

Usually option 1 is preferable since it doesn't require transfering the input data to the server and then sending back the predictions. It's also safer with respect to the user privacy, their data is not uploaded anywhere. However it requires the client to be able to compute the predictions, so probably requiring the SVM library to be installed and the extension must be able to communicate with it, so it might not be feasible.

$\endgroup$
3
  • $\begingroup$ Thanks! Any links as to where I can read more on it? Also, will it help if I write the code in Tesnorflow.js? $\endgroup$ – Shawn Sep 27 '19 at 11:17
  • $\begingroup$ Sorry I don't really know about any resources... or about anything in this topic to be honest! I also don't know Tensorflow, but as long as you can (1) train the model with it and (2) have it available client-side you should be fine. $\endgroup$ – Erwan Sep 27 '19 at 13:38
  • $\begingroup$ Oh! Thanks @Erwan $\endgroup$ – Shawn Sep 27 '19 at 14:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.