Search code examples
tensorflowneural-networkibm-watsonwatson-conversation

How can IBM Watson Assistant access saved model on Tensorflow?


I have a project that I am working on that involves training a deep neural network to generate a relevance score for users and project (building a recommender system to recommend projects for users).

I was wondering if there is any way for me to save an already trained recommender system to Tensorflow and then have the Watson Assistant query that model?

I am really just interested in setting up an interactive experience for a user to find projects that they might be interested in.

I am not experienced in either Tensorflow/Watson Assistant so I apologize in advance if this is a foolish question.


Solution

  • As I am aware of, there’s no direct way to achieve this. You may have to use IBM Watson Studio to expose your deep learning model as an API and create an application in the middle that interacts with both Watson Assistant API and Watson Studio API serving the responses to the UI for your user.

    References:

    Check here on how to deploy a tensorflow batch API - https://dat https://dataplatform.cloud.ibm.com/docs/content/analyze-data/ml-custom_libs_tensorflow.html

    https://dataplatform.cloud.ibm.com/docs/content/analyze-data/ml_dlaas_tensorflow_deploy_score.html

    Watson Assistant API reference - https://console.bluemix.net/apidocs/assistant