Skip to content

Latest commit

 

History

History
50 lines (31 loc) · 2.84 KB

README.md

File metadata and controls

50 lines (31 loc) · 2.84 KB
Raw

Conversational AI for Product related FAQs using Transformers with Azure Services

Landing Page

Bot

Problem Statement:

  • All e-commerce websites provides FAQs about specific products under the product description.
  • Those 8-10 FAQ's sometimes doesn't answer all queries about the products or give very less information related to that product due to which buyer/user doesn't get satisfied.
  • Else they provide us with millions of QNA and the website then tends to lose the users

Innovation:

  • Answers unknown questions related to that product
  • Answers right away instead of searching from a pile of thousands of questions
  • Web scraps the entire product page and answers the questions relatively
  • Even the rarest of the rarest questions are answered using AI and machine learning
  • It is pretrained on products specifications
  • Our solution uses an advanced Artificial Intelligence model GPT-2 to answer the questions and generate the output text

Novelity:

  • Not many e-commerce website uses conversational AI as an alternative to FAQs
  • Unlike other hardcoded chatbots our chatbot is trained using machine learning aspects
  • We eliminate the time period of a user asking the question and the seller answering it within 2-3 days
  • The AI chatbot seeks to answer rarest of the rarest question
  • Our solution for most of the time answers the question related to the question itself so that we don't lose customer retention due to unsatisfying results

Scalability:

  • It will be an extension and not an end product so that many e-commerce websites can use this similar approach to solve this problem
  • In the future, the product will use Image Recognition so that after identifying the features it can search the feature of the product web scrap and answer it accordingly
  • Expanding use cases with new features and capabilities
  • Creating engaging experiences beyond simple QNAs

Azure Services:

  • We used Microsoft Machine Learning Studio to script and train the machine learning model using GPT-3
  • Along with that we used Microsoft Azures Datastore to access the data stored in the containers
  • Used Microsoft Azure Devops for integration of model and web application and used API calls using REST API with the help of Microsoft Azure
  • Used Microsoft Cloud Server to deploy the product online

Architecture