PLATFORM FOR CHATBOT AUTHORING

Chatbots are the biggest talk of the town. But is your chatbot intelligent enough?

THE CHALLENGE

Our client, a global technology company, identified an opportunity to disrupt chatbot technology by building an intelligent chatbot authoring platform. The current platforms for chatbot creation require significant crafting from chabot authors and have limited domain definition capabilities. The client wanted an engine that could create AI enabled chatbots, and train them using machine learning tools to enhance their ability to develop a deeper understanding of the customer and thereby converse more intelligently.

OUR APPROACH

Existing chatbot platforms have limited abilities to deal with context — they are focused on question answering rather than dealing with a sequence of utterances. Our approach was to create a platform for chabot creation that uses domain specific advance language models that allows the system to define entities and intents of the customer talking to the chabot. This enables deeper understanding of the wants of the customer and helps the chatbot give an appropriate response.

THE METHODLOGY

  • The platform uses existing client data and uploaded human to human chat logs and domain definitions, for the machine learning algorithm to ingest and create the first version of the chatbot.
  • The advance language model learns various domain specific entities (categorizations such as person, brand etc.) and intents (for example, information search vs. purchase vs. service request). This enables the chatbot to recognize what the customer wants, and respond accordingly.
  • The machine learning retrieval model allows the author to use the chatbot in a simulated environment to create feedback for the chatbot and improve its responses.
  • Active learning is used to propose chats that need tagging to improve the existing models used by the chatbot.
  • Deep learning algorithms are used to encode chats and improve the quality of response, allowing iterative refinement of the bot through simulated interaction with the author.

TECHNIQUES, TECHNOLOGIES, TOOLS

  • Conditional Random Fields, Sentiment Analysis, Topic Modeling, Recurrent Neural Networks (RNN), LSTM

THE RESULT/IMPACT

We created the chatbot authoring platform in 6 months. The AI techniques used not only led to reduced effort in manual definition of the chatbot, but also an iterative approach to continuously improve the chatbot. As a result, the platform creates chatbots with a deeper ability to understand and converse with humans.

STILL NOT SURE WHAT TO DO?

We are glad that you preferred to contact us. Please fill our short form and one of our friendly team members will contact you back.

Form is temporarily not available. Please visit our contact page.
X
×

WRITE TO US