Decoding Problems, Coding Solutions
Nice way to start with real-world application of AI is Intent Classification and Named Entities Recognition using NLU (Natural Language Understanding). ChatBots are popular across industries. They make easy for organizations to provide customer services.
ChatBots needs to allow natural language communication in different languages. To understand what customer is asking for (means customer's intent) becomes extremely useful. For example, a bank customer types: "I want statement of previous month" on the banking chatbot. In this customer query intent is "statement" and entity is "previous month". The chatbot must send a PDF of previous month statement. But there is a small issue. From banking perspective, "statement" is an incomplete intent because statement can be of an account or the credit card.
Hence, the chatbot must display two options to the customer to select from
Once an option is selected, the PDF can be sent to the customer.
But what if customer types: "I want account statement of previous month". This time the intent should be "account statement" and entity will be "previous month". Now the chatbot should not display two options instead should directly send the PDF having an account statement. Moreover, entity like "previous month" should get resolved to actual month and year. The NLU engine must do this also.
To make such chatbots we need to have an NLU intent classifier and entity resolver that not only can be trained to understand english language but also other languages. We at Leadows Technologies, make such smart chatbots that interact using natural languages with customer. For intent classification and named entity recoginition and resolving them to proper value we use DialogFlow or Rasa.