1. Introduction #
According to Juniper Analytics, the number of conversational bots is expected to triple by 2023. These bots have gained high demand due to their ability to provide human-like communication experiences. They utilize advanced Natural Language Processing (NLP) and are trained on Frequently Asked Questions (FAQs) to deliver relevant information to user queries. Floatchat’s NLP Engine, e-sense, processes user queries, accounting for spelling and grammatical mistakes, to provide accurate responses based on a predefined match percentage.
Training a bot involves mapping correct answers to specific queries to ensure the desired output is provided. For example, if a ticket booking bot fails to provide a valid answer, the bot can be trained with a new response or linked to an existing relevant response for future inquiries.
3. Terms and Definitions #
Natural Language Processing is the ability of a computer program to understand human language as it is spoken.
These are data points or values extracted from user queries, allowing for customization and association of information.
|Frequently Asked Questions are standard queries relevant to a product or service.
4. Training Process #
The training process involves the following steps:
- A user inputs a query/question on the bot.
- The bot searches for the closest matching query saved in the FAQ module.
- The bot triggers a response mapped to the query.
- Bot administrators can monitor responses under the “Responses” section of the “Train” tab.
- In the case of an unsatisfactory or no response, users can click on the “Train” option to provide or improve the answer for the given query.
4.1 Steps to train a bot for an FAQ #
Step 1: Click on the “Train” tab in the left panel and select “Responses.”
Step 2: Click on the “Train” option from the right-hand side menu of the FAQ response that requires training.
Step 3: In the modal that opens, users can improve responses by adding a new response, adding more variations to the question, or adding entities for customization.
4.2 Adding an entity #
Users can add entities to extract specific data points or values from user queries.
4.3 Tagging an entity in a FAQ #
Entities can be tagged in FAQs to associate them with specific queries.
4.4 Query and its response #
The query asked by the user and the corresponding response can be viewed and managed.
4.5 To understand the context of the conversation while training #
Users can view the context of the conversation where a specific query was encountered by clicking on the “View Conversation” button.
4.6 Pre-training Mode #
Users can test the bot’s response to specific queries before and after training.
4.7 Training the bot #
To train the bot, users can select the closest matching response or add a new response if no valid response exists.
4.8 Similarly training for other unanswered responses. #
Users can test the bot’s responses again to ensure the trained responses are provided.