NLP-based Mental Health Chatbot for Employees on the Autism Spectrum
Expertise Delivered
Industry
Our client,
Recruit AI is a B2B SaaS platform that aims to reduce inefficiencies in traditional hiring practices. Through ethical AI and data science, Recruit AI maximizes objectivity and reduces unconscious bias in talent and recruiting. It ensures workplace happiness through the best matches between employee and employer.
Disclaimer: The name Recruit AI is a placeholder, as there is an NDA signed between both parties.
Challenge
Recruit AI won the Microsoft-organized hackathon, AI for Accessibility 2019, which inspired the founders to create an inclusive onboarding framework for employees with special needs.
They were looking to design a mental health chatbot for employees on the autistic spectrum. The chatbot would help such employees navigate their day-to-day workplace challenges before they could speak with a counselor. Recruit AI’s previous FAQ chatbot couldn’t solve the queries of autistic employees, so they were left with no choice but to see a counselor directly.
The chatbot’s job was to assist employees with personalized insights & real-time support to face workplace challenges head-on and overcome them with ease. The bot conversed with employees like a therapist to peel back the layers and find the exact source of the challenge(s) they were facing.
Categories of workplace challenges to be addressed via the bot included, but were not limited to employee onboarding, work arrangements, and office/health-related issues.
The most significant challenge of developing a contextual bot for Recruit AI was managing each user's context while interacting with the bot. Doing so required an advanced system of code that learns what each user wants at any given time based on their input.
Context management is essential as it helps make the interactions between a bot, and its user less robotic as well as scripted and more streamlined. It is easier for bots to interact with users by establishing base facts and specific ways of communicating contextual to the situation.
Solution
A contextual bot enables the creation of more natural, realistic, and human-like conversations. Keeping this in mind, the engineers at Maruti Techlabs addressed the challenge by building a text-based NLP chatbot from scratch with particular attention to contextual management.
Here is how we ensured the bot was equipped with skills to detect and manage context:
- Our NLP engineers trained more than 160 intents with the provided utterances to improve the bot's accuracy in detecting intent.
- We also built a custom logic tree for context management, making room for continuous improvement and better conversation handling at scale.
- To ensure that the bot never lost context across different conversations, the engineering team kept adding relevant intents within the syntax tree structure of the bot through a JSON file.
We used the Microsoft Bot framework - Language Understanding (LUIS) to define the chatbot conversation and deployed it on the Microsoft Azure platform, as the client was a Microsoft partner.
The chatbot was designed to act as a therapist, offering helpful advice and suggestions, leading to better resolution. E.g., the user says, “I cannot adjust to meeting new people.” The bot responds by saying, “I am sorry to hear that! You may find it difficult to adjust to meeting new people or feeling stressed. It is okay to feel this way! Here are some tips to help you adjust better.”
The bot had to be trained across different themes of questions. For example, questions around onboarding required a different contextual understanding than those around teamwork. To help us craft better responses, we sought help from the conversational experts at WotNot, Maruti Techlabs’ proprietary no-code chatbot platform.
We trained the chatbot for different conversation themes:
- Onboarding
- Work Arrangements
- Office and Health
- Organizing and Planning
- Team Work
- Communication at work
Bot Intents and Utterances
- Bot intent refer to the goal the customer has in mind when typing a question or sentence.
- Utterances refer to anything that a user says or types
Like what you read? Validate your idea/project with our AI Readiness Audit.
Process
The project was a milestone-based two-phase endeavor. It was divided into two stages - the PoC stage, which lasted two months, and the actual project stage, which continued for three months.
During the PoC phase, the Maruti Techlabs team trained Microsoft LUIS for five initial conversations to confirm that everything aligned with the client’s expectations. Once we got the go-ahead from the client, we proceeded to the project stage.
In the project stage, the team trained the bot for about 50-60 conversations. They trained 160 intents and 3000 training phrases for each conversation in LUIS. The team taught LUIS to examine user input and provide contextual output for particular user requests.
Communication & Collaboration
A dedicated Slack channel was used for daily communication with the client. We met bi-weekly over Microsoft Teams to address questions and the progress of the chatbot.
Additionally, we used Google Sheets to classify data from conversations, sort out customer intent statements, and automate responses.
Technology Stack
Our Development Process
We follow Agile, Lean, & DevOps best practices to create a superior prototype that brings your users’ ideas to fruition through collaboration & rapid execution. Our top priority is quick reaction time & accessibility.
We really want to be your extended team, so apart from the regular meetings, you can be sure that each of our team members is one phone call, email, or message away.
NLP implementation steps are often very specific to the tasks to be executed. Our fundamental process relies on various text and data analysis tools to simplify human-machine interactions and enable businesses to deliver next-generation digital experiences that are contextually relevant, highly interactive, and refreshingly human.
AI Readiness Audit
Finding the right AI partner is no easy task.
Building an AI product that delivers on value, is even more challenging.
We help you get started with a slightly different approach. Before we get into the trenches and kickstart development, we take a top-down approach with an AI Readiness Audit.
This involves really validating the idea, through qualitative and quantitative analysis of your datasets, identifying the best fit approach to model development, and putting together an implementation roadmap.
All this before writing a single line of code, and investing heavily into the idea.
Achieve more with less.
