sakshatrao/reassurancechatbot

Submission for the India PropTech Hackathon

ReassuranceChatbot

A Chatbot designed to provide reassurance to homebuyers who have invested in under-construction projects

Submission for the India PropTech Hackathon 2021

---

Demo

Watch a quick demonstration of the project prototype at - https://youtu.be/a5vpYZm-r1U

---

Overview

The chatbot is capable of performing three primary tasks -

1. FAQ Answering - One of the most common tasks of chatbots is to be able to answer general questions.
2. Personal Question Answering - This chatbot is also capable of answering personal questions specific to the user, helping in providing a personalized experience.
3. Construction Progress Updates - The chatbot also provides a dashboard for homebuyers to view construction-related details like completion, forecasts & estimated date of completion.

Motivation

• Homebuyers, especially middle-class, naturally would feel concerned after investing in under-construction projects because of the various risks like delays & frauds
• Lack of transparency in construction means construction details are not made readily available
• Most of the customer services are directed in the pre-sale phase (i.e. identifying potential customers and convincing them to buy a house) and not in the post-sale phase (i.e. reassuring homebuyers and allaying their fears and worries)

Features

FAQ Answering

• The chatbot is fed with some general questions & corresponding answers about the construction company, project & itself
• Using BERT sentence embeddings, the chatbot tries to find similarities between the user's question and the fed-in questions
• The most-similar question is selected and its answer is what is replied by the chatbot

Personal Question Answering

• Based on the user's question and using Natural Language Processing, the general topic of the question is predicted (i.e. whether the question relates to apartments, finances, etc.)
• The user's question is processed to include the user's name in the question
• Using the topic of the question and the processed question itself, an NL-to-SQL model converts the question into an SQL query
• This SQL query is then used to fetch the answer from the database

Construction Progress Updates

• To monitor construction progress, builders simply have to maintain a checklist
• Using the checklist, the chatbot estimates the completion percentage of the project by calculating checklist completion (which simply checks how many checklist items have been completed) & weighted completion (which assigns more weights to checklist items which require more efforts to complete).
• For example, if 4 out of 8 items have been checked, checklist completion if 50% but if the 4 items completed involve building 20-floor buildings while the remaining 4 items involve building parking lots, obviously the completion should be more than 50%
• Finally, the forecasts are made using two methods - by using completion datapoints in the ongoing construction & by using datapoints from previous constructions
• Using both forecasts, an estimated date of completion is determined

---

Thank You!