With the advancements in technology, we can work towards creating products that will benefit its users. Read further about Amol Birajdar’s journey with Great Learning’s PGP Artificial Intelligence and Machine Learning Course and how he implemented his learnings at work.
I’m Amol Birajdar, and I am working as a System Tester in a leading international automation company. We are using INTEGRATED Portal extensively and found that the performance of the INTEGRATED Portal Online Help System is not satisfactory, and there is a vast scope of improvement.
Our aim was to build a Chatbot that could solve the problems that we could identify. We found that the existing product “INTEGRATED Portal Online Help System” is not that intelligent enough to answer User queries. It is designed in such a way that whenever a user asks a question, it treats each word in the question as a Keyword and returns all the pages wherever those keywords are present.
E.g., End-User: Add Instruction
INTEGRATED Portal Online Help System: Let me search and return you all the pages wherever the keywords “Add” or/and “Instruction” are present.
So, at present, it returns 1000 pages to the user, which is totally irrelevant. It takes around 90 seconds to return this bundle of data, and again it seems to be very much impossible for the user to find the answer which he is looking for in those 1000 pages.
Thus, the identified key problems are,
- It doesn’t provide a relevant answer to most of the questions asked by the user.
- There is no Top Priority set that results in a bundle of irrelevant data, and the actual result is buried down in irrelevant data.
- Due to the loading of so many irrelevant data, INTEGRATED Portal Online Help system performance is also impacted.
- Since it is keyword-driven, and a user has to be very specific on what he wants to find an answer for.
- And definitely, Finding an answer is way too difficult and time-consuming.
Then, we perceived, Our INTEGRATED Portal Online Help System is not at all intelligent to understand the User query intention. And this could be solved by using Artificial intelligence. An Artificial intelligence-powered chatbot can analyze the User’s question to identify the User Intent and extract relevant entities.
A branch of Artificial intelligence called Natural language Processing deals with finding user intent and entity from human languages. So we used NLP to solve this problem. Our aim is to extract intent and entity and then use it to search in our database. In this way, instead of returning 1000s of pages, only relevant pages will be returned, which will save the user time. We solved the identified problems, and the steps included are:
- Test Data preparation
- Data Reshaping
- Pre-Processing
- Development of AI Models
- Testing the models
- Creating UI
- Deployment
We proposed an AI Chatbot, which is a Conversational solution driven by Artificial Intelligence that connects the End Users with INTEGRATED Portal through Text and Voice recognition. Our Chatbot is a smart solution that enables the End users to get their query answered in a Millisecond with 24/7 hours of availability. It is developed using AI Algorithms and learned right from scratch. It is trained to develop its own consciousness on the text, understand the user intention behind the question and provide one relevant answer to the user. Also, by understanding the query intention, it shows up similar questions to the user. The solution is leveraged to replicate the way the User’s brain works to process data and make sense of it for better decision making.
Our proposed Chatbot comes with a FAQ system that collects User feedbacks, creates a repository of responses, and goes on evolving with the responses so that customer experience can be improved.
With our solution, the User will not have to be concerned about any keyword before asking a question, doesn’t need to have any waiting time, and the best part would be, the user doesn’t have to make an effort to search the answer in a bundle of irrelevant data. Our Chatbot aims at helping End Users to get quick & most accurate replies to their questions, 24/7 hour assistance which might result in higher productivity.
E.g., End-User: Add Instruction/What is add instruction/Use of Add instruction/How Add instruction is helpful?
AI Chatbot: “Add” instruction is used to add the value at input IN1 and the value at input IN2 and query the sum at output OUT where OUT:= IN1+IN2
Apart from being able to hold meaningful conversations, instant & 24/7 responses, FAQ system, Our solution is to assist our End Users with their requested language, not just English. With advancements in Natural Language Processing and Neural Machine Translation, Our Chatbot can respond to User’s language.
E.g.,
According to a conducted survey, on average Current INTEGRATED portal search is only 27% accurate with day-to-day questions. When the same questions were passed to our Chatbot, the accuracy jumped to 74%. So, at an early stage itself, our Chatbot is three times better than the current search mechanism.
Nowadays, every major company has its own assistant like Alexa (Amazon), Siri (Apple), Cortona (Microsoft), Watson (IBM), etc. Our Chatbot will be the company’s first virtual assistant used by end customers. By implementing this, we have opened a whole new domain/market for our company.
