Brooklyn

Inspiration

A couple of weeks ago, the eight of us were neck-deep in a Valorant gaming session, we were minutes away from victory when all of a sudden, our Wi-Fi dies. Having spent the last many hours invested in securing this win one can imagine our frustration and anger. We pull out our mobiles and start furiously chatting with the wi-fi company. We went in full guns blazing trying to get our wi-fi back up, little realizing that we were interacting with a dumb chatbot that was giving us stock answers totally unrelated to our complaints. We realized at this point, that this was not a one-time thing and that we had been dealing with incompetent chatbots every time we had to communicate with any company be it our wi-fi providers or a delivery service company. At this moment we decided to channel our frustration into something productive and thus our NLP-based AI-driven intelligent customer experience platform Brooklyn was born. This tool is built on the notion that it would not just solve everyday customer issues but recognize sentiments, and emotions and execute human handoffs was born.

What it does

Our platform Brooklyn works on the following five models

Context Model - It analyses the context of the message to understand which department executive to assign the complainee to. For example, if the customer has a billing issue, the customer is assigned to the billing department executive.
Emotion Analysis Model - It analyses the emotion within a text and upon hovering over the text, it displays the underlying emotion with the respective emoticon. Our model is also capable of detecting sarcasm and indicates the same with a specific color code next to the text. For example, if the customer is unsatisfied with the executive's performance and sends the following message “You have not been helpful at all!”, upon hovering over the text an angry emoticon pops up.
Hate Speech Detection - It identifies hate speech and prevents people from sending rude and inappropriate messages, thereby ensuring a safe and professional interaction.
Personal Information Extractor - It protects the customers by hiding their sensitive information such as credit card numbers, mobile numbers, card pin numbers, etc. For example, if a customer sends their credit card information to the customer support executive, the information will be hashed entirely. Performance Model - The rating of the customer support executive keeps changing as per their performance.
Performance Model - The rating of the customer support executive is updated automatically as per their performance

How we built it

The frontend of our application is built using the lightweight React frameworks. The Central Server & APIs are built using Python Flask. Brooklyn harnesses the power of ExpertAI’s APIs & other “State of the Art” NLP models to implement the Context, Emotion Analysis, Hate Speech Detection & Personal Identifiable Information Extractor models. Brooklyn application & primary storage database are hosted on the Azure cloud.

Challenges we ran into

The primary challenge was that we wanted to develop, a "Futuristic Platform" where ExpertAI's models were synergized with our own NLP models
Developing a fraud free, multi-parameter based dynamic executive performance rating algorithm

Accomplishments that we're proud of

Accuracy of Brookyln's NLP models
Aesthetic & customer friendly User Interface
Design to Deployment in 12 days

What we learned

We identified a real-life problem which we addressed it by applying AI NLP technologies
We got to know that the current customer support platform lacks executive assistive tools like customer Emotion Capture, Personal Identifiable Information Censor Model, Hate Speech Blocker, etc.

What's next for Brooklyn

Integration of more ML/AI enabled tools to better assist and service executive and perform analytics
Brookyln can be deployed & scaled easily to cover all products and services of an enterprise
Brookyln can also be extended to support a wide range of smart devices which includes smartphones, tablets, etc.