Helped Leading Food Delivery Platform

To Eliminate Miscommunication and resources on traditional speech recognition process, leading to customer satisfaction

98.9%

Accuracy Level

30%

Saving

About the Customer

Founded in 2014, the client is one among India’s leading on-demand delivery platform that aims to elevate convenience for the urban consumer by connecting consumers to thousands of restaurants and stores in over 500+ cities. They use innovative technology to provide hassle-free, and reliable delivery experience through its fleet of independent delivery partners. The company generates terabytes of data each week from the app, which has metamorphosed from being an application to just get food, to an app that delights users every time an order is placed.

Problem

Artificial Intelligence (AI) today is being used across a plethora of platforms to make operations smooth and offer customers personalised service. Delivery giants have been expanding customer bases drastically with the help of AI; using machine learning (ML) to provide a curated list of restaurants on the customer’s landing page, based on their location and preferences; group images under food categories and more importantly, letting users search for food items using colloquial terms.

Solution

With the scale and complexity in which the app is growing, an AI intervention was required to make the platform more efficient and simpler to use. Hence, moving from a human-intelligence based model to a human- led, AI platform was a natural progression. With the application being used in multiple cities, solving unique needs of regional speech and text interactions and guided discovery through bots required the help of a third party who could identify intents in multiple languages (Intent Classification), assist in classifying and naming entities (Named Entity Recognition) into various categories, while processing the language, and carry out the speech diarization process

Results

The use of sophisticated AI-based natural language processing tools like Speech Diarization, and transcription made language identification simpler for the client, thus enabling machines to understand and process language and correcting natural language queries correctly. Using natural language processing, a single model can now perform multiple tasks such as labelling the language associated with each word, as well as perform language tasks such as label the intent for each transcribed sentence spoken by the customer simultaneously. These advanced technologies have enabled the client to eliminate miscommunications and resources on traditional speech recognition process, leading to customer satisfaction.