DataHour: Self-Supervised Address Correction for Q-Commerce

Accurate customer locations for quick-commerce is a necessary requirement for under 30-minute online food and grocery delivery systems. Inaccurate locations can negatively impact delivery time promises, ranking of relevant restaurants and the accessibility of restaurants to customers. GPS locations captured using the customer smart-phone application can be inaccurate due to scattering of electromagnetic signals by vegetation or buildings or metallic reflectors as well as due to manual error in placing the GPS pin on the application.

In this talk, we discuss the design of a location correction system for q-commerce as a cascade of a location inaccuracy classifier and a geocoder. Apart from the address text and customer location inputs, we leverage signals collected from the delivery partner’s application. Using text and numeric inputs derived from these signals, we design a self-supervised multimodal architecture for the location inaccuracy classifier. We also touch upon the shortcomings of known geocoders and ways to make them production grade for q-commerce.