Over the last decade, India has witnessed an explosion in the ecommerce industry. There is increasing adoption of e-commerce in smaller towns and cities over and above the densely populated urban centers. In this paper, we discuss the practical challenges involved with developing high-precision geocoding engines for these geographical regions in India. These challenges motivate the next iteration of our geocoding framework. In particular, we focus on addressing three core areas of improvement: 1) leveraging customer delivery data for geocoding, 2) understanding and solving for the diversity and variations in addresses for these new regions, and 3) overcoming the limited coverage of our reference corpus. To this end, we present GeoCloud. Key contributions of GeoCloud are 1) a training algorithm for learning reference-representations from delivery coordinates and 2) a retrieval algorithm for geocoding new addresses. We perform extensive testing of GeoCloud across India to capture the regional, socio-economical and linguistic diversity of our country. Our evaluation data is sampled from 72 cities and 21 states from the delivery addresses of a large e-commerce platform in India. The results show a significant improvement in precision and recall over the state-of-the-art geocoding system for India, and demonstrate the effectiveness of our intuitive, robust and generic approach. While we have shown the effectiveness of the framework for Indian addresses, we believe the framework can be applied to other countries as well, particularly where addresses are unstructured. To the best of our knowledge, this is the first instance of geocoding by learning reference-representations from large-scale delivery data.

Outfit recommendation automatically pairs user-specified reference clothing with the most suitable complement from online shops. Wearing aesthetically is a criterion for matching such fashion items. Fashion style tells a lot about one's personality and emerges from how people assemble clothing outfit from seemingly disjoint items into a cohesive concept. Experts share fashion tips showcasing their compositions to public where each item has both an image and textual meta-data. Also, retrieving products from online shopping catalogs in response to such real-world image query is essential for outfit recommendation. Our earlier tutorial focused on style and compatibility in fashion recommendation mostly based on metric and deep learning approaches. Herein, we cover several other aspects of fashion recommendation using visual signals (e.g., cross-scenario retrieval, attribute classification) and combine text input (e.g., interpretable embedding) as well. Each section concludes walking through programs executed on Jupyter workstation using real-world data sets.

Recommending lifestyle articles is of immediate interest to the e-commerce industry and is beginning to attract research attention. Often followed strategies, such as recommending popular items are inadequate for this vertical because of two reasons. Firstly, users have their own personal preference over items, referred to as personal styles, which lead to the long-tail phenomenon. Secondly, each user displays multiple personas, each persona has a preference over items which could be dictated by a particular occasion, e.g. dressing for a party would be different from dressing to go to office. Recommendation in this vertical is crucially dependent on discovering styles for each of the multiple personas. There is no literature which addresses this problem.