Products on e-commerce websites are usually organized based on seller-provided product attributes. Customers looking for a product typically have certain needs or product use-cases in mind, for e.g., a headphone for gym classes, or a printer for a small business. However, they often struggle to map these use-cases to product attributes and subsequently fail to find the product they need. In this talk, we present a use-case based shopping (UBS) ML system that facilitates use-case based customer experiences (CXs). The UBS system recommends dominant product use-cases to customers along with most relevant products for those use-cases. Use-cases and their definitions vary across product categories and marketplaces (MPs). This makes training supervised models for thousands of e-commerce categories and multiple MPs infeasible by collecting large amount training data needed to train these models. In this talk, we present our work on scaling the UBS model by instruction tuning an LLM for our task.

Rewards-based programs are popular within e-commerce online stores, with the goal of providing serendipitous incentives to delight customers. These rewards (or incentives) could be in the form of cashback, free-shipping or discount coupons on purchases within specific categories. The success of such programs relies on their ability to identify relevant rewards for customers, from a wide variety of incentives available on the online store. Estimating the likelihood of a customer redeeming an incentive is challenging due to 1) data sparsity: relatively rare occurrence of coupon redemptions as compared to issuances, and 2) delayed feedback: customers taking time to redeem, resulting in inaccurate model refresh, compounded by data drift due to new customers and coupons.

Learning individual level treatment effects from observational data is a problem of growing interest. For instance, inferring the effect of delivery promises on purchase of products on an e-commerce site or selecting the most effective treatment for a specific patient. Although the scenarios where we want to estimate the treatment effects in presence of multiple treatments is quite common in real life, most existing works related to individual treatment effect (ITE) are focused primarily on binary treatments and do not have a natural extension to the multi-treatment scenarios. In this paper we present MEMENTO ? a methodology and a framework to estimate individual treatment effect for multi-treatment scenarios, where the treatments are discrete and finite. Our approach is based on obtaining matching representations of the confounders for the various treatment types. This is achieved through minimization of an upper bound on the sum of factual and counterfactual losses. Experiments on real and semi-synthetic datasets show that MEMENTO is able to outperform known techniques for multi-treatment scenarios by close to 10% in certain use-cases. The proposed framework has been deployed for the problem of identifying minimum order quantity of a product in Amazon in an emerging marketplace and has re- sulted in a 4.7% reduction in shipping costs as proved from an A/B experiment.

Speed of delivery is critical for the success of e-commerce platforms. Faster delivery promise to the customer results in increased conversion and revenue. There are typically two mechanisms to control the delivery speed - a) replication of products across warehouses, and b) air-shipping the product. In this paper, we present a machine learning based framework to recommend air-shipping eligibility for products. Specifically, we develop a causal inference framework (referred to as Air Shipping Recommendation or ASPIRE) that balances the trade-off between revenue or conversion and delivery cost to decide whether a product should be shipped via air. We propose a doubly-robust estimation technique followed by an optimization algorithm to determine air eligibility of products and calculate the uplift in revenue and shipping cost.

Product reviews and ratings on e-commerce websites provide customers with detailed insights about various aspects of the product such as quality, usefulness, etc. Since they influence customers' buying decisions, product reviews have become a fertile ground for abuse by sellers (colluding with reviewers) to promote their own products or to tarnish the reputation of competitor's products. In this paper, our focus is on detecting such abusive entities (both sellers and reviewers) by applying tensor decomposition on the product reviews data. While tensor decomposition is mostly unsupervised, we formulate our problem as a semi-supervised binary multi-target tensor decomposition, to take advantage of currently known abusive entities. We empirically show that our multi-target semi-supervised model achieves higher precision and recall in detecting abusive entities as compared to unsupervised techniques. Finally, we show that our proposed stochastic partial natural gradient inference for our model empirically achieves faster convergence than stochastic gradient and Online-EM with sufficient statistics.

Machine Learning (ML) has become a mature technology that is being applied to a wide range of business problems such as web search, online advertising, product recommendations, object recognition, and so on. As a result, it has become imperative for researchers and practitioners to have a fundamental understanding of ML concepts and practical knowledge of end-to-end modeling. This tutorial takes a hands-on approach to introducing the audience to machine learning. The first part of the tutorial gives a broad overview and discusses some of the key concepts within machine learning. The second part of the tutorial takes the audience through the end-to-end modeling pipeline for a real-world income prediction problem.

Network structure and content in microblogging sites like Twitter influence each other ---user A on Twitter follows user B for the tweets that B posts on the network, and A may then re-tweet the content shared by B to his/her own followers. In this paper, we propose a probabilistic model to jointly model link communities and content topics by leveraging both the social graph and the content shared by users. We model a community as a distribution over users, use it as a source for topics of interest, and jointly infer both communities and topics using Gibbs sampling. While modeling communities using the social graph, or modeling topics using content have received a great deal of attention, a few recent approaches try to model topics in content-sharing platforms using both content and social graph. Our work differs from the existing generative models in that we explicitly model the social graph of users along with the user-generated content, mimicking how the two entities co-evolve in content-sharing platforms. Recent studies have found Twitter to be more of a content-sharing network and less a social network, and it seems hard to detect tightly knit communities from the follower-followee links. Still, the question of whether we can extract Twitter communities using both links and content is open. In this paper, we answer this question in the affirmative. Our model discovers coherent communities and topics, as evinced by qualitative results on sub-graphs of Twitter users. Furthermore, we evaluate our model on the task of predicting follower-followee links. We show that joint modeling of links and content significantly improves link prediction performance on a sub-graph of Twitter (consisting of about 0.7 million users and over 27 million tweets), compared to generative models based on only structure or only content and paths-based methods such as Katz.

Content sharing in social networks is a powerful mechanism for discovering content on the Internet. The degree to which content is disseminated within the network depends on the connectivity relationships among network nodes. Existing schemes for recommending connections in social networks are based on the number of common neighbors, similarity of user profiles, etc. However, such similarity-based connections do not consider the amount of content discovered.

A wide variety of clustering algorithms exist that cater to applications based on certain special characteristics of the data. Our focus is on methods that capture arbitrary shaped clusters in data, the so called spatial clustering algorithms. With the growing size of spatial datasets from diverse sources, the need for scalable algorithms is paramount. We propose a shape-based clustering algorithm, ABACUS, that scales to large datasets. ABACUS is based on the idea of identifying the intrinsic structure for each cluster, which we also refer to as the backbone of that cluster. The backbone comprises of a much smaller set of points, thus giving this method the desired ability to scale to larger datasets. ABACUS operates in two stages. In the first stage, we identify the backbone of each cluster via an iterative process made up of globbing (or point merging) and point movement operations. The backbone enables easy identification of the true clusters in a subsequent stage. Experiments on a range of real (images from geospatial satellites, etc.) and synthetic datasets demonstrate the efficiency and effectiveness of our approach. In particular, ABACUS is over an order of magnitude faster than existing shape-based clustering methods, yet it provides a comparable or better clustering quality.

Online social networks offer opportunities to analyze user behavior and social connectivity and leverage resulting in-sights for effective online advertising. We study the adoption of a paid product by members of a large and well-connected Instant Messenger (IM) network. This product is important to the business and poses unique challenges to advertising due to its low baseline adoption rate. We find that adop-tion by highly connected individuals is correlated with their social connections (friends) adopting after them. However, there is little evidence of social influence by these high degree individuals. Further, the spread of adoption remains mostly local to first-adopters and their immediate friends. We ob-serve strong evidence of peer pressure wherein future adop-tion by an individual is more likely if the product has been

The Semantic Web community, until now, has used traditional database systems for the storage and querying of RDF data. The SPARQL query language also closely follows SQL syntax. As a natural consequence, most of the SPARQL query processing techniques are based on database query processing and optimization techniques. For SPARQL join query optimization, previous works like RDF-3X and Hexastore have proposed to use 6-way indexes on the RDF data. Although these indexes speed up merge-joins by orders of magnitude, for complex join queries generating large intermediate join results, the scalability of the query processor still remains a challenge.

Given a large directed graph, rapidly answering reachability queries between source and target nodes is an important problem. Existing methods for reachability trade-off indexing time and space versus query time performance. However, the biggest limitation of existing methods is that they simply do not scale to very large real-world graphs. We present a very simple, but scalable reachability index, called GRAIL, that is based on the idea of randomized interval labeling, and that can effectively handle very large graphs. Based on an extensive set of experiments, we show that while more sophisticated methods work better on small graphs, GRAIL is the only index that can scale to millions of nodes and edges. GRAIL has linear indexing time and space, and the query time ranges from constant time to being linear in the graph order and size.

Clustering is one of the fundamental data mining tasks. Many different clustering paradigms have been developed over the years, which include partitional, hierarchical, mixture model based, density-based, spectral, subspace, and so on. The focus of this paper is on full-dimensional, arbitrary shaped clusters. Existing methods for this problem suffer either in terms of the memory or time complexity (quadratic or even cubic). This shortcoming has restricted these algorithms to datasets of moderate sizes. In this paper we propose SPARCL, a simple and scalable algorithm for finding clusters with arbitrary shapes and sizes, and it has linear space and time complexity. SPARCL consists of two stages - the first stage runs a carefully initialized version of the K-means algorithm to generate many small seed clusters. The second stage iteratively merges the generated clusters to obtain the final shape-based clusters. Experiments were conducted on a variety of datasets to highlight the effectiveness, efficiency, and scalability of our approach. On the large datasets SPARCL is an order of magnitude faster than the best existing approaches.

In this paper, we introduce the concept of alpha-orthogonal patterns to mine a representative set of graph patterns. Intuitively, two graph patterns are alpha-orthogonal if their similarity is bounded above by alpha. Each alpha-orthogonal pattern is also a representative for those patterns that are at least beta similar to it. Given user defined alpha, beta isin [0,1], the goal is to mine an alpha-orthogonal, beta-representative set that minimizes the set of unrepresented patterns. We present ORIGAMI, an effective algorithm for mining the set of representative orthogonal patterns. ORIGAMI first uses a randomized algorithm to randomly traverse the pattern space, seeking previously unexplored regions, to return a set of maximal patterns. ORIGAMI then extracts an alpha-orthogonal, beta-representative set from the mined maximal patterns. We show the effectiveness of our algorithm on a number of real and synthetic datasets. In particular, we show that our method is able to extract high quality patterns even in cases where existing enumerative graph mining methods fail to do so.