Re-ranking plays a crucial role in product search by reassessing products from the primary retrieval system based on specific engagement and relevance criteria. While transformer-based models like the cross encoder have advanced the relevance of ranking models in recent years, a significant challenge arises from the high latency cost associated with running a cross encoder model at runtime. This challenge becomes more pronounced in the long-tail segment, where conventional techniques like caching prove ineffective. To tackle these issues, our paper introduces a scalable framework featuring a BERT-based cross encoder model for re-ranking, deployed in the Walmart search engine. We employ strategies such as intermediate representations, operator fusion, and vectorization to improve the inference latency of the cross encoder model. Furthermore, we provide a detailed discussion on the runtime implementation, highlighting key learnings and practical tricks that ensured minimal impact on response latency during production. Finally, we present the results of online experiments, including manual evaluation and interleaving test conducted on real-world e-commerce search traffic.

Existing prompting strategies for adapting pretrained vision language models to the downstream task of finegrained attribute classification learn visual variance in a class-specific manner. We present DisPoL, a method for learning disentangled representations that improves the transferability and performance of continuous prompts for downstream classification tasks. Our method decomposes a prompt into separate sub-prompts, then performs late fusion of the corresponding output embeddings in a novel manner. We combine the fixed embedding of a static, context-constraining object sub-prompt and the tunable embedding of a soft sub-prompt for a task-specific attribute using self-attention. By avoiding joint learning of these tokens, the resulting disentangled prompt embeddings are more transferable to unseen objects. We also demonstrate how to use hand-crafted templates to initialize the task-specific soft prompt, improving training efficiency. Through extensive experiments, we show that DisPoL exceeds the performance of existing methods in few-shot settings and highlight its contribution as a parameter-efficient fine-tuning method.

Embedding-based neural retrieval (EBR) is an effective search retrieval method in product search for tackling the vocabulary gap between customer search queries and products. The initial launch of our EBR system at Walmart yielded significant gains in relevance and add-to-cart rates [1]. However, despite EBR generally retrieving more relevant products for reranking, we have observed numerous instances of relevance degradation. Enhancing retrieval performance is crucial, as it directly influences product reranking and affects the customer shopping experience. Factors contributing to these degradations include false positives/negatives in the training data and the inability to handle query misspellings. To address these issues, we present several approaches to further strengthen the capabilities of our EBR model in terms of retrieval relevance. We introduce a Relevance Reward Model (RRM) based on human relevance feedback. We utilize RRM to remove noise from the training data and distill it into our EBR model through a multi-objective loss. In addition, we present the techniques to increase the performance of our EBR model, such as typo-aware training, and semi-positive generation. The effectiveness of our EBR is demonstrated through offline relevance evaluation, online AB tests, and successful deployments to live production.

In embedding-based retrieval, Approximate Nearest Neighbor (ANN) search enables efficient retrieval of similar items from large-scale datasets. While maximizing recall of relevant items is usually the goal of retrieval systems, a low precision may lead to a poor search experience. Unlike lexical retrieval, which inherently limits the size of the retrieved set through keyword matching, dense retrieval via ANN search has no natural cutoff. Moreover, the cosine similarity scores of embedding vectors are often optimized via contrastive or ranking losses, which make them difficult to interpret. Consequently, relying on top-K or cosine-similarity cutoff is often insufficient to filter out irrelevant results effectively. This issue is prominent in product search, where the number of relevant products is often small. This paper introduces a novel relevance filtering component (called "Cosine Adapter") for embedding-based retrieval to address this challenge. Our approach maps raw cosine similarity scores to interpretable scores using a query-dependent mapping function. We then apply a global threshold on the mapped scores to filter out irrelevant results. We are able to significantly increase the precision of the retrieved set, at the expense of a small loss of recall. The effectiveness of our approach is demonstrated through experiments on both public MS MARCO dataset and internal Walmart product search data. Furthermore, online A/B testing on the Walmart site validates the practical value of our approach in real-world e-commerce settings.

Query Understanding (QU) is a fundamental process in E-commerce search engines by extracting the shopping intents of customers. It usually includes a set of different tasks such as named entity recognization and query classification. Traditional approaches often tackle each task separately by its own network, which leads to excessive workload for development and maintenance as well as increased latency and resource usage in large-scale E-commerce platforms. To tackle these challenges, this paper presents a multi-task learning approach to query understanding at Walmart. We experimented with several state-of-the-art multi-task learning architectures including MTDNN, MMoE, and PLE. Furthermore, we propose a novel large-scale entity-aware multi-task learning model (EAMT)1 by retrieving entities from engagement data as query context to augment the query representation. To the best of our knowledge, there exists no prior work on multi-task learning for E-commerce query understanding. Comprehensive offline experiments are conducted on industry-scale datasets (up to 965M queries) to illustrate the effectiveness of our approach. The results from online experiments show substantial gains in key accuracy and latency metrics. https://github.com/zhiyuanpeng/KDD2023-EAMT

In product search, the retrieval of candidate products before re-ranking is more mission critical and challenging than other search like web search, especially for tail queries, which have a complex and specific search intent. In this paper, we present a hybrid system for e-commerce search deployed at Walmart that combines traditional inverted index and embedding-based neural retrieval to better answer user tail queries. Our system significantly improved the relevance of the search engine, measured by both offline and online evaluations. The improvements were achieved through a combination of different approaches. We present a new technique to train the neural model at scale. and describe how the system was deployed in production with little impact on response time. We highlight multiple learnings and practical tricks that were used in the deployment of this system.

Product ranking is a crucial component for many e-commerce services. One of the major challenges in product search is the vocabulary mismatch between query and products, which may be a larger vocabulary gap problem compared to other information retrieval domains. While there is a growing collection of neural learning to match methods aimed specifically at overcoming this issue, they do not leverage the recent advances of large language models for product search. On the other hand, product ranking often deals with multiple types of engagement signals such as clicks, add-to-cart, and purchases, while most of the existing works are focused on optimizing one single metric such as click-through rate, which may suffer from data sparsity. In this work, we propose a novel end-to-end multi-task learning framework for product ranking with BERT to address the above challenges. The proposed model utilizes domain-specific BERT with fine-tuning to bridge the vocabulary gap and employs multi-task learning to optimize multiple objectives simultaneously, which yields a general end-to-end learning framework for product search. We conduct a set of comprehensive experiments on a real-world e-commerce dataset and demonstrate significant improvement of the proposed approach over the state-of-the-art baseline methods.

In e-commerce, product content, especially product images have a significant influence on a customer's journey from product discovery to evaluation and finally, purchase decision. Since many e-commerce retailers sell items from other third-party marketplace sellers besides their own, the content published by both internal and external content creators needs to be monitored and enriched, wherever possible. Despite guidelines and warnings, product listings that contain offensive and non-compliant images continue to enter catalogs. Offensive and non-compliant content can include a wide range of objects, logos, and banners conveying violent, sexually explicit, racist, or promotional messages. Such images can severely damage the customer experience, lead to legal issues, and erode the company brand. In this paper, we present a computer vision driven offensive and non-compliant image detection system for extremely large image datasets. This paper delves into the unique challenges of applying deep learning to real-world product image data from retail world. We demonstrate how we resolve a number of technical challenges such as lack of training data, severe class imbalance, fine-grained class definitions etc. using a number of practical yet unique technical strategies. Our system combines state-of-the-art image classification and object detection techniques with budgeted crowd-sourcing to develop a solution customized for a massive, diverse, and constantly evolving product catalog.

When shopping for fashion, customers often look for products which can complement their current outfit. For example, customers want to buy a jacket which can go well with their jeans and sneakers. To address the task of fashion matching, we propose a neural compatibility model for ranking fashion products based on the compatibility matching with the input outfit. The contribution of our work is twofold. First, we demonstrate that product descriptions contain rich information about product comparability which has not been fully utilized in the prior work. Secondly, we exploit such useful information from text data by taking advantages of semantic matching and lexical matching both of which are important for fashion matching. The proposed model is evaluated on a real-world fashion outfit dataset and achieves the state-of-the-art results by comparing to the competitive baselines. In the future work, we plan to extend the model by incorporating product images which are the major data source in the prior work on fashion matching.

Classifying products precisely and efficiently is a major challenge in modern e-commerce. The high traffic of new products uploaded daily and the dynamic nature of the categories raise the need for machine learning models that can reduce the cost and time of human editors. In this paper, we propose a decision level fusion approach for multi-modal product classification based on text and image neural network classifiers. We train input specific state-of-the-art deep neural networks for each input source, show the potential of forging them together into a multi-modal architecture and train a novel policy network that learns to choose between them. Finally, we demonstrate that our multi-modal network improves classification accuracy over both networks on a real-world large-scale product classification dataset that we collected from Walmart.com. While we focus on image-text fusion that characterizes e-commerce businesses, our algorithms can be easily applied to other modalities such as audio, video, physical sensors, etc.

In e-commerce, content quality of the product catalog plays a key role in delivering a satisfactory experience to the customers. In particular, visual content such as product images influences customers' engagement and purchase decisions. With the rapid growth of e-commerce and the advent of artificial intelligence, traditional content management systems are giving way to automated scalable systems. In this paper, we present a machine learning driven visual content management system for extremely large e-commerce catalogs. For a given product, the system aggregates images from various suppliers, understands and analyzes them to produce a superior image set with optimal image count and quality, and arranges them in an order tailored to the demands of the customers. The system makes use of an array of technologies, ranging from deep learning to traditional computer vision, at different stages of analysis. In this paper, we outline how the system works and discuss the unique challenges related to applying machine learning techniques to real-world data from e-commerce domain. We emphasize how we tune state-of-the-art image classification techniques to develop solutions custom made for a massive, diverse, and constantly evolving product catalog. We also provide the details of how we measure the system's impact on various customer engagement metrics.

The performance of a kernel-based learning algorithm depends very much on the choice of the kernel. Recently, much attention has been paid to the problem of learning the kernel itself from given training examples. The main emphasis has been on formulating the problem as a tractable convex optimization problem. Only for a few very special cases such as support vector machines are explicit convex formulations known. In this paper, we show that, in a wide variety of kernel-based learning algorithms, the kernel learning problem can be formulated as a convex optimization problem which interior-point methods can solve globally and efficiently. The kernel learning method is illustrated with a regression problem that arises in petroleum engineering.

Wireless sensors are expected to be widely deployed in the near future for a vast variety of applications. Sensing data is not useful unless the location where the data is collected is also known to the end users. Although much work has been done on the positioning of sensor nodes, proposed algorithms often are not applicable in certain environments. Furthermore, there is always a need to reduce their complexity and cost, and to improve accuracy for these algorithms. This work is primarily motivated by a new research project called WINES to deploy wireless sensors to monitor the conditions of London underground and water systems. Global Positioning System (GPS) based positioning methods are not applicable in the underground environments. In addition, the tunnels and water pipes are not located at the same horizontal level, so existing algorithms proposed for two-dimensional deployment areas become inadequate. To overcome the challenge, we propose and study here a new localization algorithm by extending one proposed for two-dimensional service areas by Capkun et al. Our new scheme is a distributed, self-organized, infrastructure-free positioning algorithm that enables easy and flexible sensor deployment in the harsh environments. Furthermore, our computer simulation reveals that the new scheme achieves a satisfactory, relative average position error of less than 5% when the errors for distance estimations among sensors have a standard deviation of no more than 5%.

In Kernel Fisher discriminant analysis (KFDA), we carry out Fisher linear discriminant analysis in a high dimensional feature space defined implicitly by a kernel. The performance of KFDA depends on the choice of the kernel; in this paper, we consider the problem of finding the optimal kernel, over a given convex set of kernels. We show that this optimal kernel selection problem can be reformulated as a tractable convex optimization problem which interior-point methods can solve globally and efficiently. The kernel selection method is demonstrated with some UCI machine learning benchmark examples.

We consider the problem of choosing a linear classifier that minimizes misclassification probabilities in two-class classification, which is a bi-criterion problem, involving a trade-off between two objectives. We assume that the class-conditional distributions are Gaussian. This assumption makes it computationally tractable to find Pareto optimal linear classifiers whose classification capabilities are inferior to no other linear ones. The main purpose of this paper is to establish several robustness properties of those classifiers with respect to variations and uncertainties in the distributions. We also extend the results to kernel-based classification. Finally, we show how to carry out trade-off analysis empirically with a finite number of given labeled data.

We consider the problem of finding the optimal one-bit quantizer for symmetric source distributions, with the Euclidean norm as the measure of distortion. For fixed rate quantizers, we prove that for (symmetric) monotonically decreasing source distributions with ellipsoidal level curves, the centroids of the optimal 1-bit quantizer must be on the major axis of the ellipsoids. Under the same assumptions on the source distribution, the centroids of the optimal one-bit variable-rate quantizer lie on one of the axes of the ellipsoid. If further, the source distribution f(x) is log-concave in x, the optimal 1-bit fixed-rate quantizer is unique and symmetric about the origin. (The Gaussian is an example of a distribution that satisfies all these conditions.) Under a further set of conditions on the source distributions, we show that there is a threshold below which the optimal fixed rate and variable rate quantizer are the same.

Fisher linear discriminant analysis (LDA) can be sensitive to the problem data. Robust Fisher LDA can systematically alleviate the sensitivity problem by explicitly incorporating a model of data uncertainty in a classification problem and optimizing for the worst-case scenario under this model. The main contribution of this paper is show that with general convex uncertainty models on the problem data, robust Fisher LDA can be carried out using convex optimization. For a certain type of product form uncertainty model, robust Fisher LDA can be carried out at a cost comparable to standard Fisher LDA. The method is demonstrated with some numerical examples. Finally, we show how to extend these results to robust kernel Fisher discriminant analysis, i.e., robust Fisher LDA in a high dimensional feature space.