Abstract Dense embedding-based semantic matching is widely used in e-commerce product search to address the shortcomings of lexical matching such as sensitivity to spelling variants. The recent advances in BERT-like language model encoders, have however, not found their way to realtime search due to the strict inference latency requirement imposed on e-commerce websites. While bi-encoder BERT architectures enable fast approximate nearest neighbor search, training them effectively on query-product data remains a challenge due to training instabilities and the persistent generalization gap with cross-encoders. In this work, we propose a four-stage training procedure to leverage large BERT-like models for product search while preserving low inference latency. We introduce query-product interaction pre-finetuning to effectively pretrain BERT bi-encoders for matching and improve generalization. Through offline experiments on an e-commerce product dataset, we show that a distilled small BERT-based model (75M params) trained using our approach improves the search relevance metric by up to 23% over a baseline DSSM-based model with similar inference latency. The small model only suffers a 3% drop in relevance metric compared to the 20x larger teacher. We also show using online A/B tests at scale, that our approach improves over the production model in exact and substitute products retrieved.

Templated graphical models (TGMs) encode model structure using rules that capture recurring relationships between multiple random variables. While the rules in TGMs are interpretable, it is not clear how they can be used to generate explanations for the individual predictions of the model. Further, learning these rules from data comes with high computational costs: it typically requires an expensive combinatorial search over the space of rules and repeated optimization over rule weights. In this work, we propose a new structure learning algorithm, Explainable Structured Model Search (ESMS), that learns a templated graphical model and an explanation framework for its predictions. ESMS uses a novel search procedure to efficiently search the space of models and discover models that trade-off predictive accuracy and explainability. We introduce the notion of relational stability and prove that our proposed explanation framework is stable. Further, our proposed piecewise pseudolikelihood (PPLL) objective does not require re-optimizing the rule weights across models during each iteration of the search. In our empirical evaluation on three realworld datasets, we show that our proposed approach not only discovers models that are explainable, but also significantly outperforms existing state-out-the-art structure learning approaches.

Abstract Statistical relational learning (SRL) frameworks are effective at defining probabilistic models over complex relational data. They often use weighted first-order logical rules where the weights of the rules govern probabilistic interactions and are usually learned from data. Existing weight learning approaches typically attempt to learn a set of weights that maximizes some function of data likelihood; however, this does not always translate to optimal performance on a desired domain metric, such as accuracy or F1 score. In this paper, we introduce a taxonomy of search-based weight learning approaches for SRL frameworks that directly optimize weights on a chosen domain performance metric. To effectively apply these search-based approaches, we introduce a novel projection, referred to as scaled space (SS), that is an accurate representation of the true weight space. We show that SS removes redundancies in the weight space and captures the semantic distance between the possible weight configurations. In order to improve the efficiency of search, we also introduce an approximation of SS which simplifies the process of sampling weight configurations. We demonstrate these approaches on two state-of-the-art SRL frameworks: Markov logic networks and probabilistic soft logic. We perform empirical evaluation on five real-world datasets and evaluate them each on two different metrics. We also compare them against four other weight learning approaches. Our experimental results show that our proposed search-based approaches outperform likelihood-based approaches and yield up to a 10% improvement across a variety of performance metrics. Further, we perform an extensive evaluation to measure the robustness of our approach to different initializations and hyperparameters. The results indicate that our approach is both accurate and robust.

Abstract Statistical relational learning (SRL) and graph neural networks (GNNs) are two powerful approaches for learning and inference over graphs. Typically, they are evaluated in terms of simple metrics such as accuracy over individual node labels. Complexaggregate graph queries(AGQ) involving multiple nodes, edges, and labels are common in the graph mining community and are used to estimate important network properties such as social cohesion and influence. While graph mining algorithms support AGQs, they typically do not take into account uncertainty, or when they do, make simplifying assumptions and do not build full probabilistic models. In this paper, we examine the performance of SRL and GNNs on AGQs over graphs with partially observed node labels. We show that, not surprisingly, inferring the unobserved node labels as a first step and then evaluating the queries on the fully observed graph can lead to sub-optimal estimates, and that a better approach is to compute these queries as an expectation under the joint distribution. We propose a sampling framework to tractably compute the expected values of AGQs. Motivated by the analysis of subgroup cohesion in social networks, we propose a suite of AGQs that estimate the community structure in graphs. In our empirical evaluation, we show that by estimating these queries as an expectation, SRL-based approaches yield up to a 50-fold reduction in average error when compared to existing GNN-based approaches.

Statistical relational learning (SRL) frameworks allow users to create large, complex graphical models using a compact, rule-based representation. However, these models can quickly become prohibitively large and not fit into machine memory. In this work we address this issue by introducing a novel technique called tandem inference (ti). The primary idea of ti is to combine grounding and inference such that both processes happen in tandem. ti uses an out-of-core streaming approach to overcome memory limitations. Even when memory is not an issue, we show that our proposed approach is able to do inference faster while using less memory than existing approaches. To show the effectiveness of ti, we use a popular SRL framework called Probabilistic Soft Logic (PSL). We implement ti for PSL by proposing a gradient-based inference engine and a streaming approach to grounding. We show that we are able to run an SRL model with over 1B cliques in under nine hours and using only 10 GB of RAM; previous approaches required more than 800 GB for this model and are infeasible on common hardware. To the best of our knowledge, this is the largest SRL model ever run.

Probabilistic soft logic (PSL) is a statistical relational learning framework that represents complex relational models with weighted first-order logical rules. The weights of the rules in PSL indicate their importance in the model and influence the effectiveness of the model on a given task. Existing weight learning approaches often attempt to learn a set of weights that maximizes some function of data likelihood. However, this does not always translate to optimal performance on a desired domain metric, such as accuracy or F1 score. In this paper, we introduce a new weight learning approach called Bayesian optimization for weight learning (BOWL) based on Gaussian process regression that directly optimizes weights on a chosen domain performance metric. The key to the success of our approach is a novel projection that captures the semantic distance between the possible weight configurations. Our experimental results show that our proposed approach outperforms likelihood-based approaches and yields up to a 10% improvement across a variety of performance metrics. Further, we performed experiments to measure the scalability and robustness of our approach on various realworld datasets.

Fast propagation, ease-of-access, and low cost have made social media an increasingly popular means for news consumption. However, this has also led to an increase in the preponderance of fake news. Widespread propagation of fake news can be detrimental to society, and this has created enormous interest in fake news detection on social media. Many approaches to fake news detection use the news content, social context, or both. In this work, we look at fake news detection as a problem of estimating the credibility of both the news publishers and users that propagate news articles. We introduce a new approach called the credibility score-based model that can jointly infer fake news and credibility scores for publishers and users. We use a state-of-the-art statistical relational learning framework called probabilistic soft logic to perform this joint inference effectively. We show that our approach is accurate at both fake news detection and inferring credibility scores. Further, our model can easily integrate any auxiliary information that can aid in fake news detection. Using the FakeNewsNet dataset, we show that our approach significantly outperforms previous approaches at fake news detection by up to 10% in recall and 4% in accuracy. Furthermore, the credibility scores learned for both publishers and users are representative of their true behavior.

We develop a variant of multiclass logistic regression that is significantly more robust to noise. The algorithm has one weight vector per class and the surrogate loss is a function of the linear activations (one per class). The surrogate loss of an example with linear activation vector $\mathbf{a}$ and class $c$ has the form $-\log_{t_1} \exp_{t_2} (a_c - G_{t_2}(\mathbf{a}))$ where the two temperatures $t_1$ and $t_2$ “temper” the $\log$ and $\exp$, respectively, and $G_{t_2}(\mathbf{a})$ is a scalar value that generalizes the log-partition function. We motivate this loss using the Tsallis divergence. Our method allows transitioning between non-convex and convex losses by the choice of the temperature parameters. As the temperature $t_1$ of the logarithm becomes smaller than the temperature $t_2$ of the exponential, the surrogate loss becomes “quasi convex”. Various tunings of the temperatures recover previous methods and tuning the degree of non-convexity is crucial in the experiments. In particular, quasi-convexity and boundedness of the loss provide significant robustness to the outliers. We explain this by showing that $t_1 < 1$ caps the surrogate loss and $t_2 >1$ makes the predictive distribution have a heavy tail. We show that the surrogate loss is Bayes-consistent, even in the non-convex case. Additionally, we provide efficient iterative algorithms for calculating the log-partition value only in a few number of iterations. Our compelling experimental results on large real-world datasets show the advantage of using the two-temperature variant in the noisy as well as the noise free case.

E-commerce search engines are the primary means by which customers shop for products online. Each customer query contains multiple facets such as product type, color, brand, etc. A successful search engine retrieves products that are relevant to the query along each of these attributes. However, due to lexical (erroneous title, description, etc.) and behavioral irregularities (clicks or purchases of products that do not belong to the same facet as the query), some mismatched products are often included in search results. These irregularities can be detected using simple binary classifiers like gradient boosted decision trees or logistic regression. Typically, these binary classifiers use strong independence assumptions between the results and ignore structural relationships available in the data, such as the connections between products and queries. In this paper, we use the connections that exist between products and query to identify a special kind of structure we refer to as a micrograph. Further, we make use of Statistical Relational Learning (SRL) to incorporate these micrographs in the data and pose the problem as a structured prediction problem. We refer to this approach as structured mismatch classification (\SMC). In addition, we show that naive addition of structure does not improve the performance of the model and hence introduce a variation of \SMC, strong \SMC~(\SSMC), which improves over the baseline by passing information from high-confidence predictions to lower confidence predictions. In our empirical evaluation we show that our proposed approach outperforms the baseline classification methods by up to 12% in precision. Furthermore, we use quasi-Newton methods to make our method viable for real-time inference in a search engine and show that our approach is up to 150 times faster than existing ADMM-based solvers.

We study the problem of scaling Multinomial Logistic Regression (MLR) to datasets with very large number of data points in the presence of large number of classes. At a scale where neither data nor the parameters are able to fit on a single machine, we argue that simultaneous data and model parallelism (Hybrid Parallelism) is inevitable. The key challenge in achieving such a form of parallelism in MLR is the log-partition function which needs to be computed across all K classes per data point, thus making model parallelism non-trivial. To overcome this problem, we propose a reformulation of the original objective that exploits double-separability, an attractive property that naturally leads to hybrid parallelism. Our algorithm (DS-MLR) is asynchronous and completely de-centralized, requiring minimal communication across workers while keeping both data and parameter workloads partitioned. Unlike standard data parallel approaches, DS-MLR avoids bulk-synchronization by maintaining local normalization terms on each worker and accumulating them incrementally using a token-ring topology. We demonstrate the versatility of DS-MLR under various scenarios in data and model parallelism, through an empirical study consisting of real-world datasets. In particular, to demonstrate scaling via hybrid parallelism, we created a new benchmark dataset (Reddit-Full) by pre-processing 1.7 billion reddit user comments spanning the period 2007-2015. We used DS-MLR to solve an extreme multi-class classification problem of classifying 211 million data points into their corresponding subreddits. Reddit-Full is a massive data set with data occupying 228 GB and 44 billion parameters occupying 358 GB. To the best of our knowledge, no other existing methods can handle MLR in this setting.

Search queries are appropriate when users have explicit intent, but they perform poorly when the intent is difficult to express or if the user is simply looking to be inspired. Visual browsing systems allow e-commerce platforms to address these scenarios while offering the user an engaging shopping experience. Here we explore extensions in the direction of adaptive personalization and item diversification within Stream, a new form of visual browsing and discovery by Amazon. Our system presents the user with a diverse set of interesting items while adapting to user interactions. Our solution consists of three components (1) a Bayesian regression model for scoring the relevance of items while leveraging uncertainty, (2) a submodular diversification framework that re-ranks the top scoring items based on category, and (3) personalized category preferences learned from the user's behavior. When tested on live traffic, our algorithms show a strong lift in click-through-rate and session duration.