Query understanding is essential in modern relevance systems, where user queries are often short, ambiguous, and highly context-dependent. Traditional approaches often rely on multiple task-specific Named Entity Recognition models to extract structured facets as seen in job search applications. However, this fragmented architecture is brittle, expensive to maintain, and slow to adapt to evolving taxonomies and language patterns. In this paper, we introduce a unified query understanding framework powered by a Large Language Model (LLM), designed to address these limitations. Our approach jointly models the user query and contextual signals such as profile attributes to generate structured interpretations that drive more accurate and personalized recommendations. The framework improves relevance quality in online A/B testing while significantly reducing system complexity and operational overhead. The results demonstrate that our solution provides a scalable and adaptable foundation for query understanding in dynamic web applications.

LinkedIn, one of the world's largest platforms for professional networking and job seeking, encounters various modeling challenges in building recommendation systems for its job matching product, including cold-start, filter bubbles, and biases affecting candidate-job matching. To address these, we developed the STAR (Signal integration for Talent And Recruiters) system, leveraging the combined strengths of Large Language Models (LLMs) and Graph Neural Networks (GNNs). LLMs excel at understanding textual data, such as member profiles and job postings, while GNNs capture intricate relationships and mitigate cold-start issues through network effects. STAR integrates diverse signals by uniting LLM and GNN capabilities with industrial-scale paradigms including adaptive sampling and version management. It provides an end-to-end solution for developing and deploying embeddings in large-scale recommender systems. Our key contributions include a robust methodology for building embeddings in industrial applications, a scalable GNN-LLM integration for high-performing recommendations, and practical insights for real-world model deployment.

Web-scale search and recommendation systems depend on efficient retrieval to manage massive datasets and user traffic.This paper chronicles our evolutionary path in building the retrieval layer at LinkedIn, progressing from a CPU-based inverted index system to a GPU-accelerated embedding-based retrieval system.Initially anchored by traditional term-based retrieval, we enhanced relevance and productivity through learning-to-retrieve approaches by generating mappings among inferred attributes.As these early efforts encountered limitations in inferring and matching attributes at scale, we transitioned to embedding-based retrieval for greater flexibility and performance, but found that existing infrastructure couldn't support large-scale production needs.This led us to develop a GPUbased retrieval system designed for high performance, flexible modeling, and multi-objective business optimization.We present the infrastructure innovations, optimizations, and key lessons learned throughout this transition, offering practical insights for building scalable, flexible retrieval systems.

Graph recommendation methods, representing a connected interaction perspective, reformulate user–item interactions as graphs to leverage graph structure and topology to recommend and have proved practical effectiveness at scale. Large language models (LLMs), representing a textual generative perspective, excel at modeling user languages, understanding behavioral contexts, capturing user–item semantic relationships, analyzing textual sentiments, and generating coherent and contextually relevant texts as recommendations. However, there is a gap between the connected graph perspective and the text generation perspective as the task formulations are different. A research question arises: how can we effectively integrate the two perspectives for more personalized RecSys? To fill this gap, we propose to incorporate graph-edge information into LLMs via prompt and attention innovations. We reformulate recommendations as a probabilistic generative problem using prompts. We develop a framework to incorporate graph edge information from the prompt and attention mechanisms for graph-structured LLM recommendations. We develop a new prompt design that brings in both first-order and second-order graph relationships; we devise an improved LLM attention mechanism to embed direct the spatial and connectivity information of edges. Our evaluation of real-world datasets demonstrates the framework’s ability to understand connectivity information in graph data and to improve the relevance and quality of recommendation results. Our code is released at: https://github.com/anord-wang/LLM4REC.git .

Web-scale search systems typically tackle the scalability challenge with a two-step paradigm: retrieval and ranking. The retrieval step, also known as candidate selection, often involves extracting entities, creating an inverted index, and performing term matching for retrieval. Such traditional methods require manual and time-consuming development of retrieval models. In this paper, we propose a framework for constructing a graph that integrates human knowledge with user activity data analysis. The learned links are utilized for retrieval purposes. The model is easy to explain, debug, and tune. The system implementation is straightforward and can directly leverage existing inverted index systems. We applied this retrieval framework to enhance the job search and recommendation systems on a large professional networking portal, resulting in significant performance improvements.

Job marketplace is a heterogeneous graph composed of interactions among members (job-seekers), companies, and jobs. Understanding and modeling job marketplace can benefit both job seekers and employers, ultimately contributing to the greater good of the society. However, existing graph neural network (GNN)-based methods have shallow understandings of the associated textual features and heterogeneous relations. To address the above challenges, we propose PLM4Job, a job marketplace foundation model that tightly couples pretrained language models (PLM) with job market graph, aiming to fully utilize the pretrained knowledge and reasoning ability to model member/job textual features as well as various member-job relations simultaneously. In the pretraining phase, we propose a heterogeneous ego-graph-based prompting strategy to model and aggregate member/job textual features based on the topological structure around the target member/job node, where entity type embeddings and graph positional embeddings are introduced accordingly to model different entities and their heterogeneous relations. Meanwhile, a proximity-aware attention alignment strategy is designed to dynamically adjust the attention of the PLM on ego-graph node tokens in the prompt, such that the attention can be better aligned with job marketplace semantics. Extensive experiments at LinkedIn demonstrate the effectiveness of PLM4Job.

Recently, there has been growing interest in developing the next-generation recommender systems (RSs) based on pretrained large language models (LLMs). However, the semantic gap between natural language and recommendation tasks is still not well addressed, leading to multiple issues such as spuriously correlated user/item descriptors, ineffective language modeling on user/item data, inefficient recommendations via auto-regression, etc. In this paper, we propose CLLM4Rec, the first generative RS that tightly integrates the LLM paradigm and ID paradigm of RSs, aiming to address the above challenges simultaneously. We first extend the vocabulary of pretrained LLMs with user/item ID tokens to faithfully model user/item collaborative and content semantics. Accordingly, a novel soft+hard prompting strategy is proposed to effectively learn user/item collaborative/content token embeddings via language modeling on RS-specific corpora, where each document is split into a prompt consisting of heterogeneous soft (user/item) tokens and hard (vocab) tokens and a main text consisting of homogeneous item tokens or vocab tokens to facilitate stable and effective language modeling. In addition, a novel mutual regularization strategy is introduced to encourage CLLM4Rec to capture recommendation-related information from noisy user/item content. Finally, we propose a novel recommendation-oriented finetuning strategy for CLLM4Rec, where an item prediction head with multinomial likelihood is added to the pretrained CLLM4Rec backbone to predict hold-out items based on soft+hard prompts established from masked user-item interaction history, where recommendations of multiple items can be generated efficiently without hallucination.

Recommender systems (RSs) have become an indispensable part of online platforms. With the growing concerns of algorithmic fairness, RSs are not only expected to deliver high-quality personalized content, but are also demanded not to discriminate against users based on their demographic information. However, existing RSs could capture undesirable correlations between sensitive features and observed user behaviors, leading to biased recommendations. Most fair RSs tackle this problem by completely blocking the influences of sensitive features on recommendations. But since sensitive features may also affect user interests in a fair manner (e.g., race on culture-based preferences), indiscriminately eliminating all the influences of sensitive features inevitably degenerate the recommendations quality and necessary diversities. To address this challenge, we propose a path-specific fair RS (PSF-RS) for recommendations. Specifically, we summarize all fair and unfair correlations between sensitive features and observed ratings into two latent proxy mediators, where the concept of path-specific bias (PS-Bias) is defined based on path-specific counterfactual inference. Inspired by Pearl's minimal change principle, we address the PS-Bias by minimally transforming the biased factual world into a hypothetically fair world, where a fair RS model can be learned accordingly by solving a constrained optimization problem. For the technical part, we propose a feasible implementation of PSF-RS, i.e., PSF-VAE, with weakly-supervised variational inference, which robustly infers the latent mediators such that unfairness can be mitigated while necessary recommendation diversities can be maximally preserved simultaneously. Experiments conducted on semi-simulated and real-world datasets demonstrate the effectiveness of PSF-RS.

Online marketplace is a digital platform that connects buyers (demand) and sellers (supply) and provides exposure opportunities that individual participants would not otherwise have access to. Online marketplaces exist in a diverse set of domains and industries, for example, rideshare (Lyft, DiDi, Uber), house rental (Airbnb), real estate (Beke), online retail (Amazon, Ebay), job search (LinkedIn, Indeed.com, CareerBuilder), and food ordering and delivery (Doordash, Meituan). Besides academia, many companies and institutions are researching on topics specific to their particular domains. The fundamental mechanism of an online marketplace is to match supply and demand to generate transactions, with objectives considering service quality, participants experience, financial and operational efficiency. It is valuable to bring together researchers and practitioners from different application domains to discuss their experiences, challenges, and opportunities to leverage cross-domain knowledge. The goal of this workshop is to offer an opportunity to appreciate the diversity in applications, to draw connections to inform decision optimization across different industries, and to discover new problems that are fundamental to marketplaces of different domains. This workshop will follow a dual-track format. Track 1 covers the issues and algorithms pertinent to general online marketplaces as well as specific problems and applications arising from those diverse domains, such as ridesharing, online retail, food delivery, house rental, real estate, and more. Track 2 focuses on the state of the art advances in the computational jobs marketplace. Interesting challenges in this domain include the drastic increase of work from home or remote work, the imbalance between the demand and supply of the job market, the popularity of independent workers, the capability of helping job seekers on their whole job seeking journey and career development, the different objectives and behaviors of all major stakeholders in the ecosystem, e.g. job seekers, employers, recruiters and job agents.

The widespread usage of e-commerce websites in daily life and the resulting wealth of implicit feedback data form the foundation for systems that train and test e-commerce search ranking algorithms. While convenient to collect, implicit feedback data inherently suffers from various types of bias since user feedback is limited to products they are exposed to by existing search ranking algorithms and impacted by how the products are displayed. In the literature, a vast majority of existing methods have been proposed towards unbiased learning to rank for list-based web search scenarios. However, such methods cannot be directly adopted by e-commerce websites mainly for two reasons. First, in e-commerce websites, search engine results pages (SERPs) are displayed in 2-dimensional grids. The existing methods have not considered the difference in user behavior between list-based web search and grid-based product search. Second, there can be multiple types of implicit feedback (e.g., clicks and purchases) on e-commerce websites. We aim to utilize all types of implicit feedback as the supervision signals. In this work, we extend unbiased learning to rank to the world of e-commerce search via considering a grid-based product search scenario. We propose a novel framework which (1) forms the theoretical foundations to allow multiple types of implicit feedback in unbiased learning to rank and (2) incorporates the row skipping and slower decay click models to capture unique user behavior patterns in grid-based product search for inverse propensity scoring. Through extensive experiments on real-world e-commerce search log datasets across browsing devices and product taxonomies, we show that the proposed framework outperforms the state of the art unbiased learning to rank algorithms. These results also reveal important insights on how user behavior patterns vary in e-commerce SERPs across browsing devices and product taxonomies.

User engagement plays a central role in companies operating online services, such as search engines, news portals, e-commerce sites, entertainment services, and social networks. A main challenge is to leverage collected knowledge about the daily online behavior of millions of users to understand what engages them short-term and more importantly long-term. Two critical steps of improving user engagement are metrics and their optimization. The most common way that engagement is measured is through various online metrics, acting as proxy measures of user engagement. This tutorial will review these metrics, their advantages and drawbacks, and their appropriateness to various types of online services. Once metrics are defined, how to optimize them will become the key issue. We will survey methodologies including machine learning models and experimental designs that are utilized to optimize these metrics via direct or indirect ways. As case studies, we will focus on four types of services, news, search, entertainment, and e-commerce.

It is common in the internet industry to use offline-developed algorithms to power online products that contribute to the success of a business. Offline-developed algorithms are guided by offline evaluation metrics, which are often different from online business key performance indicators (KPIs). To maximize business KPIs, it is important to pick a north star among all available offline evaluation metrics. By noting that online products can be measured by online evaluation metrics, the online counterparts of offline evaluation metrics, we decompose the problem into two parts. As the offline A/B test literature works out the first part: counterfactual estimators of offline evaluation metrics that move the same way as their online counterparts, we focus on the second part: causal effects of online evaluation metrics on business KPIs. The north star of offline evaluation metrics should be the one whose online counterpart causes the most significant lift in the business KPI. We model the online evaluation metric as a mediator and formalize its causality with the business KPI as dose-response function (DRF). Our novel approach, causal meta-mediation analysis, leverages summary statistics of many existing randomized experiments to identify, estimate, and test the mediator DRF. It is easy to implement and to scale up, and has many advantages over the literature of mediation analysis and meta-analysis. We demonstrate its effectiveness by simulation and implementation on real data.

There is an increasing attention on next-item recommendation systems to infer the dynamic user preferences with sequential user interactions. While the semantics of an item can change over time and across users, the item correlations defined by user interactions in the short term can be distilled to capture such change, and help in uncovering the dynamic user preferences. Thus, we are motivated to develop a novel next-item recommendation framework empowered by sequential hypergraphs. Specifically, the framework: (i) adopts hypergraph to represent the short-term item correlations and applies multiple convolutional layers to capture multi-order connections in the hypergraph; (ii) models the connections between different time periods with a residual gating layer; and (iii) is equipped with a fusion layer to incorporate both the dynamic item embedding and short-term user intent to the representation of each interaction before feeding it into the self-attention layer for dynamic user modeling. Through experiments on datasets from the ecommerce sites Amazon and Etsy and the information sharing platform Goodreads, the proposed model can significantly outperform the state-of-the-art in predicting the next interesting item for each user.

Currently, most sequence-based recommendation models aim to predict a user's next actions (e.g. next purchase) based on their past actions. These models either capture users' intrinsic preference (e.g. a comedy lover, or a fan of fantasy) from their long-term behavior patterns or infer their current needs by emphasizing recent actions. However, in e-commerce, intrinsic user behavior may be shifted by occasions such as birthdays, anniversaries, or gifting celebrations (Valentine's Day or Mother's Day), leading to purchases that deviate from long-term preferences and are not related to recent actions. In this work, we propose a novel next-item recommendation system which models a user's default, intrinsic preference, as well as two different kinds of occasion-based signals that may cause users to deviate from their normal behavior. More specifically, this model is novel in that it: (1) captures a personal occasion signal using an attention layer that models reoccurring occasions specific to that user (e.g. a birthday); (2) captures a global occasion signal using an attention layer that models seasonal or reoccurring occasions for many users (e.g. Christmas); (3) balances the user's intrinsic preferences with the personal and global occasion signals for different users at different timestamps with a gating layer. We explore two real-world e-commerce datasets (Amazon and Etsy) and show that the proposed model outperforms state-of-the-art models by 7.62% and 6.06% in predicting users' next purchase.

Users exhibit different intents across e-commerce services (e.g. discovering items, purchasing gifts, etc.) which drives them to interact with a wide variety of items in multiple ways (e.g. click, add-to-cart, add-to-favorites, purchase). To give better recommendations, it is important to capture user intent, in addition to considering their historic interactions. However these intents are by definition latent, as we observe only a user’s interactions, and not their underlying intent. To discover such latent intents, and use them effectively for recommendation, in this paper we propose an Attentive Sequential model of Latent Intent (ASLI in short). Our model first learns item similarities from users’ interaction histories via a self-attention layer, then uses a Temporal Convolutional Network layer to obtain a latent representation of the user’s intent from her actions on a particular category. We use this representation to guide an attentive model to predict the next item. Results from our experiments show that our model can capture the dynamics of user behavior and preferences, leading to state-of-the-art performance across datasets from two major e-commerce platforms, namely Etsy and Alibaba.

Aesthetic style is the crux of many purchasing decisions. When considering an item for purchase, buyers need to be aligned not only with the functional aspects (e.g. description, category, ratings) of an item's specification, but also its stylistic and aesthetic aspects (e.g. modern, classical, retro) as well. Style becomes increasingly important on e-commerce sites like Etsy, an online marketplace for handmade and vintage goods, where hundreds of thousands of items can differ by style and aesthetic alone. As such, it is important for industry recommender systems to properly model style when understanding shoppers' buying preference. In past work, because of its abstract nature, style is often approached in an unsupervised manner, represented by nameless latent factors or embeddings. As a result, there has been no previous work on predictive models nor analysis devoted to understanding how style, or even the presence of style, impacts a buyer's purchase decision. In this paper, we discuss a novel process by which we leverage 43 named styles given by merchandising experts in order to bootstrap large-scale style prediction and analysis of how style impacts purchase decision. We train a supervised, style-aware deep neural network that is shown to predict item style with high accuracy, while generating style-aware embeddings that can be used in downstream recommendation tasks. We share in our analysis, based on over a year's worth of transaction data and show that these findings are crucial to understanding how to more explicitly leverage style signal in industry-scale recommender systems.

E-commerce companies have a number of online products, such as organic search, sponsored search, and recommendation modules, to fulfill customer needs. Although each of these products provides a unique opportunity for users to interact with a portion of the overall inventory, they are all similar channels for users and compete for limited time and monetary budgets of users. To optimize users' overall experiences on an E-commerce platform, instead of understanding and improving different products separately, it is important to gain insights into the evidence that a change in one product would induce users to change their behaviors in others, which may be due to the fact that these products are functionally similar. In this paper, we introduce causal mediation analysis as a formal statistical tool to reveal the underlying causal mechanisms. Existing literature provides little guidance on cases where multiple unmeasured causally-dependent mediators exist, which are common in A/B tests. We seek a novel approach to identify in those scenarios direct and indirect effects of the treatment. In the end, we demonstrate the effectiveness of the proposed method in data from Etsy's real A/B tests and shed lights on complex relationships between different products.

Matching between two information objects is the core of many different information retrieval (IR) applications including Web search, question answering, and recommendation. Recently, deep learning methods have yielded immense success in speech recognition, computer vision, and natural language processing, significantly advancing state-of-the-art of these areas. In the IR community, deep learning has also attracted much attention, and researchers have proposed a large number of deep matching models to tackle the matching problem for different IR applications. Despite the fact that deep matching models have gained significant progress in these areas, there are still many challenges to be addressed when applying these models to real IR scenarios. In this workshop, we focus on the applicability of deep matching models to practical applications. We aim to discuss the issues of applying deep matching models to production systems, as well as to shed some light on the fundamental characteristics of different matching tasks in IR. website : https://wsdm2019-dapa.github.io/index.html

Online A/B tests play an instrumental role for Internet companies to improve products and technologies in a data-driven manner. An online A/B test, in its most straightforward form, can be treated as a static hypothesis test where traditional statistical tools such as p-values and power analysis might be applied to help decision makers determine which variant performs better. However, a static A/B test presents both time cost and the opportunity cost for rapid product iterations. For time cost, a fast-paced product evolution pushes its shareholders to consistently monitor results from online A/B experiments, which usually invites peeking and altering experimental designs as data collected. It is recognized that this flexibility might harm statistical guarantees if not introduced in the right way, especially when online tests are considered as static hypothesis tests. For opportunity cost, a static test usually entails a static allocation of users into different variants, which prevents an immediate roll-out of the better version to larger audience or risks of alienating users who may suffer from a bad experience. While some works try to tackle these challenges, no prior method focuses on a holistic solution to both issues. In this paper, we propose a unified framework utilizing sequential analysis and multi-armed bandit to address time cost and the opportunity cost of static online tests simultaneously. In particular, we present an imputed sequential Girshick test that accommodates online data and dynamic allocation of data. The unobserved potential outcomes are treated as missing data and are imputed using empirical averages. Focusing on the binomial model, we demonstrate that the proposed imputed Girshick test achieves Type-I error and power control with both a fixed allocation ratio and an adaptive allocation such as Thompson Sampling through extensive experiments. In addition, we also run experiments on historical Etsy.com A/B tests to show the reduction in opportunity cost when using the proposed method.

Building price- and budget-aware recommender systems is critical in settings where one wishes to produce recommendations that balance users' preferences (what they like) with a model of purchase likelihood (what they will buy). A trivial solution consists of learning global budget terms for each user based on their past expenditure. To more accurately model user budgets, we also consider a user's within-session budget, which may deviate from their global budget depending on their shopping context. In this paper, we find that users implicitly reveal their session-specific budgets through the sequence of items they browse within that session. Specifically, we find that some users "browse down," by purchasing the cheapest item among alternatives under consideration, others "browse up" (selecting the most expensive), and others ultimately purchase items around the middle. Surprisingly, this mixture of behaviors is difficult to observe globally, as individual users tend to belong firmly to one of the three segments. To model this behavior, we develop an interpretable budget model that combines a clustering component to detect different user segments, with a model of segment-specific purchase profiles. We apply our model on a dataset of browsing and purchasing sessions from Etsy, a large e-commerce website focused on handmade and vintage goods, where it outperforms strong baselines and existing production systems.

In recent years, product search engines have emerged as a key factor for online businesses. According to a recent survey, over 55% of online customers begin their online shopping journey by searching on an E-Commerce (EC) website like Amazon as opposed to a generic web search engine like Google. Information retrieval research to date has been focused on optimizing search ranking algorithms for web documents while little attention has been paid to product search. There are several intrinsic differences between web search and product search that make the direct application of traditional search ranking algorithms to EC search platforms difficult. First, the success of web and product search is measured differently; one seeks to optimize for relevance while the other must optimize for both relevance and revenue. Second, when using real-world EC transaction data, there is no access to manually annotated labels. In this paper, we address these differences with a novel learning framework for EC product search called LETORIF (LEarning TO Rank with Implicit Feedback). In this framework, we utilize implicit user feedback signals (such as user clicks and purchases) and jointly model the different stages of the shopping journey to optimize for EC sales revenue. We conduct experiments on real-world EC transaction data and introduce a a new evaluation metric to estimate expected revenue after re-ranking. Experimental results show that LETORIF outperforms top competitors in improving purchase rates and total revenue earned.

User engagement plays a central role in companies operating online services, such as search engines, news portals, e-commerce sites, and social networks. A main challenge is to leverage collected knowledge about the daily online behavior of millions of users to understand what engage them short-term and more importantly long-term. The most common way that engagement is measured is through various online metrics, acting as proxy measures of user engagement. This tutorial will review these metrics, their advantages and drawbacks, and their appropriateness to various types of online services. As case studies, we will focus on three types of services, news, search and e-commerce. We will also briefly discuss how to develop better machine learning models to optimize online metrics, and design experiments to test these models.

In this work, we propose to improve long-term user engagement in a recommender system from the perspective of sequential decision optimization, where users' click and return behaviors are directly modeled for online optimization. A bandit-based solution is formulated to balance three competing factors during online learning, including exploitation for immediate click, exploitation for expected future clicks, and exploration of unknowns for model estimation. We rigorously prove that with a high probability our proposed solution achieves a sublinear upper regret bound in maximizing cumulative clicks from a population of users in a given period of time, while a linear regret is inevitable if a user's temporal return behavior is not considered when making the recommendations. Extensive experimentation on both simulations and a large-scale real-world dataset collected from Yahoo frontpage news recommendation log verified the effectiveness and significant improvement of our proposed algorithm compared with several state-of-the-art online learning baselines for recommendation.

Recent advances in neural networks have inspired people to design hybrid recommendation algorithms that can incorporate both (1) user-item interaction information and (2) content information including image, audio, and text. Despite their promising results, neural network-based recommendation algorithms pose extensive computational costs, making it challenging to scale and improve upon. In this paper, we propose a general neural network-based recommendation framework, which subsumes several existing state-of-the-art recommendation algorithms, and address the efficiency issue by investigating sampling strategies in the stochastic gradient descent training for the framework. We tackle this issue by first establishing a connection between the loss functions and the user-item interaction bipartite graph, where the loss function terms are defined on links while major computation burdens are located at nodes. We call this type of loss functions "graph-based" loss functions, for which varied mini-batch sampling strategies can have different computational costs. Based on the insight, three novel sampling strategies are proposed, which can significantly improve the training efficiency of the proposed framework (up to $\times 30$ times speedup in our experiments), as well as improving the recommendation performance. Theoretical analysis is also provided for both the computational cost and the convergence. We believe the study of sampling strategies have further implications on general graph-based loss functions, and would also enable more research under the neural network-based recommendation framework.

Recommending personalized content to users is a long-standing challenge to many online services including Facebook, Yahoo, Linkedin and Twitter. Traditional recommendation models such as latent factor models and feature-based models are usually trained for all users and optimize an "average" experience for them, yielding sub-optimal solutions. Although multi-task learning provides an opportunity to learn personalized models per user, learning algorithms are usually tailored to specific models (e.g., generalized linear model, matrix factorization and etc.), creating obstacles for a unified engineering interface, which is important for large Internet companies. In this paper, we present an empirical framework to learn user-specific personal models for content recommendation by utilizing gradient information from a global model. Our proposed method can potentially benefit any model that can be optimized through gradients, offering a lightweight yet generic alternative to conventional multi-task learning algorithms for user personalization. We demonstrate the effectiveness of the proposed framework by incorporating it in three popular machine learning algorithms including logistic regression, gradient boosting decision tree and matrix factorization. Our extensive empirical evaluation shows that the proposed framework can significantly improve the efficiency of personalized recommendation in real-world datasets.

Latent factor models and decision tree based models are widely used in tasks of prediction, ranking and recommendation. Latent factor models have the advantage of interpreting categorical features by a low-dimensional representation, while such an interpretation does not naturally fit numerical features. In contrast, decision tree based models enjoy the advantage of capturing the nonlinear interactions of numerical features, while their capability of handling categorical features is limited by the cardinality of those features. Since in real-world applications we usually have both abundant numerical features and categorical features with large cardinality (e.g. geolocations, IDs, tags etc.), we design a new model, called GB-CENT, which leverages latent factor embedding and tree components to achieve the merits of both while avoiding their demerits. With two real-world data sets, we demonstrate that GB-CENT can effectively (i.e. fast and accurately) achieve better accuracy than state-of-the-art matrix factorization, decision tree based models and their ensemble.

Feature-based collaborative filtering models, such as state-of-the-art factorization machines and regression-based latent factor models, rarely consider features' structural information, ignoring the heterogeneity of inter-type and intra-type relationships. Naïvely treating all feature pairs equally would potentially deteriorate the overall recommendation performance. In addition, human prior knowledge and other hierarchical or graphical structures are often available for some features, e.g., the country-state-city hierarchy for geographic features and the topical taxonomy for article features. It is a challenge to utilize the prior knowledge to further boost performance of state-of-the-art models. In this paper we employ rich features from both user and item sides to enhance latent factors learnt from interaction data, uncovering hidden structures from features' relationships and learning sparse pairwise and tree structural connections among features. Our framework borrows the modeling strengh from both structural sparsity modeling and latent factor models. Experiments on a real-world large-scale recommendation data set demonstrated that the proposed model outperforms several strong state-of-the-art baselines.

Many internet companies, such as Yahoo, Facebook, Google and Twitter, rely on content recommendation systems to deliver the most relevant content items to individual users through personalization. Delivering such personalized user experiences is believed to increase the long term engagement of users. While there has been a lot of progress in designing effective personalized recommender systems, by exploiting user interests and historical interaction data through implicit (item click) or explicit (item rating) feedback, directly optimizing for users' satisfaction with the system remains challenging. In this paper, we explore the idea of using item-level dwell time as a proxy to quantify how likely a content item is relevant to a particular user. We describe a novel method to compute accurate dwell time based on client-side and server-side logging and demonstrate how to normalize dwell time across different devices and contexts. In addition, we describe our experiments in incorporating dwell time into state-of-the-art learning to rank techniques and collaborative filtering models that obtain competitive performances in both offline and online settings.

We propose a novel probabilistic topic model that jointly models authors, documents, cited authors, and venues simultaneously in one integrated framework, as compared to previous work which embeds fewer components. This model is designed for three typical applications in academic network analysis: the problems of expert ranking, cited author prediction and venue prediction. Experiments based on two real world data sets demonstrate the model to be effective, and it outperforms several state-of-the-art algorithms in all three applications.

Online user engagement optimization is key to many Internet business. Several research areas are related to the concept of online user engagement optimization, including machine learning, data mining, information retrieval, recommender systems, online A/B (bucket) testing and psychology. In the past, research efforts in this direction are pursued in separate communities and conferences, yielding potential disconnected and repeated results. In addition, researchers and practitioners are sometimes only exposed to a specific aspect of the topic, which might be incomplete and suboptimal to the whole picture. Here, we organize the first workshop on the topic of online user engagement optimization, explicitly targeting the topic as a whole and bring researchers and practitioners together to foster the field. We invite two leading researchers from industry to give keynote talks about online machine learning and online experimentations. In addition, several invited talks from industry and academic researchers have covered the topics of content personalization, online experimental platforms and recommender systems. Also, six novel submissions are included as short papers in the workshop such that new results are discussed and shared among the workshop.

Users of popular services like Twitter and Facebook are often simultaneously overwhelmed with the amount of information delivered via their social connections and miss out on much content that they might have liked to see, even though it was distributed outside of their social circle. Both issues serve as difficulties to the users and drawbacks to the services.

With the availability of cheap location sensors, geotagging of messages in online social networks is proliferating. For instance, Twitter, Facebook, Foursquare, and Google+ provide these services both explicitly by letting users choose their location or implicitly via a sensor. This paper presents an integrated generative model of location and message content. That is, we provide a model for combining distributions over locations, topics, and over user characteristics, both in terms of location and in terms of their content preferences. Unlike previous work which modeled data in a flat pre-defined representation, our model automatically infers both the hierarchical structure over content and over the size and position of geographical locations. This affords significantly higher accuracy --- location uncertainty is reduced by 40% relative to the best previous results [21] achieved on location estimation from Tweets.

As online social media further integrates deeper into our lives, we spend more time consuming social update streams that come from our online connections. Although social update streams provide a tremendous opportunity for us to access information on-the-fly, we often complain about its relevance. Some of us are flooded with a steady stream of information and simply cannot process it in full. Ranking the incoming content becomes the only solution for the overwhelmed users. For some others, in contrast, the incoming information stream is pretty weak, and they have to actively search for relevant information which is quite tedious. For these users, augmenting their incoming content flow with relevant information from outside their first-degree network would be a viable solution. In that case, the problem of relevance becomes even more prominent. In this paper, we start an open discussion on how to build effective systems for ranking social updates from a unique perspective of LinkedIn -- the largest professional network in the world. More specifically, we address this problem as an intersection of learning to rank, collaborative filtering, and clickthrough modeling, while leveraging ideas from information retrieval and recommender systems. We propose a novel probabilistic latent factor model with regressions on explicit features and compare it with a number of non-trivial baselines. In addition to demonstrating superior performance of our model, we shed some light on the nature of social updates on LinkedIn and how users interact with them, which might be applicable to social update streams in general.

Micro-blogging services have become indispensable communication tools for online users for disseminating breaking news, eyewitness accounts, individual expression, and protest groups. Recently, Twitter, along with other online social networking services such as Foursquare, Gowalla, Facebook and Yelp, have started supporting location services in their messages, either explicitly, by letting users choose their places, or implicitly, by enabling geo-tagging, which is to associate messages with latitudes and longitudes. This functionality allows researchers to address an exciting set of questions: 1) How is information created and shared across geographical locations, 2) How do spatial and linguistic characteristics of people vary across regions, and 3) How to model human mobility. Although many attempts have been made for tackling these problems, previous methods are either complicated to be implemented or oversimplified that cannot yield reasonable performance. It is a challenge task to discover topics and identify users' interests from these geo-tagged messages due to the sheer amount of data and diversity of language variations used on these location sharing services. In this paper we focus on Twitter and present an algorithm by modeling diversity in tweets based on topical diversity, geographical diversity, and an interest distribution of the user. Furthermore, we take the Markovian nature of a user's location into account. Our model exploits sparse factorial coding of the attributes, thus allowing us to deal with a large and diverse set of covariates efficiently. Our approach is vital for applications such as user profiling, content recommendation and topic tracking. We show high accuracy in location estimation based on our model. Moreover, the algorithm identifies interesting topics based on location and language.

With hundreds of millions of participants, social media services have become commonplace. Unlike a traditional social network service, a microblogging network like Twitter is a hybrid network, combining aspects of both social networks and information networks. Understanding the structure of such hybrid networks and predicting new links are important for many tasks such as friend recommendation, community detection, and modeling network growth. We note that the link prediction problem in a hybrid network is different from previously studied networks. Unlike the information networks and traditional online social networks, the structures in a hybrid network are more complicated and informative. We compare most popular and recent methods and principles for link prediction and recommendation. Finally we propose a novel structure-based personalized link prediction model and compare its predictive performance against many fundamental and popular link prediction methods on real-world data from the Twitter microblogging network. Our experiments on both static and dynamic data sets show that our methods noticeably outperform the state-of-the-art.

In recent years social media have become indispensable tools for information dissemination, operating in tandem with traditional media outlets such as newspapers, and it has become critical to understand the interaction between the new and old sources of news. Although social media as well as traditional media have attracted attention from several research communities, most of the prior work has been limited to a single medium. In addition temporal analysis of these sources can provide an understanding of how information spreads and evolves. Modeling temporal dynamics while considering multiple sources is a challenging research problem. In this paper we address the problem of modeling text streams from two news sources - Twitter and Yahoo! News. Our analysis addresses both their individual properties (including temporal dynamics) and their inter-relationships. This work extends standard topic models by allowing each text stream to have both local topics and shared topics. For temporal modeling we associate each topic with a time-dependent function that characterizes its popularity over time. By integrating the two models, we effectively model the temporal dynamics of multiple correlated text streams in a unified framework. We evaluate our model on a large-scale dataset, consisting of text streams from both Twitter and news feeds from Yahoo! News. Besides overcoming the limitations of existing models, we show that our work achieves better perplexity on unseen data and identifies more coherent topics. We also provide analysis of finding real-world events from the topics obtained by our model.

Text corpora with documents from a range of time epochs are natural and ubiquitous in many fields, such as research papers, newspaper articles and a variety of types of recently emerged social media. People not only would like to know what kind of topics can be found from these data sources but also wish to understand the temporal dynamics of these topics and predict certain properties of terms or documents in the future. Topic models are usually utilized to find latent topics from text collections, and recently have been applied to temporal text corpora. However, most proposed models are general purpose models to which no real tasks are explicitly associated. Therefore, current models may be difficult to apply in real-world applications, such as the problems of tracking trends and predicting popularity of keywords. In this paper, we introduce a real-world task, tracking trends of terms, to which temporal topic models can be applied. Rather than building a general-purpose model, we propose a new type of topic model that incorporates the volume of terms into the temporal dynamics of topics and optimizes estimates of term volumes. In existing models, trends are either latent variables or not considered at all which limits the potential for practical use of trend information. In contrast, we combine state-space models with term volumes with a supervised learning model, enabling us to effectively predict the volume in the future, even without new documents. In addition, it is straightforward to obtain the volume of latent topics as a by-product of our model, demonstrating the superiority of utilizing temporal topic models over traditional time-series tools (e.g., autoregressive models) to tackle this kind of problem. The proposed model can be further extended with arbitrary word-level features which are evolving over time. We present the results of applying the model to two datasets with long time periods and show its effectiveness over non-trivial baselines.

Unlike a traditional social network service, a microblogging network like Twitter is a hybrid network, combining aspects of both social networks and information networks. Understanding the structure of such hybrid networks and to predict new links are important for many tasks such as friend recommendation, community detection, and network growth models. In this paper, by analyzing data collected over time, we find that 90% of new links are to people just two hops away and dynamics of friend acquisition are also related to users' account age. Finally, we compare two popular sampling methods which are widely used for network analysis and find that ForestFire does not preserve properties required for the link prediction task.

Social tagging systems have become increasingly popular for sharing and organizing web resources. Tag prediction is a common feature of social tagging systems. Social tagging by nature is an incremental process, meaning that once a user has saved a web page with tags, the tagging system can provide more accurate predictions for the user, based on user's incremental behaviors. However, existing tag prediction methods do not consider this important factor, in which their training and test datasets are either split by a fixed time stamp or randomly sampled from a larger corpus. In our temporal experiments, we perform a time-sensitive sampling on an existing public dataset, resulting in a new scenario which is much closer to "real-world".

Discussion boards and online forums are important platforms for people to share information. Users post questions or problems onto discussion boards and rely on others to provide possible solutions and such question-related content sometimes even dominates the whole discussion board. However, to retrieve this kind of information automatically and effectively is still a non-trivial task. In addition, the existence of other types of information (e.g., announcements, plans, elaborations, etc.) makes it difficult to assume that every thread in a discussion board is about a question. We consider the problems of identifying question-related threads and their potential answers as classification tasks. Experimental results across multiple datasets demonstrate that our method can significantly improve the performance in both question detection and answer finding subtasks. We also do a careful comparison of how different types of features contribute to the final result and show that non-content features play a key role in improving overall performance. Finally, we show that a ranking scheme based on our classification approach can yield much better performance than prior published methods.