Search

E-commerce search and recommendation usually operate on structured data such as product catalogs and taxonomies. However, creating better search and recommendation systems often requires a large variety of unstructured data including customer reviews and articles on the web. Traditionally, the solution has always been converting unstructured data into structured data through information extraction, and conducting search over the structured data. However, this is a costly approach that often has low quality. In this paper, we envision a solution that does entirely the opposite. Instead of converting unstructured data (web pages, customer reviews, etc) to structured data, we instead convert structured data (product inventory, catalogs, taxonomies, etc) into textual data, which can be easily integrated into the text corpus that trains LLMs. Then, search and recommendation can be performed through a Q/A mechanism through an LLM instead of using traditional information retrieval methods over structured data.

Efficient and accurate product relevance assessment is critical for user experiences and business success. Training a proficient relevance assessment model requires high-quality query-product pairs, often obtained through negative sampling strategies. Unfortunately, current methods introduce pooling bias by mistakenly sampling false negatives, diminishing performance and business impact. To address this, we present Bias-mitigating Hard Negative Sampling (BHNS), a novel negative sampling strategy tailored to identify and adjust for false negatives, building upon our original False Negative Estimation algorithm. Our experiments in the Instacart search setting confirm BHNS as effective for practical e-commerce use. Furthermore, comparative analyses on public dataset showcase its domain-agnostic potential for diverse applications., Comment: Submitted to WWW'24 Industry Track

The key to e-commerce search is how to best utilize the large yet noisy log data. In this paper, we present our embedding-based model for grocery search at Instacart. The system learns query and product representations with a two-tower transformer-based encoder architecture. To tackle the cold-start problem, we focus on content-based features. To train the model efficiently on noisy data, we propose a self-adversarial learning method and a cascade training method. AccOn an offline human evaluation dataset, we achieve 10% relative improvement in RECALL@20, and for online A/B testing, we achieve 4.1% cart-adds per search (CAPS) and 1.5% gross merchandise value (GMV) improvement. We describe how we train and deploy the embedding based search model and give a detailed analysis of the effectiveness of our method., Comment: Accepted by SIGIR eCom, July 15, 2022

With the prevalence of deep learning based embedding approaches, recommender systems have become a proven and indispensable tool in various information filtering applications. However, many of them remain difficult to diagnose what aspects of the deep models' input drive the final ranking decision, thus, they cannot often be understood by human stakeholders. In this paper, we investigate the dilemma between recommendation and explainability, and show that by utilizing the contextual features (e.g., item reviews from users), we can design a series of explainable recommender systems without sacrificing their performance. In particular, we propose three types of explainable recommendation strategies with gradual change of model transparency: whitebox, graybox, and blackbox. Each strategy explains its ranking decisions via different mechanisms: attention weights, adversarial perturbations, and counterfactual perturbations. We apply these explainable models on five real-world data sets under the contextualized setting where users and items have explicit interactions. The empirical results show that our model achieves highly competitive ranking performance, and generates accurate and effective explanations in terms of numerous quantitative metrics and qualitative visualizations.

Recently, there is an effort to extend fine-grained entity typing by using a richer and ultra-fine set of types, and labeling noun phrases including pronouns and nominal nouns instead of just named entity mentions. A key challenge for this ultra-fine entity typing task is that human annotated data are extremely scarce, and the annotation ability of existing distant or weak supervision approaches is very limited. To remedy this problem, in this paper, we propose to obtain training data for ultra-fine entity typing by using a BERT Masked Language Model (MLM). Given a mention in a sentence, our approach constructs an input for the BERT MLM so that it predicts context dependent hypernyms of the mention, which can be used as type labels. Experimental results demonstrate that, with the help of these automatically generated labels, the performance of an ultra-fine entity typing model can be improved substantially. We also show that our approach can be applied to improve traditional fine-grained entity typing after performing simple type mapping., Comment: ACL 2021

Federated learning learns a neural network model by aggregating the knowledge from a group of distributed clients under the privacy-preserving constraint. In this work, we show that this paradigm might inherit the adversarial vulnerability of the centralized neural network, i.e., it has deteriorated performance on adversarial examples when the model is deployed. This is even more alarming when federated learning paradigm is designed to approximate the updating behavior of a centralized neural network. To solve this problem, we propose an adversarially robust federated learning framework, named Fed_BVA, with improved server and client update mechanisms. This is motivated by our observation that the generalization error in federated learning can be naturally decomposed into the bias and variance triggered by multiple clients' predictions. Thus, we propose to generate the adversarial examples via maximizing the bias and variance during server update, and learn the adversarially robust model updates with those examples during client update. As a result, an adversarially robust neural network can be aggregated from these improved local clients' model updates. The experiments are conducted on multiple benchmark data sets using several prevalent neural network models, and the empirical results show that our framework is robust against white-box and black-box adversarial corruptions under both IID and non-IID settings., Comment: Accepted by CIKM-2022

A natural language interface (NLI) to databases is an interface that translates a natural language question to a structured query that is executable by database management systems (DBMS). However, an NLI that is trained in the general domain is hard to apply in the spatial domain due to the idiosyncrasy and expressiveness of the spatial questions. Inspired by the machine comprehension model, we propose a spatial comprehension model that is able to recognize the meaning of spatial entities based on the semantics of the context. The spatial semantics learned from the spatial comprehension model is then injected to the natural language question to ease the burden of capturing the spatial-specific semantics. With our spatial comprehension model and information injection, our NLI for the spatial domain, named SpatialNLI, is able to capture the semantic structure of the question and translate it to the corresponding syntax of an executable query accurately. We also experimentally ascertain that SpatialNLI outperforms state-of-the-art methods., Comment: 10 pages

Question answering (QA) has become a popular way for humans to access billion-scale knowledge bases. Unlike web search, QA over a knowledge base gives out accurate and concise results, provided that natural language questions can be understood and mapped precisely to structured queries over the knowledge base. The challenge, however, is that a human can ask one question in many different ways. Previous approaches have natural limits due to their representations: rule based approaches only understand a small set of "canned" questions, while keyword based or synonym based approaches cannot fully understand the questions. In this paper, we design a new kind of question representation: templates, over a billion scale knowledge base and a million scale QA corpora. For example, for questions about a city's population, we learn templates such as What's the population of $city?, How many people are there in $city?. We learned 27 million templates for 2782 intents. Based on these templates, our QA system KBQA effectively supports binary factoid questions, as well as complex questions which are composed of a series of binary factoid questions. Furthermore, we expand predicates in RDF knowledge base, which boosts the coverage of knowledge base by 57 times. Our QA system beats all other state-of-art works on both effectiveness and efficiency over QALD benchmarks.

Relational database management systems (RDBMSs) are powerful because they are able to optimize and answer queries against any relational database. A natural language interface (NLI) for a database, on the other hand, is tailored to support that specific database. In this work, we introduce a general purpose transfer-learnable NLI with the goal of learning one model that can be used as NLI for any relational database. We adopt the data management principle of separating data and its schema, but with the additional support for the idiosyncrasy and complexity of natural languages. Specifically, we introduce an automatic annotation mechanism that separates the schema and the data, where the schema also covers knowledge about natural language. Furthermore, we propose a customized sequence model that translates annotated natural language queries to SQL statements. We show in experiments that our approach outperforms previous NLI methods on the WikiSQL dataset and the model we learned can be applied to another benchmark dataset OVERNIGHT without retraining.

Verbs are important in semantic understanding of natural language. Traditional verb representations, such as FrameNet, PropBank, VerbNet, focus on verbs' roles. These roles are too coarse to represent verbs' semantics. In this paper, we introduce verb patterns to represent verbs' semantics, such that each pattern corresponds to a single semantic of the verb. First we analyze the principles for verb patterns: generality and specificity. Then we propose a nonparametric model based on description length. Experimental results prove the high effectiveness of verb patterns. We further apply verb patterns to context-aware conceptualization, to show that verb patterns are helpful in semantic-related tasks., Comment: 7 pages, 3 figures, camera-ready version published on AAAI 2016

Entity suggestion by example (ESbE) refers to a type of entity acquisition query in which a user provides a set of example entities as the query and obtains in return some entities that best complete the concept underlying the given query. Such entity acquisition queries can be useful in many applications such as related-entity recommendation and query expansion. A number of ESbE query processing solutions exist in the literature. However, they mostly build only on the idea of entity co-occurrences either in text or web lists, without taking advantage of the existence of many web-scale conceptual taxonomies that consist of hierarchical isA relationships between entity-concept pairs. This paper provides a query processing method based on the relevance models between entity sets and concepts. These relevance models can be used to obtain the fine-grained concepts implied by the query entity set, and the entities that belong to a given concept, thereby providing the entity suggestions. Extensive evaluations with real data sets show that the accuracy of the queries processed with this new method is significantly higher than that of existing solutions.

Online encyclopedia such as Wikipedia has become one of the best sources of knowledge. Much effort has been devoted to expanding and enriching the structured data by automatic information extraction from unstructured text in Wikipedia. Although remarkable progresses have been made, their effectiveness and efficiency is still limited as they try to tackle an extremely difficult natural language understanding problems and heavily relies on supervised learning approaches which require large amount effort to label the training data. In this paper, instead of performing information extraction over unstructured natural language text directly, we focus on a rich set of semi-structured data in Wikipedia articles: linked entities. The idea of this paper is the following: If we can summarize the relationship between the entity and its linked entities, we immediately harvest some of the most important information about the entity. To this end, we propose a novel rank aggregation approach to remove noise, an effective clustering and labeling algorithm to extract knowledge.

Computing shortest distances is one of the fundamental problems on graphs, and remains a {\em challenging} task today. {\em Distance} {\em landmarks} have been recently studied for shortest distance queries with an auxiliary data structure, referred to as {\em landmark} {\em covers}. This paper studies how to apply distance landmarks for fast {\em exact} shortest distance query answering on large road graphs. However, the {\em direct} application of distance landmarks is {\em impractical} due to the high space and time cost. To rectify this problem, we investigate novel techniques that can be seamlessly combined with distance landmarks. We first propose a notion of {\em hybrid landmark covers}, a revision of landmark covers. Second, we propose a notion of {\em agents}, each of which represents a small subgraph and holds good properties for fast distance query answering. We also show that agents can be computed in {\em linear time}. Third, we introduce graph partitions to deal with the remaining subgraph that cannot be captured by agents. Fourth, we develop a unified framework that seamlessly integrates our proposed techniques and existing optimization techniques, for fast shortest distance query answering. Finally, we experimentally verify that our techniques significantly improve the efficiency of shortest distance queries, using real-life road graphs.

Shortest path computation is one of the most fundamental operations for managing and analyzing large social networks. Though existing techniques are quite effective for finding the shortest path on large but sparse road networks, social graphs have quite different characteristics: they are generally non-spatial, non-weighted, scale-free, and they exhibit small-world properties in addition to their massive size. In particular, the existence of hubs, those vertices with a large number of connections, explodes the search space, making the shortest path computation surprisingly challenging. In this paper, we introduce a set of novel techniques centered around hubs, collectively referred to as the Hub-Accelerator framework, to compute the k-degree shortest path (finding the shortest path between two vertices if their distance is within k). These techniques enable us to significantly reduce the search space by either greatly limiting the expansion scope of hubs (using the novel distance- preserving Hub-Network concept) or completely pruning away the hubs in the online search (using the Hub2-Labeling approach). The Hub-Accelerator approaches are more than two orders of magnitude faster than BFS and the state-of-the-art approximate shortest path method Sketch for the shortest path computation. The Hub- Network approach does not introduce additional index cost with light pre-computation cost; the index size and index construction cost of Hub2-Labeling are also moderate and better than or comparable to the approximation indexing Sketch method., Comment: 12 pages

We study the problem of answering k-hop reachability queries in a directed graph, i.e., whether there exists a directed path of length k, from a source query vertex to a target query vertex in the input graph. The problem of k-hop reachability is a general problem of the classic reachability (where k=infinity). Existing indexes for processing classic reachability queries, as well as for processing shortest path queries, are not applicable or not efficient for processing k-hop reachability queries. We propose an index for processing k-hop reachability queries, which is simple in design and efficient to construct. Our experimental results on a wide range of real datasets show that our index is more efficient than the state-of-the-art indexes even for processing classic reachability queries, for which these indexes are primarily designed. We also show that our index is efficient in answering k-hop reachability queries., Comment: VLDB2012

Many studies have been conducted on seeking the efficient solution for subgraph similarity search over certain (deterministic) graphs due to its wide application in many fields, including bioinformatics, social network analysis, and Resource Description Framework (RDF) data management. All these works assume that the underlying data are certain. However, in reality, graphs are often noisy and uncertain due to various factors, such as errors in data extraction, inconsistencies in data integration, and privacy preserving purposes. Therefore, in this paper, we study subgraph similarity search on large probabilistic graph databases. Different from previous works assuming that edges in an uncertain graph are independent of each other, we study the uncertain graphs where edges' occurrences are correlated. We formally prove that subgraph similarity search over probabilistic graphs is #P-complete, thus, we employ a filter-and-verify framework to speed up the search. In the filtering phase,we develop tight lower and upper bounds of subgraph similarity probability based on a probabilistic matrix index, PMI. PMI is composed of discriminative subgraph features associated with tight lower and upper bounds of subgraph isomorphism probability. Based on PMI, we can sort out a large number of probabilistic graphs and maximize the pruning capability. During the verification phase, we develop an efficient sampling algorithm to validate the remaining candidates. The efficiency of our proposed solutions has been verified through extensive experiments., Comment: VLDB2012

The ability to handle large scale graph data is crucial to an increasing number of applications. Much work has been dedicated to supporting basic graph operations such as subgraph matching, reachability, regular expression matching, etc. In many cases, graph indices are employed to speed up query processing. Typically, most indices require either super-linear indexing time or super-linear indexing space. Unfortunately, for very large graphs, super-linear approaches are almost always infeasible. In this paper, we study the problem of subgraph matching on billion-node graphs. We present a novel algorithm that supports efficient subgraph matching for graphs deployed on a distributed memory store. Instead of relying on super-linear indices, we use efficient graph exploration and massive parallel computing for query processing. Our experimental results demonstrate the feasibility of performing subgraph matching on web-scale graph data., Comment: VLDB2012

This paper describes the LDL++ system and the research advances that have enabled its design and development. We begin by discussing the new nonmonotonic and nondeterministic constructs that extend the functionality of the LDL++ language, while preserving its model-theoretic and fixpoint semantics. Then, we describe the execution model and the open architecture designed to support these new constructs and to facilitate the integration with existing DBMSs and applications. Finally, we describe the lessons learned by using LDL++ on various tested applications, such as middleware and datamining.