arXiv subjects

J. Shane Culpepper

Publications and source records attributed to J. Shane Culpepper.

At least 19 recordsLinked to original sources

arXiv2026

Unified Data Discovery across Query Modalities and User Intents

Data discovery - retrieving relevant tables from a data lake in response to user queries - is a fundamental building block for downstream analytics. In practice, data discovery must support different query modalities, including natural language (NL) statements and tables, and accommodate diverse user intents, ranging from open-ended enrichment to task-driven inference for applications such as table question answering and fact verification. However, most existing methods are designed for a single query modality or a specific user intent, limiting their generalizability. We propose UniDisc, a unified data discovery framework that supports both NL statements and tables as queries and generalizes across diverse user intents without intent-specific representations or relevance modeling. UniDisc learns a common cross-modal representation model that produces unified representations for queries of different modalities and candidate tables, enabling uniform relevance assessment across discovery scenarios. Since learning such a model typically requires large labeled collections of query-table pairs, which are expensive to obtain, UniDisc instead exploits contextual signals naturally available in data lakes. Specifically, it models NL statements and tables as nodes in a heterogeneous graph with multiple edge types, and applies dual-view neighbor aggregation and joint optimization to learn robust, context-aware representations under limited supervision. These representations support flexible relevance estimation during retrieval. Experiments on seven datasets show that UniDisc consistently outperforms strong baselines on both NL- and table-based discovery.

cs.DB↗

arXiv2026

Decomposition-Driven Multi-Table Retrieval and Reasoning for Numerical Question Answering

In this paper, we study the problem of numerical multi-table question answering (MTQA) over large-scale table collections (e.g., online data repositories). This task is essential in many analytical applications. Existing MTQA solutions, such as text-to-SQL or open-domain MTQA methods, are designed for databases and struggle when applied to large-scale table collections. The key limitations include: (1) Limited support for complex table relationships; (2) Ineffective retrieval of relevant tables at scale; (3) Inaccurate answer generation. To overcome these limitations, we propose DMRAL, a Decomposition-driven Multi-table Retrieval and Answering framework for MTQA over large-scale table collections, which consists of: (1) constructing a table relationship graph to capture complex relationships among tables; (2) Table-Aligned Question Decomposer and Coverage-Aware Retriever, which jointly enable the effective identification of relevant tables from large-scale corpora by enhancing the question decomposition quality and maximizing the question coverage of retrieved tables; and (3) Sub-question Guided Reasoner, which produces correct answers by progressively generating and refining the reasoning program based on sub-questions. Experiments on two MTQA datasets demonstrate that DMRAL significantly outperforms existing state-of-the-art MTQA methods, with an average improvement of 24% in table retrieval and 55% in answer accuracy.

cs.DB↗

arXiv2026

Revisiting Human-vs-LLM judgments using the TREC Podcast Track

Using large language models (LLMs) to annotate relevance is an increasingly important technique in the information retrieval community. While some studies demonstrate that LLMs can achieve high user agreement with ground truth (human) judgments, other studies have argued for the opposite conclusion. To the best of our knowledge, these studies have primarily focused on classic ad-hoc text search scenarios. In this paper, we conduct an analysis on user agreement between LLM and human experts, and explore the impact disagreement has on system rankings. In contrast to prior studies, we focus on a collection composed of audio files that are transcribed into two-minute segments -- the TREC 2020 and 2021 podcast track. We employ five different LLM models to re-assess all of the query-segment pairs, which were originally annotated by TREC assessors. Furthermore, we re-assess a small subset of pairs where LLM and TREC assessors have the highest disagreement, and found that the human experts tend to agree with LLMs more than with the TREC assessors. Our results reinforce the previous insights of Sormunen in 2002 -- that relying on a single assessor leads to lower user agreement.

cs.IR↗

arXiv2025

The Effects of Demographic Instructions on LLM Personas

Social media platforms must filter sexist content in compliance with governmental regulations. Current machine learning approaches can reliably detect sexism based on standardized definitions, but often neglect the subjective nature of sexist language and fail to consider individual users' perspectives. To address this gap, we adopt a perspectivist approach, retaining diverse annotations rather than enforcing gold-standard labels or their aggregations, allowing models to account for personal or group-specific views of sexism. Using demographic data from Twitter, we employ large language models (LLMs) to personalize the identification of sexism.

cs.IR↗

arXiv2025

Dataset Discovery via Line Charts

Line charts are a valuable tool for data analysis and exploration, distilling essential insights from a dataset. However, access to the underlying dataset behind a line chart is rarely readily available. In this paper, we explore a novel dataset discovery problem, dataset discovery via line charts, focusing on the use of line charts as queries to discover datasets within a large data repository that are capable of generating similar line charts. To solve this problem, we propose a novel approach called Fine-grained Cross-modal Relevance Learning Model (FCM), which aims to estimate the relevance between a line chart and a candidate dataset. To achieve this goal, FCM first employs a visual element extractor to extract informative visual elements, i.e., lines and y-ticks, from a line chart. Then, two novel segment-level encoders are adopted to learn representations for a line chart and a dataset, preserving fine-grained information, followed by a cross-modal matcher to match the learned representations in a fine-grained way. Furthermore, we extend FCM to support line chart queries generated based on data aggregation. Last, we propose a benchmark tailored for this problem since no such dataset exists. Extensive evaluation on the new benchmark verifies the effectiveness of our proposed method. Specifically, our proposed approach surpasses the best baseline by 30.1% and 41.0% in terms of prec@50 and ndcg@50, respectively.

cs.DB↗

arXiv2025

Distinctiveness Maximization in Datasets Assemblage

In this paper, given a user's query set and budget, we aim to use the limited budget to help users assemble a set of datasets that can enrich a base dataset by introducing the maximum number of distinct tuples (i.e., maximizing distinctiveness). We prove this problem to be NP-hard. A greedy algorithm using exact distinctiveness computation attains an approximation ratio of (1-1/e)/2, but it lacks efficiency and scalability due to its frequent computation of the exact distinctiveness marginal gain of any candidate dataset for selection. This requires scanning through every tuple in candidate datasets and thus is unaffordable in practice. To overcome this limitation, we propose an efficient machine learning (ML)-based method for estimating the distinctiveness marginal gain of any candidate dataset. This effectively eliminates the need to test each tuple individually. Estimating the distinctiveness marginal gain of a dataset involves estimating the number of distinct tuples in the tuple sets returned by each query in a query set across multiple datasets. This can be viewed as the cardinality estimation for a query set on a set of datasets, and the proposed method is the first to tackle this cardinality estimation problem. This is a significant advancement over prior methods that were limited to single-query cardinality estimation on a single dataset and struggled with identifying overlaps among tuple sets returned by each query in a query set across multiple datasets. Extensive experiments using five real-world data pools demonstrate that our algorithm, which utilizes ML-based distinctiveness estimation, outperforms all relevant baselines in effectiveness, efficiency, and scalability. A case study on two downstream ML tasks also highlights its potential to find datasets with more useful tuples to enhance the performance of ML tasks.

cs.DB↗

arXiv2023

A Simple Yet High-Performing On-disk Learned Index: Can We Have Our Cake and Eat it Too?

While in-memory learned indexes have shown promising performance as compared to B+-tree, most widely used databases in real applications still rely on disk-based operations. Based on our experiments, we observe that directly applying the existing learned indexes on disk suffers from several drawbacks and cannot outperform a standard B+-tree in most cases. Therefore, in this work we make the first attempt to show how the idea of learned index can benefit the on-disk index by proposing AULID, a fully on-disk updatable learned index that can achieve state-of-the-art performance across multiple workload types. The AULID approach combines the benefits from both traditional indexing techniques and the learned indexes to reduce the I/O cost, the main overhead under disk setting. Specifically, three aspects are taken into consideration in reducing I/O costs: (1) reduce the overhead in updating the index structure; (2) induce shorter paths from root to leaf node; (3) achieve better locality to minimize the number of block reads required to complete a scan. Five principles are proposed to guide the design of AULID which shows remarkable performance gains and meanwhile is easy to implement. Our evaluation shows that AULID has comparable storage costs to a B+-tree and is much smaller than other learned indexes, and AULID is up to 2.11x, 8.63x, 1.72x, 5.51x, and 8.02x more efficient than FITing-tree, PGM, B+-tree, ALEX, and LIPP.

cs.DB↗

arXiv2023

Updatable Learned Indexes Meet Disk-Resident DBMS -- From Evaluations to Design Choices

Although many updatable learned indexes have been proposed in recent years, whether they can outperform traditional approaches on disk remains unknown. In this study, we revisit and implement four state-of-the-art updatable learned indexes on disk, and compare them against the B+-tree under a wide range of settings. Through our evaluation, we make some key observations: 1) Overall, the B+-tree performs well across a range of workload types and datasets. 2) A learned index could outperform B+-tree or other learned indexes on disk for a specific workload. For example, PGM achieves the best performance in write-only workloads while LIPP significantly outperforms others in lookup-only workloads. We further conduct a detailed performance analysis to reveal the strengths and weaknesses of these learned indexes on disk. Moreover, we summarize the observed common shortcomings in five categories and propose four design principles to guide future design of on-disk, updatable learned indexes: (1) reducing the index's tree height, (2) better data structures to lower operation overheads, (3) improving the efficiency of scan operations, and (4) more efficient storage layout.

cs.DB↗

arXiv2021

Let Trajectories Speak Out the Traffic Bottlenecks

Traffic bottlenecks are a set of road segments that have an unacceptable level of traffic caused by a poor balance between road capacity and traffic volume. A huge volume of trajectory data which captures real-time traffic conditions in road networks provides promising new opportunities to identify the traffic bottlenecks. In this paper, we define this problem as trajectory-driven traffic bottleneck identification: Given a road network R, a trajectory database T , find a representative set of seed edges of size K of traffic bottlenecks that influence the highest number of road segments not in the seed set. We show that this problem is NP-hard and propose a framework to find the traffic bottlenecks as follows. First, a traffic spread model is defined which represents changes in traffic volume for each road segment over time. Then, the traffic diffusion probability between two connected segments and the residual ratio of traffic volume for each segment can be computed using historical trajectory data. We then propose two different algorithmic approaches to solve the problem. The first one is a best-first algorithm BF, with an approximation ratio of 1-1/e. To further accelerate the identification process in larger datasets, we also propose a sampling-based greedy algorithm SG. Finally, comprehensive experiments using three different datasets compare and contrast various solutions, and provide insights into important efficiency and effectiveness trade-offs among the respective methods.

cs.DB↗

arXiv2021

Spatial Object Recommendation with Hints: When Spatial Granularity Matters

Existing spatial object recommendation algorithms generally treat objects identically when ranking them. However, spatial objects often cover different levels of spatial granularity and thereby are heterogeneous. For example, one user may prefer to be recommended a region (say Manhattan), while another user might prefer a venue (say a restaurant). Even for the same user, preferences can change at different stages of data exploration. In this paper, we study how to support top-k spatial object recommendations at varying levels of spatial granularity, enabling spatial objects at varying granularity, such as a city, suburb, or building, as a Point of Interest (POI). To solve this problem, we propose the use of a POI tree, which captures spatial containment relationships between POIs. We design a novel multi-task learning model called MPR (short for Multi-level POI Recommendation), where each task aims to return the top-k POIs at a certain spatial granularity level. Each task consists of two subtasks: (i) attribute-based representation learning; (ii) interaction-based representation learning. The first subtask learns the feature representations for both users and POIs, capturing attributes directly from their profiles. The second subtask incorporates user-POI interactions into the model. Additionally, MPR can provide insights into why certain recommendations are being made to a user based on three types of hints: user-aspect, POI-aspect, and interaction-aspect. We empirically validate our approach using two real-life datasets, and show promising performance improvements over several state-of-the-art methods.

cs.IR↗

arXiv2020

A Survey on Trajectory Data Management, Analytics, and Learning

Recent advances in sensor and mobile devices have enabled an unprecedented increase in the availability and collection of urban trajectory data, thus increasing the demand for more efficient ways to manage and analyze the data being produced. In this survey, we comprehensively review recent research trends in trajectory data management, ranging from trajectory pre-processing, storage, common trajectory analytic tools, such as querying spatial-only and spatial-textual trajectory data, and trajectory clustering. We also explore four closely related analytical tasks commonly used with trajectory data in interactive or real-time processing. Deep trajectory learning is also reviewed for the first time. Finally, we outline the essential qualities that a trajectory data management system should possess in order to maximize flexibility.

cs.DB↗

arXiv2020

Boosting Search Performance Using Query Variations

Rank fusion is a powerful technique that allows multiple sources of information to be combined into a single result set. However, to date fusion has not been regarded as being cost-effective in cases where strict per-query efficiency guarantees are required, such as in web search. In this work we propose a novel solution to rank fusion by splitting the computation into two parts -- one phase that is carried out offline to generate pre-computed centroid answers for queries with broadly similar information needs, and then a second online phase that uses the corresponding topic centroid to compute a result page for each query. We explore efficiency improvements to classic fusion algorithms whose costs can be amortized as a pre-processing step, and can then be combined with re-ranking approaches to dramatically improve effectiveness in multi-stage retrieval systems with little efficiency overhead at query time. Experimental results using the ClueWeb12B collection and the UQV100 query variations demonstrate that centroid-based approaches allow improved retrieval effectiveness at little or no loss in query throughput or latency, and with reasonable pre-processing requirements. We additionally show that queries that do not match any of the pre-computed clusters can be accurately identified and efficiently processed in our proposed ranking pipeline.

cs.IR↗

arXiv2020

RMITB at TREC COVID 2020

Search engine users rarely express an information need using the same query, and small differences in queries can lead to very different result sets. These user query variations have been exploited in past TREC CORE tracks to contribute diverse, highly-effective runs in offline evaluation campaigns with the goal of producing reusable test collections. In this paper, we document the query fusion runs submitted to the first and second round of TREC COVID, using ten queries per topic created by the first author. In our analysis, we focus primarily on the effects of having our second priority run omitted from the judgment pool. This run is of particular interest, as it surfaced a number of relevant documents that were not judged until later rounds of the task. If the additional judgments were included in the first round, the performance of this run increased by 35 rank positions when using RBP p=0.5, highlighting the importance of judgment depth and coverage in assessment tasks.

cs.IR↗

arXiv2020

Dynamic Ridesharing in Peak Travel Periods

In this paper, we study a variant of the dynamic ridesharing problem with a specific focus on peak hours: Given a set of drivers and rider requests, we aim to match drivers to each rider request by achieving two objectives: maximizing the served rate and minimizing the total additional distance, subject to a series of spatio-temporal constraints. Our problem can be distinguished from existing work in three aspects: (1) Previous work did not fully explore the impact of peak travel periods where the number of rider requests is much greater than the number of available drivers. (2) Existing solutions usually rely on single objective optimization techniques, such as minimizing the total travel cost. (3) When evaluating the overall system performance, the runtime spent on updating drivers' trip schedules as per incoming rider requests should be incorporated, while it is excluded by most existing solutions. We propose an index structure together with a set of pruning rules and an efficient algorithm to include new riders into drivers' existing trip schedule. To answer new rider requests effectively, we propose two algorithms that match drivers with rider requests. Finally, we perform extensive experiments on a large-scale test collection to validate the proposed methods.

cs.DB↗

arXiv2020

Temporal Network Representation Learning via Historical Neighborhoods Aggregation

Network embedding is an effective method to learn low-dimensional representations of nodes, which can be applied to various real-life applications such as visualization, node classification, and link prediction. Although significant progress has been made on this problem in recent years, several important challenges remain, such as how to properly capture temporal information in evolving networks. In practice, most networks are continually evolving. Some networks only add new edges or nodes such as authorship networks, while others support removal of nodes or edges such as internet data routing. If patterns exist in the changes of the network structure, we can better understand the relationships between nodes and the evolution of the network, which can be further leveraged to learn node representations with more meaningful information. In this paper, we propose the Embedding via Historical Neighborhoods Aggregation (EHNA) algorithm. More specifically, we first propose a temporal random walk that can identify relevant nodes in historical neighborhoods which have impact on edge formations. Then we apply a deep learning model which uses a custom attention mechanism to induce node embeddings that directly capture temporal information in the underlying feature representation. We perform extensive experiments on a range of real-world datasets, and the results demonstrate the effectiveness of our new approach in the network reconstruction task and the link prediction task.

cs.LG↗

arXiv2019

The Potential of Learned Index Structures for Index Compression

Inverted indexes are vital in providing fast key-word-based search. For every term in the document collection, a list of identifiers of documents in which the term appears is stored, along with auxiliary information such as term frequency, and position offsets. While very effective, inverted indexes have large memory requirements for web-sized collections. Recently, the concept of learned index structures was introduced, where machine learned models replace common index structures such as B-tree-indexes, hash-indexes, and bloom-filters. These learned index structures require less memory, and can be computationally much faster than their traditional counterparts. In this paper, we consider whether such models may be applied to conjunctive Boolean querying. First, we investigate how a learned model can replace document postings of an inverted index, and then evaluate the compromises such an approach might have. Second, we evaluate the potential gains that can be achieved in terms of memory requirements. Our work shows that learned models have great potential in inverted indexing, and this direction seems to be a promising area for future research.

cs.IR↗

arXiv2018

The Maximum Trajectory Coverage Query in Spatial Databases

With the widespread use of GPS-enabled mobile devices, an unprecedented amount of trajectory data is becoming available from various sources such as Bikely, GPS-wayPoints, and Uber. The rise of innovative transportation services and recent break-throughs in autonomous vehicles will lead to the continued growth of trajectory data and related applications. Supporting these services in emerging platforms will require more efficient query processing in trajectory databases. In this paper, we propose two new coverage queries for trajectory databases: (i) k Maximizing Reverse Range Search on Trajectories (kMaxRRST); and (ii) a Maximum k Coverage Range Search on Trajectories (MaxkCovRST). We propose a novel index structure, the Trajectory Quadtree (TQ-tree) that utilizes a quadtree to hierarchically organize trajectories into different quadtree nodes, and then applies a z-ordering to further organize the trajectories by spatial locality inside each node. This structure is highly effective in pruning the trajectory search space, which is of independent interest. By exploiting the TQ-tree data structure, we develop a divide-and-conquer approach to compute the trajectory "service value", and a best-first strategy to explore the trajectories using the appropriate upper bound on the service value to efficiently process a kMaxRRST query. Moreover, to solve the MaxkCovRST, which is a non-submodular NP-hard problem, we propose a greedy approximation which also exploits the TQ-tree. We evaluate our algorithms through an extensive experimental study on several real datasets, and demonstrate that our TQ-tree based algorithms outperform common baselines by two to three orders of magnitude.

cs.DB↗

arXiv2017

Efficient and Effective Tail Latency Minimization in Multi-Stage Retrieval Systems

Scalable web search systems typically employ multi-stage retrieval architectures, where an initial stage generates a set of candidate documents that are then pruned and re-ranked. Since subsequent stages typically exploit a multitude of features of varying costs using machine-learned models, reducing the number of documents that are considered at each stage improves latency. In this work, we propose and validate a unified framework that can be used to predict a wide range of performance-sensitive parameters which minimize effectiveness loss, while simultaneously minimizing query latency, across all stages of a multi-stage search architecture. Furthermore, our framework can be easily applied in large-scale IR systems, can be trained without explicitly requiring relevance judgments, and can target a variety of different efficiency-effectiveness trade-offs, making it well suited to a wide range of search scenarios. Our results show that we can reliably predict a number of different parameters on a per-query basis, while simultaneously detecting and minimizing the likelihood of tail-latency queries that exceed a pre-specified performance budget. As a proof of concept, we use the prediction framework to help alleviate the problem of tail-latency queries in early stage retrieval. On the standard ClueWeb09B collection and 31k queries, we show that our new hybrid system can reliably achieve a maximum query time of 200 ms with a 99.99% response time guarantee without a significant loss in overall effectiveness. The solutions presented are practical, and can easily be used in large-scale distributed search engine deployments with a small amount of additional overhead.

cs.IR↗