Data provenance is any information about the origin of a piece of data and the process that led to its creation. Most database provenance work has focused on creating models and semantics to query and generate this provenance information. While comprehensive, provenance information remains large and overwhelming, making it hard for data provenance systems to support data exploration. We present a new approach to provenance exploration that builds on data summarization techniques. We contribute novel summarization schemes for the provenance of aggregation queries and techniques for the fast generation of these summarization schemes. We introduce two types of summaries for aggregate queries.Impact summariestake into account the impact of specific groups of tuples in the provenance of the query on an aggregate result, andcomparative summariesallow users to compare the provenance of two aggregate results. We also present algorithms for efficient computation of these summaries, implement optimizations using data sampling and feature selection, and conduct experiments and a user survey to show the feasibility and relevance of our approaches.

Data provenance is any information about the origin of a piece of data and the process that led to its creation. Most database provenance work has focused on creating models and semantics to query and generate this information. While comprehensive, provenance information remains large and overwhelming, which can make it hard for provenance systems to support data exploration. We present a new approach to provenance exploration that builds on data summarization techniques. We contribute two novel summarization schemes for the provenance of aggregation queries: Impact summaries, and comparative summaries. We show with experiments that our techniques incur little overhead compared to basic summaries. We conduct a survey to show that our approaches are useful to users.

Provenance information can be large and overwhelming to users. We present a set of criteria that any provenance exploration tool must have and introduce Pastwatch, a provenance exploration system that adheres to those criteria. We also address the issues associated with provenance of aggregation queries, including the creation of a summarization method that makes provenance of aggregation queries manageable for users. Finally, we conduct a quantitative user study to show statistically significant results that Pastwatch makes provenance information more efficient and easier to use than standard approaches.

Formulating efficient SQL queries requires several cycles of tuning and execution. We examine methods that can accelerate and improve this interaction by providing insights about SQL queries prior to execution. We achieve this by predicting properties such as the query answer size, its run-time, and error class. Unlike existing approaches, our approach does not rely on any statistics from the database instance or query execution plans. Our approach is based on using data-driven machine learning techniques that rely on large query workloads to model SQL queries and their properties. Empirical results show that the neural network models are more accurate in predicting several query properties.

Query recommendation is critical to assisting first-time users, who may not have the knowledge necessary to know how to issue effective SQL queries, especially in scientific databases. To help users learn how to issue SQL queries, we turn to recommendation. In particular, we consider how to recommend SQL queries and other relevant features to users using nearest-neighbor collaborative filtering. This paper presents Skyrec Summary, an improved summary method which represents the number of resulting tuples queried from each session as an importance rating. The paper adopts SkyServer, an astronomy log dataset, and presents SkyServer Surfliner, a recommender system that supports interactive database exploration, for comparative evaluation of different summary methods. A user study on SkyServer Surfliner shows that Skyrec summary method can quickly provide helpful query recommendations.

Standard collaborative filtering approaches for top-N recommendation are biased toward popular items. As a result, they recommend items that users are likely aware of and under-represent long-tail items. This is inadequate, both for consumers who prefer novel items and because concentrating on popular items poorly covers the item space, whereas high item space coverage increases providers' revenue. We present an approach that relies on historical rating data to learn user long-tail novelty preferences. We integrate these preferences into a generic re-ranking framework that customizes balance between accuracy and coverage. We empirically validate that our proposed framework increases the novelty of recommendations. Furthermore, by promoting long-tail items to the right group of users, we significantly increase the system's coverage while scalably maintaining accuracy. Our framework also enables personalization of existing non-personalized algorithms, making them competitive with existing personalized algorithms in key performance metrics, including accuracy and coverage.

Aggregate queries in integration contexts often do not have one“true”answer; there can be multiple correct answers for the same aggregate query. This is due to the existence of duplicate or overlapping data points, possibly with di↵erent values, across the data sources. Depending on the choice of data source combinations that are used to answer the query, di↵erent answers can be generated. Thus, representing the answer to the aggregate query as an answer distribution instead of a single scalar value, will allow the users to better understand the range of possible answers. This work provides a suite of methods for extracting statistics that convey meaningful information about aggregate query answers in heterogeneous integration settings. We focus on the following challenges: 1. determining which statistics best represent an answer’s distribution; and 2. eciently computing the desired statistics. Our solution includes the following answer statistics 1. a set of point estimates with confidence intervals; 2. a high coverage interval that unveils “hot areas” in a distribution; and 3. a stability score that measures the impact of source dynamics. We optimize the extraction of the above statistical information by minimizing the sampling load and applying fast approximate algorithms. We verify the e↵ectiveness and eciency of our methods with empirical studies using real-life and synthetic, scaled data sets.

Extracting ontological relationships (e.g., ISA and HASA) from free-text repositories (e.g., engineering documents and instruction manuals) can improve users' queries, as well as benefit applications built for these domains.

In a Peer Data Management System (PDMS), autonomous peers share semantically rich data. For queries to be translated across peers, a peer must provide a mapping to other peers in the PDMS; peers connected by such mappings are called acquaintances. To maximize PDMS query answering performance, a peer needs to optimize its choice of acquaintances. This paper investigates the acquaintance selection problem and introduces a novel framework for performing this acquaintance selection. Our framework includes two selection schemes that effectively and efficiently estimate mapping effectiveness. The "one-shot" scheme clusters peers and estimates the improvement in query answering based on cluster properties. The "two-hop" scheme estimates using locally available information at multiple rounds. Our empirical study shows that both schemes effectively help acquaintance selection and scale to large PDMSs.

Peer Data Management Systems (PDMSs) allow the efficient sharing of data between peers with overlapping sources of information. These sources share data through mappings between peers. In current systems, queries are asked over each peer's local schema and then translated using the mappings between peers. While this allows the data to be accessed uniformly, users lack access to information that is not in their own schemas. In this paper, we propose a light-weight, automatic method to create a mediated schema in a PDMS. Our work benefits PDMSs by allowing access to more data and without unduly stressing the peer's resources or requiring additional resources such as ontologies. We present our system — MePSys, which creates a mediated schema in PDMSs automatically using the existing mappings provided to translate queries. We further discuss how to update the mediated schema in a stable state, i.e. after the system setup period.

Manipulating graph-structured schemas (ontologies, models, etc.) requires the result to remain fully connected. In certain cases, e.g., calculating the difference of two schemas, support structures may be needed in the result. We describe our engine to process support structures in the context of a schema management system and describe schema reintegration experiments which validate the performance and correctness of our system

We address the problem of generating a mediated schema from a set of relational data source schemas and conjunctive queries that specify where those schemas overlap. Unlike past approaches that generate only the mediated schema, our algorithm also generates view definitions, i.e., source-to-mediated schema mappings.

Most previous schema mapping works focus on creating mappings in specific data models for data transformation, failing to capture a richer set of possible relationships between schema elements. For example, most schema matching approaches might discover that 'TA' in one schema equals 'grad TA' in another one, even though the relationship can be modeled more accurately by saying that 'grad TA' is a specialization of 'TA'. Deepening the mapping semantics in turn allow richer application semantics. This paper presents and proves the effectiveness of SeMap, a system that constructs a complex, semantically richer mapping (including 'Has-a', 'Is-a', 'Associates' and 'Equivalent' relationship types) that can be used across data models. We achieve this goal by: (1) exploiting semantic evidence for possible matches; (2) finding a globally optimal match assignment; (3) identifying the relationship embedded in the selected matches. We implemented our semantic matching approach within a prototype system, SeMap, and showed its accuracy and effectiveness.

No abstract available.