Learning to create data-integrating queries

Abstract

The number of potentially-related data resources available for querying -databases, data warehouses, virtual integrated schemascontinues to grow rapidly. Perhaps no area has seen this problem as acutely as the life sciences, where hundreds of large, complex, interlinked data resources are available on fields like proteomics, genomics, disease studies, and pharmacology. The schemas of individual databases are often large on their own, but users also need to pose queries across multiple sources, exploiting foreign keys and schema mappings. Since the users are not experts, they typically rely on the existence of pre-defined Web forms and associated query templates, developed by programmers to meet the particular scientists' needs. Unfortunately, such forms are scarce commodities, often limited to a single database, and mismatched with biologists' information needs that are often context-sensitive and span multiple databases.We present a system with which a non-expert user can author new query templates and Web forms, to be reused by anyone with related information needs. The user poses keyword queries that are matched against source relations and their attributes; the system uses sequences of associations (e.g., foreign keys, links, schema mappings, synonyms, and taxonomies) to create multiple ranked queries linking the matches to keywords; the set of queries is attached to a Web query form. Now the user and his or her associates may pose specific queries by filling in parameters in the form. Importantly, the answers to this query are ranked and annotated with data provenance, and the user provides feedback on the utility of the answers, from which the system ultimately learns to assign costs to sources and associations according to the user's specific information need, as a result changing the ranking of the queries used to generate results. We evaluate the effectiveness of our method against "gold standard" costs from domain experts and demonstrate the method's scalability.

Incrementally improving dataspaces based on user feedback

“…In this paper, we described an approach in which we assumed that the integration schema together with a set of candidate schema mappings are given, and showed how those mappings can be annotated and refined over time to identify the mappings that meet user requirements. While the techniques we presented in this paper are in the spirit of dataspaces, they represent a point in the broad space of dataspace solutions, which includes aspects such as bootstrapping dataspaces to provide users with services from the start [49], providing users with the means for querying heterogeneous data sources in the absence of schema mappings, using e.g., keyword search [52,51], indexing dataspaces [17], profiling dataspaces in the absence of schema information [30]. In this section, we analyze and compare these proposals to ours.…”

A Functional Model for Dataspace Management Systems

“…Matches are chosen based on their utility with respect to a query workload that is provided in advance. ORCHESTRA [62,36], a collaborative data sharing system, covering the three phases initialization, usage and improvement, uses a generic graph structure to store the schemas and matches between schema elements, which are derived semi-automatically and annotated with costs representing the bias of the system against using the matches. Mappings in the form of query templates are derived from keyword queries posed by the user and matched against the schemas and matches.…”

“…u := gather(r) This operation is construed as providing the means by which a set u of feedback instances can be gathered. Incremental improvement based on user feedback can take a variety of forms: through the manual provision of mappings (e.g., [64]); through the annotation of query results as to which items are spurious or which should be ranked higher (e.g., [62]); through a more intensively interactive approach requiring a fair amount of user input during the integration process (e.g., [37]), or through a process by which mappings are debugged (e.g., [50]). We observe that all these approaches require, to different degrees, an understanding of the syntax and semantics of mapping and schema languages on the part of the person providing the feedback.…”

“…In our own work, we have focussed on query evaluation, but there are many other ways in which a dataspace could be used, e.g., browsing [35], keyword searching [14,43,46,62], or interaction based on the notion of trails [18,64]. Improvement: This stage is characteristic of dataspaces and aims to counteract the shortcomings ensuing from the reliance on automation for bootstrapping (the other characteristic feature of DSMSs in this context).…”

“…In our own work [6], we have used feedback on query results to annotate, select and refine the collection of mappings that can be used to answer a given query by taking into account the possibilities for trading off precision and recall. It is possible and useful to gather other kinds of feedback, e.g., on queries [15]; on mappings [1]; or for query specification [61,62]. Maintenance: Changes to the sources must be propagated throughout.…”