Supporting Schema References in Keyword Queries Over Relational Databases

Abstract

Relational Keyword Search (R-KwS) systems enable naive/informal users to explore and retrieve information from relational databases without knowing schema details or query languages. They take a keyword query, locate their corresponding elements in the target database, and connect them using information on PK/FK constraints. Although there are many such systems in the literature, most of them only support queries with keywords referring to the contents of the database and just very few support queries with keywords refering the database schema. We propose Lathe, a novel R-KwS that supports such queries.To this end, we first generalize the well-known concepts of Candidate Joining Networks (CJNs) and Query Matches (QMs) to handle keywords referring to schema elements and propose new algorithms to generate them. Then, we introduce two major innovations: a ranking algorithm for selecting better QMs, yielding the generation of fewer but better CJNs, and an eager evaluation strategy for pruning void useless CJNs. We present experiments performed with query sets and datasets previously experimented with state-of-theart R-KwS systems. Our results indicate that Lathe can handle a wider variety of queries while remaining highly effective, even for databases with intricate schemas.

SEREIA - Document Store Exploration through Keywords

“…In general, many combinations of KMs of a keyword query may return documents that, when correctly connected, satisfy the query given as input. However, in accordance with previous work [12,13,15,16], we consider in this work only combinations that correspond to a "total" and "minimal" cover of a keyword query. More precisely, a total cover guarantees that all terms from the keyword query are present in the result, and a minimal cover guarantees that there are no documents with redundant information.…”

“…State-of-the-art systems in this category employ an early evaluation to assess which CJNs better represent the original keyword query. This evaluation improves not only the overall quality of the process, but also its efficiency [12][13][14]16]. The second category represents instance-based systems that aim to materialize database tuples in a graph called Data Graph.…”

“…Since many QMs can be generated for a single keyword query, SEREIA uses a ranking algorithm based on the CNRank algorithm [12] to identify QMs that are useful and more prone to satisfy the user's intent. SEREIA estimates the relevance of QMs based on a Bayesian Belief Network model for the current state of the underlying database [16]. In practice, this model evaluates two types of relevance evidence when ranking QMs.…”

“…In our work, based on the terminology used in previous work [13,16], we call these intermediate collections keyword-free matches, which are essential for the definition and generation of Pipelines Sketches. Specifically, these matches assist when generating PLSs by acting as intermediate nodes to other KMs.…”

“…A sound PLS indicates that a KM cannot point to the same collection more times than the number of attributes that reference that same collection. This concept of soundness is adapted from previous work [16].…”