A Paradigm for Learning Queries on Big Data

Abstract

International audienceSpecifying a database query using a formal query language is typically a challenging task for non-expert users. In the context of big data, this problem becomes even harder as it requires the users to deal with database instances of big sizes and hence difficult to visualize. Such instances usually lack a schema to help the users specify their queries, or have an incomplete schema as they come from disparate data sources. In this paper, we propose a novel paradigm for interactive learning of queries on big data, without assuming any knowledge of the database schema. The paradigm can be applied to different database models and a class of queries adequate to the database model. In particular, in this paper we present two instantiations that validated the proposed paradigm for learning relational join queries and for learning path queries on graph databases. Finally, we discuss the challenges of employing the paradigm for further data models and for learning cross-model schema mappings

Graph Queries

“…In other words, a node is certain with a label α if labeling it explicitly with α does not eliminate any query from C(G, S). The notion of certain nodes is inspired by possible world semantics and certain answers [29], and already employed for XML querying for nonexpert users [20] and for the inference of relational joins [14,15]. Additionally, given a graph G, a sample S, and a node ν, we say that ν is informative (w.r.t.…”

“…In Section 4.2, we have described a polynomial algorithm for learning graph queries of the basic RPQ class. More expressive fragments, for example those including conjunctions, can benefit from our previous work on learning relational queries [14]. A possible unification between the two lines of research would be desirable given the actual occurrences of C2RPQ in real-world query logs [18].…”

“…A possible unification between the two lines of research would be desirable given the actual occurrences of C2RPQ in real-world query logs [18]. Another direction of future work would be to actually ameliorate the interactive paradigm presented in [14,13,12]. In these instances, the initial sample is empty and the user gradually fills the sample by providing positive and negative labels, until the inferred query and the user goal query coincide.…”

Big-Graphs: Querying, Mining, and Beyond

“…(5) visual query [16,52], (6) natural language questions [119,28], (7) incorporating users' feedback [19,134], (8) query auto-completion and recommendation [77], (9) answers explanation [131,150,59], (10) conversational QA [176], etc. A one-time answer might not be satisfactory.…”

Query By Provenance

“…We do not impose such restrictions and are able to infer complex Conjunctive Queries. Last, there is a prominent line of work on query-by-example in the context of data exploration [26,7,2,6]. Here users typically provide an initial set of examples, leading to the generation of a consistent query (or multiple such queries); the queries and/or their results are presented to users, who may in turn provide feedback used to refine the queries, and so on.…”