Natural-language database interface with automated keyword mapping and join-path inferences

A method for inferring a join path from a natural-language database query, the method comprising:

• receiving, by a database management system (DBMS), a natural-language query (NLQ) submitted by a user of the DBMS, wherein the NLQ requests access to information organized into at least two elements of a database managed by the DBMS, wherein each element of the at least two elements is either a relation of the database or an attribute of the database;
• extracting, by the DBMS, a set of keywords from the NLQ;
• selecting, by the DBMS, a set of candidate mapping configurations, wherein each configuration of the set of candidate mapping configurations consists of a unique combination of keyword mappings that maps the entire set of extracted keywords onto a set of query fragments;
• retrieving, by the DBMS, a query fragment graph (QFG) that counts occurrences, in previously logged structured queries, of each query fragment identified by a keyword mapping of any configuration of the set of candidate mapping configurations;
• choosing, by the DBMS, an optimal configuration from the set of candidate mapping configurations as a function of information contained in the QFG;
• identifying, by the DBMS, that the optimal configuration comprises at least two keyword mappings, wherein each of the at least two keyword mappings spans two or more tables of the database;
• inferring, by the DBMS, an optimal join path by comparing the relative numbers of times that each distinct pair of query fragments, of an optimal set of query fragments, co-occurred in a query of the previously logged structured queries,
• wherein the optimal join path joins at least two query fragments, of the optimal set of query fragments, each of which represents a keyword of one of the at least two keyword mappings;
• generating, by the DBMS, a structured query that comprises the optimal join path, and
• wherein the optimal set of query fragments contains query fragments, occurring in the previously logged structured queries, and co-occurring in any candidate configuration of the set of candidate mapping configurations more often than does any set of query fragments identified by other configurations of the candidate mapping configurations;
• and
• forwarding the structured query, by the DBMS, to a downstream component of the DBMS.

A computer program product, comprising a computer-readable hardware storage device having a computer-readable program code stored therein, the program code configured to be executed by a database management system (DBMS) comprising a processor, a memory coupled to the processor, and a computer-readable hardware storage device coupled to the processor, the storage device containing program code configured to be run by the processor via the memory to implement a method for a natural-language database interface with automated keyword mapping and join-path inferences, the method comprising:

• receiving, by the DBMS, a natural-language query (NLQ) submitted by a user of the DBMS, wherein the NLQ requests access to information organized into at least two elements of a database managed by the DBMS, wherein each element of the at least two elements is either a relation of the database or an attribute of the database;
• extracting, by the DBMS, a set of keywords from the NLQ;
• selecting, by the DBMS, a set of candidate mapping configurations, wherein each configuration of the set of candidate mapping configurations identifies a unique combination of keyword mappings that maps the entire set of extracted keywords onto a set of query fragments;
• retrieving, by the DBMS, a query fragment graph (QFG) that counts occurrences, in previously logged structured queries, of each query fragment identified by a keyword mapping of any configuration of the set of candidate mapping configurations;
• choosing, by the DBMS, an optimal configuration from the set of candidate mapping configurations,
• wherein the optimal configuration identifies an optimal set of query fragments, and
• wherein the optimal set of query fragments contains pairs of query fragments that are identified by the QFG as co-occurring in the previously logged structured queries more frequently than do pairs of query fragments identified by other configurations of the candidate mapping configurations;
• generating, by the DBMS, a structured query that comprises the optimal set of query fragments;
• and
• forwarding the structured query, by the DBMS, to a downstream component of the DBMS.

A method, comprising:

• creating a Query Fragment Graph (QFG) data structure that identifies a number of co-occurrences of query fragments that occur within any query of a set of structured queries stored in a log of a database;
• given a keyword extracted from a Natural Language Query (NLQ), using the QFG data structure to rank a set of keyword mappings,
• wherein each mapping of the keyword mappings associates the keyword with an attribute or relation of the database, and
• wherein each ranked mapping is each ranked as a function of a relative number of occurrences, in the set of structured queries, of a database attribute or relation associated with the keyword by the ranked mapping;
• given an attribute or relation associated with the keyword by one of the keyword mappings, using the QFG data structure to identify a ranked list of join paths,
• wherein each ranked join path joins the associated attribute or relation to another attribute or relation of the database, and
• wherein the ranked join paths are each ranked as a function of a relative number of occurrences, in the set of structured queries, of each ranked join path;
• and
• using a Natural Language Interface for Databases (NLIDB) to translate the NLQ into a translated structured query that incorporates a highest-ranked join path,
• wherein the translated structured query has a semantic meaning analogous to an intended meaning of the NLQ.