Search

We are extending the keyword-based query interface DdQl for relational databases which is based on contextual background knowledge such as suitable join conditions and which was proposed in [{Dietmar Seipel, 2021]. In the previous paper, join conditions were extracted from existing referential integrity (foreign key) constraints of the database schema, or they could be learned from other, previous database queries. In this paper, we describe a speech-to-text component for entering the query keywords based on the system Whisper. Keywords, which have been recognized wrongly by Whisper can be corrected to similarly sounding words. Again, the context of the database schema can help here. For users with a limited knowledge of the schema and the contents of the database, the approach of DdQl can help to provide useful suggestions for query implementations in Sql or Datalog, from which the user can choose one. Our tool DdQl can be run in a docker image; it yields the possible queries in Sql and a special domain specific rule language that extends Datalog. The Datalog variant allows for additional user-defined aggregation functions which are not possible in Sql.

We are proposing a keyword-based query interface for knowledge bases - including relational or deductive databases - based on contextual background knowledge such as suitable join conditions or synonyms. Join conditions could be extracted from existing referential integrity (foreign key) constaints of the database schema. They could also be learned from other, previous database queries, if the database schema does not contain foreign key constraints. Given a textual representation - a word list - of a query to a relational database, one may parse the list into a structured term. The intelligent and cooperative part of our approach is to hypothesize the semantics of the word list and to find suitable links between the concepts mentioned in the query using contextual knowledge, more precisely join conditions between the database tables. We use a knowledge-based parser based on an extension of Definite Clause Grammars (Dcg) that are interweaved with calls to the database schema to suitably annotate the tokens as table names, table attributes, attribute values or relationships linking tables. Our tool DdQl yields the possible queries in a special domain specific rule language that extends Datalog, from which the user can choose one.

Definite Clause Grammars (DCGs) are a convenient way to specify possibly non-context-free grammars for natural and formal languages. They can be used to progressively build a parse tree as grammar rules are applied by providing an extra argument in the DCG rule’s head. In the simplest way, this is a structure that contains the name of the used nonterminal. This extension of a DCG has been proposed for natural language processing in the past and can be done automatically in Prolog using term expansion. We extend this approach by a meta-nonterminal to specify optional and sequences of nonterminals, as these structures are common in grammars for formal, domain-specific languages. We specify a term expansion that represents these sequences as lists while preserving the grammar’s ability to be used both for parsing and serialising, i.e. to create a parse tree by a given source code and vice-versa. We show that this mechanism can be used to lift grammars specified in extended Backus-Naur form (EBNF) to generate parse trees. As a case study, we present a parser for the Prolog programming language itself based only on the grammars given in the ISO Prolog standard which produces corresponding parse trees.

As with Amazon’s Echo and its conversational agent Alexa, smart voice-controlled devices become ever more present in daily life, and many different applications can be integrated into this platform. In this paper, we present a framework that eases the development of skills in Prolog. As Prolog has a long history in natural language processing, we may integrate well-established techniques, such as reasoning about knowledge with Attempto Controlled English, instead of depending on example phrases and pre-defined slots.

In this paper, we use declarative and domain-specific languages for representing expert knowledge in the field of change management in organisational psychology. Expert rules obtained in practical case studies are represented as declarative rules in a deductive database. The expert rules are annotated by information describing their provenance and confidence. Additional provenance information for the whole - or parts of the - rule base can be given by ontologies. Deductive databases allow for declaratively defining the semantics of the expert knowledge with rules; the evaluation of the rules can be optimised and the inference mechanisms could be changed, since they are specified in an abstract way. As the logical syntax of rules had been a problem in previous applications of deductive databases, we use specially designed domain-specific languages to make the rule syntax easier for non-programmers. The semantics of the whole knowledge base is declarative. The rules are written declaratively in an extension datalogs of the well-known deductive database language datalog on the data level, and additional datalogs rules can configure the processing of the annotated rules and the ontologies.