Browse Prior Art Database

QUANTRA - Query Understander And Translator

IP.com Disclosure Number: IPCOM000019861D
Original Publication Date: 2003-Oct-02
Included in the Prior Art Database: 2003-Oct-02
Document File: 3 page(s) / 34K

Publishing Venue

IBM

Abstract

QUANTRA is a simple, yet powerful and portable natural language front end to relational databases. It answers queries given by the user in English by converting them into SQL and executing the SQL queries against the database. It is easily configurable to any relational database with minimal manual intervention.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 3

QUANTRA - Query Understander And Translator

  Disclosed is a system that can be a simple, yet powerful and portable natural language front end to relational databases. It answers queries given by the user in English by converting them into SQL and executing the SQL queries against the database. It is easily configurable to any relational database with minimal manual intervention.

QUANTRA takes a unique approach of performing syntax parsing like MASQUE/SQL [1] and using DB-Models similar to semantic graphs of [2] and [3] as knowledge. Unlike [1], it does not suffer from the limitations of tedious knowledge coding phases. Also, QUANTRA does not use statistical methods to disambiguate at any level and hence does not suffer from the limitations of
[2] and [3].

The basic problem that QUANTRA is solving is this. Given an input database query in English, obtain the corresponding database query in SQL. In order to do this, there is a need to have knowledge of the database structures and the lexicon that can be used by the users in issuing English queries. The initial step is to generate knowledge from the database using reverse engineering techniques [4]. This will give DB models (similar to ER Models) that semantically represent the database and are structured as graphs. Facilities to manually add more knowledge into it are provided. This task of generating the semantic knowledge from the database and storing them as graphs is a one time process.

Next, note that a DB model (the main source of knowledge) for QUANTRA is a graph. A subgraph of this graph represents a query (similar to [2]). Once this subgraph is identified, an SQL query can be generated out of it. Hence the problem now gets reduced to this. From the input English database query , obtain a subgraph of the DB model that represents the user's query.

At runtime, i.e. when the user provides the English query, the input English expression is parsed using a grammar like a dependency grammar. This parsing will give rise to syntactic relationships

1

[This page contains 2 pictures or other non-text objects]

Page 2 of 3

between the words uttered in the natural language input. These syntactic relationships are captured in the form of a syntactic graph...