UNIVERSITY OF DURHAM
LABORATORY FOR NATURAL LANGUAGE ENGINEERING
DEPARTMENT OF COMPUTER SCIENCE
Applications are invited for a research assistant to work on the
DEAR project which is building a natural language database interface
using the LOLITA natural-language processor and the Glasgow Haskell Compiler.
Candidates should have experience in the use of functional languages for programming
substantial software systems (preferably with lazy functional languages). Experience
in natural language processing and databases would
also be helpful.
The post is available immediately for a period of two years (but applicants
should be able to start by mid October at the latest). The salary will be
on the research 1A scale (14,317 to 17,466 pounds sterling), depending on experience.
Applicants should contact Rick Morgan:
Email: [EMAIL PROTECTED]
Phone: +44 191 3743654
Address: Department of Computer Science
Science Laboratories
South Road
Durham DH1 3LE
UK
The DEAR Project
The DEAR (Database EAsy Retrieval) project aims to build a sophisticated
natural language interface for a commercial database. This will be achieved
by building on four pieces of technology :-
- LOLITA (Large-scale, Object-based, Linguistic Interactor, Translator
and Analyser) a sophisticated natural-language processor which is the result
of a eight-year research programme at the University of Durham. LOLITA has
been designed as a general purpose NL base system upon which specific
applications can be built. It is written in Haskell and, at over 40,000
lines, is one of the world's largest lazy functional programs.
- GHC, the Glasgow Haskell Compiler, aims to provide Haskell
programmers with a state-of-the art compiler and to serve as the driving
force behind, and test-bed for, research in language design and
implementation technology. GHC is intended to compile code for parallel
machines.
- A large commercial database provided by the project's `heavy uncle'
Rolls Royce and Associates.
- A commercial multiprocessor (Sun SparcCentre 2000)
The system will take NL queries, translate them into SQL, pass them on to
the (unmodified) underlying database and display the result.
The project aims to speed the analysis by executing LOLITA on a
multiprocessor using a parallel implementation of the functional language
Haskell. The significance of the project goes well beyond a single Rolls
Royce database. Firstly, since the interface to the database is achieved
through SQL, the system should be usable with almost any database. Secondly
the work will be applicable to knowledge bases other than relational DBMSs.
Finally, the project will be a convincing (and demanding) application of
parallel Haskell which will identify the important issues in large-scale
programming in the purely functional paradigm.
