[1] From: "David L. Gants" <dgants@parallel.park.uga.edu> (192)
Subject: Job Advertisement
[2] From: "David L. Gants" <dgants@parallel.park.uga.edu> (23)
Subject: UCSD Cognitive Science - Junior Faculty Position
--[1]------------------------------------------------------------------
Date: Thu, 6 Nov 1997 08:24:49 -0500 (EST)
From: "David L. Gants" <dgants@parallel.park.uga.edu>
Subject: Job Advertisement
>> From: Steel S W D <sam@csc18.essex.ac.uk>
================================================================
Jobs offered
================================================================
Senior Research Officer
Spoken language dialogue access to partially structured data
The University of Essex has a vacancy for a senior research officer
to work on a project to develop spoken language dialogue access to
partially structured data. The project will develop a system that can
take part in spoken dialogue. It will use artificial intelligence,
natural language processing and knowledge representation techniques to
extract useful information from partially structured data.
The project is based in the Computer Science department of the
University of Essex. This post is available from January 1998. It is
funded by a telecommunications company.
The successful applicant will be expected to take part in the design of
the system, to carry out the implementation, and take part in
evaluations and demonstrations.
The position is especially suitable for those intending to make a
career in artificial intelligence. Applicants should have a solid
background in computer science and artificial intelligence: for
instance, a good honours degree or MSc. Experience of programming in C
is an advantage.
Appointment, initially for one year with possibility of continuation for
two more years, will be made on the Grade 1A scale for auxiliary
(research) staff, 15,159-22,785 pounds per annum. Starting salary will
depend on previous experience.
Application details may be obtained by phoning Colchester (01206) 872462
(24 hours), quoting reference R/227, or by writing to the Personnel
Section, University of Essex, Wivenhoe Park, Colchester CO4 3SQ, or by
email to
staffing@essex.ac.uk
Closing date: 18 November 1997
You can also send questions to Dr Sam Steel, sam@essex.ac.uk
================================================================
Sam Steel
----------------------------------------------------------------
FURTHER PARTICULARS
Department of Computer Science
Senior Research Officer grade IA
Job title and grade: Senior Research Officer
Department: Computer Science
Responsible to: Head of Computer Science Department
Reports on day to day basis to: Dr Steel, Ms De Roeck, Prof Turner, Dr
Scott
Purpose of job:
Research and development on a project to enable spoken language dialogue
access to partially structured data
Duties of post:
The duties of a research officer are naturally wide-ranging. In
general, they are to advance the construction of the system by design,
implementation and testing as far as possible in the time available.
The post-holder may expect continuous direction and advice from the
principal investigators, but they will also have wide scope for
initiative and invention.
The successful applicant will be expected to take part in the design of
the system, to carry out the implementation, and take part in
evaluations and demonstrations.
They will also be responsible for preparation, revision and input of the
data on which the system will operate.
They may have to assist in presenting descriptions of the system both
internally and externally, and both in writing and orally.
They will perform any other such duties as may be assigned from time to
time by the Head of Department or his/her nominee
These duties are a guide to the work that the postholder may initially
be required to undertake. They may be changed from time to time to meet
changing circumstances and do not form part of the contract of
employment'
Essential experience, skills and attributes of the postholder
Applicants should have a solid background in computer science and
artificial intelligence: for instance, a good honours degree or MSc.
However, candidate with relevant work experience in computing or
artificial intelligence are also welcome.
They will show ability to think constructively and inventively about
problems.
Desirable experience, skills and attributes of the postholder
Experience of any of these is an advantage:
programming in C
programming in Prolog
speech and signal processing
natural language processing; linguistics
formal methods in computer science; logic
However, it may well be the case that applicants have skills which they
feel are relevant to the project but which are not listed here. They
should mention these in their application.
----------------------------------------------------------------
Other relevant information
Aims of the project
The goal is a wholly artificial system that can respond sensibly to
spoken enquiries about partially structured data. The most interesting
and difficult aspects are not the voice input, but the control of
dialogue and the retrieval of information from the data. It would build
on extant ``spoken language dialogue systems'' (SLDS), but would be
unusual in its attempt to deal with a large pre-existing body of
knowledge.
The assistant must work, even if only crudely, after one year. If the
data was which organizations sold what goods and services, it should
then be able to participate in dialogues such as this,
User: I want to buy some teaspoons.
System: What are teaspoons?
User: Teaspoons are small spoons.
System: What are spoons?
User: Spoons are used for eating.
System: Do you buy spoons at food shops?
User: No.
System: Do you buy spoons at catering suppliers?
User: Yes.
System: [Lists catering suppliers.]
Do you buy spoons at restaurants?
User: That's enough.
System: Goodbye.
SLDSs already have a fairly standard design. It contains:
A SPEECH RECOGNIZER to convert voice input into strings of words;
A SPEECH-TO-MESSAGE CONVERTER to try to recognize possible messages of
known form in the string of words, even if that loses information;
A DIALOGUE MANAGER to attempts to see user input as a step in a known
class of dialogues: if this succeeds, the next move will be made by the
system;
A DATABASE AND ``INFERENCE ENGINE'' because the questions will often not
be instantly answerable from the database.
The standard SLDS is inadequate for this project because of
BRITTLENESS OF DIALOGUE (because at any point which responses are
allowable is determined by the system) and
BRITTLENESS OF DOMAIN KNOWLEDGE (when the system could know that
teaspoons were cutlery, and that house furnishings were sold at J
Smith's, but not know that J Smith's might sell teaspoons because it did
not know that cutlery was a sort of house furnishing).
We will reduce brittleness by trading away correctness for coverage. To
help do this, the system should be
DECLARATIVE, with a clear separation of the knowledge of dialogue and of
the domain from the machinery that deploys that knowledge;
SHALLOW, so that the machinery can run even with knowledge that is very
incomplete and very shallow;
CORRECTABLE, so that improving the system's knowledge is part of the way
it is used, not just part of the way it was constructed;
ERRATIC BUT EXPOSED, where by accepting a high (though declining) error
rate in the system's understanding and responses, compensated for by
providing a visible model of its model of the dialogue.
The dialogue scripts will rely on interrogating the database. Though
those enquiries are simple, the information by itself will not be
enough. For instance, any adult can guess that ironmongers or home
furnishing shops are good places to try. There are four ways of
approaching that.
DEEP KNOWLEDGE where the database would contain facts where the
primitives are a privileged a priori set of ``concepts''.
SHALLOW KNOWLEDGE where the primitive types and most of the primitive
relations are lexemes --- word stems. Such a representation is not
far from
the original text, and will relatively little translation.
TEXTUAL RETRIEVAL which is simple, easy and fast; and it misses very
little.
``ASK THE USER'', because ``What sort of thing are TEASPOONS?'' or
``What sort of trader sells TEASPOONS?'' are reasonable questions.
We assume that we start with the basic information in machine-readable
form, perhaps as a simple table of the form
goods/service offered, trader, address, phone number
but which could also contain the cross-referencing of categories and
sections of free text. We must translate this into a database, rapidly
and shallowly.
The system's knowledge will be assumed to be permanently incomplete.
When the system fails in an irritating way --- that is, when the user
feels sure that the original information contains the answer ---, the
whole enquiry will be logged for a maintainer to look at later.
Failures of use of knowledge will not be seen as bugs needing debugging,
but as ignorance needing education.
Robustness can only be achieved at the cost of making mistakes. We will
accept a high error rate, but compensate by making the system explain
its responses to the user, and able to accept the user's corrections.
This will use a subsidiary, conventional, channel.
It is important that the project provide demonstrations of what it is
doing at every stage of its development.
The principal investigators are Dr Sam Steel, Ms Anne De Roeck, Dr Paul
Scott and Professor Ray Turner.
----------------------------------------------------------------
Appointments, initially for one year with possibility of continuation
for two more years, will be made on the Grade 1A scale for auxiliary
(research) staff, 15,159-22,785 pounds per annum. Starting salary will
depend on previous experience.
The appointment will be initially for one year with possibility of
continuation for two more years, depending on the performance of the
person appointed and the continuation of the project into a third year
by the funder.
Applicants should note that the University operates a No Smoking policy.
A completed application cover sheet, curriculum vitae, comprising a
statement of the candidate's experience and qualifications, and a
covering letter should be sent to
The Personnel Section
University of Essex
Wivenhoe Park
Colchester
Essex CO4 3SQ
to arrive not later than 18 November 1997
Please quote job reference number R/227 in your application.
--[2]------------------------------------------------------------------
Date: Thu, 6 Nov 1997 08:27:06 -0500 (EST)
From: "David L. Gants" <dgants@parallel.park.uga.edu>
Subject: UCSD Cognitive Science - Junior Faculty Position
From: elman@crl.ucsd.edu (Jeff Elman)
FACULTY POSITION
IN COGNITIVE SCIENCE
UNIVERSITY OF CALIFORNIA, SAN DIEGO
The Department of Cognitive Science at the University of California, San
Diego invites applications for a faculty position at the assistant
professor level (tenure-track) starting July 1, 1998, the salary
commensurate with the experience of the successful applicant and based on
the UC pay scale, and subject to the availability of funds.
Applicants are sought in the area of higher cognitive function and must
have a PhD (or ABD). A broad interdisciplinary perspective and experience
with multiple methodologies will be highly valued. Postdoctoral experience
is desirable.
Women and minorities are encouraged to apply. The University of California,
San Diego is an affirmative action/equal opportunity employer. All
applications received by January 1, 1998 will receive thorough
consideration until position is filled. Candidates should include a vita,
reprints, a short letter describing their background and interests, and
names and addresses of at least three references to:
University of California, San Diego
Faculty Search Committee
Department of Cognitive Science 0515-JE
9500 Gilman Drive
La Jolla, CA 92093-0515
-------------------------------------------------------------------------
Humanist Discussion Group
Information at <http://www.kcl.ac.uk/humanities/cch/humanist/>
<http://www.princeton.edu/~mccarty/humanist/>
=========================================================================