[Ltg] LTG Seminar [Rolf Schwitter and Robert Dale, 2008-09-08, E6A 202, 11am]

[Ilya Anisimoff]

LTG Seminar
 - see: http://www.clt.mq.edu.au/Events/Seminars.html

 Monday, 9th September, 2008, 11am
 Macquarie University, E6A, Room 202
----

Title: Creating and Querying Linguistically Motivated Ontologies
Speaker: Rolf Schwitter

I will do a dry-run of a talk that I have to give at a KR 2008
workshop in Sydney on September 17. The story goes as follows:
This paper argues that a formal ontology (in our case a description
logic knowledge base) should be created in a linguistically motivated
way so that it can be queried easily by non-specialists. This can best
be achieved by using a strict naming convention that is based on those
linguistic expressions that occur in the application domain for which
the ontology will be created. We will see that ABox and TBox
statements that closely follow this naming convention can be written
directly in a controlled natural language and that the same controlled
natural language can be used to query the description logic knowledge
base. Both ABox and TBox statements written in controlled natural
language are translated automatically into the Knowledge
Representation System Specification (KRSS) syntax and questions are
translated into RacerPro's new query language nRQL and answered over
the description logic knowledge base. Using a controlled natural
language as a high-level interface language abstracts away from any
formal notation and allows for true collaboration between humans and
machines.

---

Title: Architecture and Representations in the Thinking Head
Speaker: Robert Dale

In this talk I'll sketch my current thinking in regards to the
architecture of the dialog management system we aim to build for the
Thinking Head, and talk about the representations of different kinds
of knowledge that this will involve.  There are three key desiderata
that bear on the design decisions here: (1) we want to maintain a
symmetry between the stages of language analysis and language
generation (something that is not usually the case in most systems);
(2) we want to allow for the use of canned expressions and templates
as well as 'reasoned-from-first-principles' representations at various
levels; and (3) we want to enable incremental analysis and generation
processes. We're a long way from sorting out all the details here, so
the aim of this session is to sketch current thinking and get
feedback.