|
Our Approach to Natural
Language Understanding
ACE’s ground-breaking solutions for transforming
and repairing free-format Fax, Telex, Email messages into structured
SWIFT or any other required format for Straight Through Processing
relies on Natural Language Understanding technology. However our
approach is unique, accounting for the technology’s inherent shortcomings
to construct a value-added solution that is practical and at the
same time powerful. The natural language understanding approach
we have taken is based on a liberalized Conceptual Dependency
representation and modified Semantic Analysis we call Conceptual
Analysis. During Conceptual Analysis we perform transformation
of the natural language phrases and sentences into our conceptual
dependency-like representation using concepts conveyed in the
sentence. We believe that the best way to approach the problem
of building an intelligent program is to emulate human conceptual
mechanisms that deal with language. STP White Paper
The Conceptual Dependency
Theory
Our approach is based on the original Theory
of Conceptual Dependency proposed by Prof. Roger Schank of Yale
University, New Haven, Connecticut, USA. This is a theory of the
representation of the meaning of sentences. The basic axiom of
the theory is:
A. For any two sentences that are identical in
meaning, regardless of language, there should be only one representation.
B. Any information in a sentence that is implicit
must be made explicit in the representation of the meaning of
that sentence. (Corollary)
Principles of Natural
Language Understanding Utilized
- People, in speaking and writing, consistently leave out information
that they feel can easily be inferred by the listener or reader.
They try to be concise and therein begins the root of the problem.
In order to understand people, we have to use additional knowledge
to fill in these "gaps".
- People use a great deal of knowledge to understand. There
are two classes of knowledge that people bring to bear during
the understanding process: General Knowledge (plans) and Specific
Knowledge (scripts).
- General Knowledge enables a person to understand and interpret
another person's actions simply because the other person is
a human being with certain standard needs who lives in a world
which has certain standard methods of getting those needs fulfilled.
Specific detailed Knowledge about a situation allows us to do
less processing and wondering about frequently experienced events.
- In a closed environment, the contexts are known and hence
most of the information that is used by people while understanding
or assumed by people while communicating with an intelligent
entity is finite and known in advance. In such cases, most of
the understanding can be Script based. Scripts are intended
to account for such a specific knowledge that people have.
Advantages of Our Approach
- Language Independence - Multi-lingual approach
- Robustness - can handle ungrammatical and incomplete sentences
- Flexible and easily expandable inferencing and understanding
by using additional knowledge sources
- Knowledge sources can be modularized and hence manage and
expand intelligence.
|
