One of the advantages of conceptual dependency is that it allows the computer to make inferences from a natural language system in the same way human beings do. The CD representation of a sentence is built using conceptual primitives and not with primitives of the words contained in the sentence. These conceptual primitives represent thoughts and the relationships between thoughts. Using conceptual dependency facilitates the use of inference rules, because many inferences are already contained in the representation itself.
There are several primitives to represent actions, the more commonly used are the following, taken from Schank [19]:
Basically each primitive contains components:
An Actor: He is the one that performs the ACT.
An ACT: Performed by the actor, done to an object.
An Object: The action is performed on it.
A Direction: The location that an ACT is directed towards.
A State: The state that an object is in, represented by scales which have numerical values. For example Health goes from -10 to 10, with 10 meaning perfect health, 0 all right and -10 dead.
Following are some examples of CD's, and their representation.
* PTRANS, the transfer of the physical location of an object, has the following representation:
(PTRANS (ACTOR NIL) (OBJECT NIL) (FROM NIL) (TO NIL))
For example the sentence: "Robot, go to the kitchen," assuming that the robot is in the living room, has the following representation:
(PTRANS (ACTOR Robot) (OBJECT Robot) (FROM living-room) (TO kitchen))
When the verb "go" is found in the sentence a ptrans structure is issued. The empty slots (NIL) need to be filled finding the missing elements in the sentence. The actor is the robot, the object is the robot, meaning that the robot is moving itself, from the living room to the kitchen.
* MTRANS is the transfer of mental information, and it has the following representation:
(MTRANS (ACTOR NIL) (OBJECT NIL) (FROM NIL) (TO NIL))
For example the sentence: "Robert told Susan about it," is represented by:
(MTRANS (ACTOR Robert) (OBJECT it) (FROM Robert's brain) (TO Susan's brain))
* ATRANS is the transfer of an abstract relationship, or object, and has the following representation:
(ATRANS (ACTOR NIL) (OBJECT NIL) (FROM NIL) (TO NIL))
For instance the phrase: "John gives the book to Mary," can be represented by the following CD:
(ATRANS (ACTOR John) (OBJECT book) (FROM John) (TO Mary))
* INGEST is the ingestion of an object by an animal, and has the following representation:
(INGEST (ACTOR NIL) (OBJECT NIL) (FROM NIL) (TO NIL))
For instance the phrase: "John drinks milk," is represented by the following CD:
(INGEST (ACTOR John) (OBJECT milk) (FROM nil) (TO John's mouth))
Here we don't know where the milk is coming from; to complete this structure we need aside information that can be obtained from the previous, or next sentence, or by context.
* PROPEL is the application of physical force to an object, and has the following representation:
(PROPEL (ACTOR NIL) (OBJECT NIL) (FROM NIL) (TO NIL))
For instance the phrase: "Mary killed the spider by throwing a shoe to it," is represented by the following CD:
(PROPEL (ACTOR Mary) (OBJECT shoe) (FROM Mary) (TO spider))
The physical state of the spider changes from a value bigger than -10 to -10. In this example the actor needs to move its arm to throw the shoe, that means, that we may infer that the action MOVE took place. MOVE is the instrumental for PROPEL. As in this example an action may require other instrumental actions to be performed.
Questions are handled by putting the symbol Q in the conceptual spot being inquired about. For example "Where is the dog," can be represented as follows:
(PTRANS (ACTOR dog) (OBJECT dog) (TO Q) (FROM dog's initial place))
and the system needs to find its place and it answers with an MTRANS, that is:
(MTRANS (ACTOR computer) (OBJECT "dog's new place") (TO inquirer) (FROM computer))
In this section we explain how to transform a natural language sentence into a structure which can be manipulated more easily. These structure facilitate the inference process, but the inference problem is still not solved, what they do is to reduce the number of verbs into a small number of ACTs, from which inferences can be made.
CDs can be used for representing simple actions. They are also well suited for representing commands or simple questions, but are not very useful for representing complex sentence. Schank and Owens [21] explain that for the following sentence it will take about two pages of CD forms: "John bet Sam fifty dollars that the Mets would win the World Series."
Some of the conceptual dependency techniques explained in this section were incorporated in our speech expert system, as is described in the next sections.