OpenCyc.org HomepageCyc-NL Documentation

E-Mail Comments to: opencyc-doc@cyc.com
Last Update: 03/28/2002
Copyright© 1996-2002 Cycorp. All rights reserved.

Return to Table of Contents

This is an introduction to the Dictionary Assistant, which allows users to add lexical information to Cyc for natural language parsing and generation.


The Dictionary Assistant: Walkthrough Examples

The Dictionary Assistant allows a user to hook natural language terms up with Cyc concepts so that those words can then be used to refer to those concepts. In this way, Cyc's vast store of knowledge is accessed via natural language. When a natural language, say English, word is hooked up with a concept, we say that the concept has been lexified. You use the Dictionary Assistant to "lexify" Cyc concepts.

To launch the Dictionary Assistant, look for the link labelled [Lexify] in the upper left-hand frame of the browser, when you are viewing a constant.

Cyc will first prompt you to enter the concept about which you want to add lexical information. As a default, the constant you were browsing will appear in the typein box.

Suppose the term you want to add new lexical information to is Dog; that is, suppose you want to refer to the concept Dog in a way currently unfamiliar to Cyc. For our example we'll add the English word "whelp" to Cyc's concept Dog so that in the future whenever anyone says "whelp" to Cyc, it will present Dog as a possible interpretation.

First, type "dog" into the "Term" box, and type "whelp" into the "Phrase" box.

On the next screen, Cyc asks whether the phrase "whelp" is a name. Since "whelp" is a regular noun, not a proper name like "Fred Jones" or "Dallas" or "the FBI", you would hit the "No" button.

Next, Cyc will ask you to choose the usage of your new word that sounds best. In this case, "whelp" sounds fine in both the first and third contexts. However, since we want to enter a noun that refers to dog, and not a verb meaning to have puppies, you would select the first option.

The next screen checks to see whether "whelp" can be pluralized. You would click "Yes" to this question. Finally, you are asked whether "whelp" is the most common way of referring to the concept Dog. In this case, the answer would be No, since terms like "dog" are much more common.

Now you can go ahead and type in "whelp" whenever you want to refer to the concept Dog. For example, typing "whelp" into the constant search box in the top of the browser pane and hitting Enter will bring up the constant Dog.

Sometimes you will want to add new lexical information to relations rather than to individuals or collections, as we just did above, and this process is slightly different.

Lexifying predicates or functions is a little trickier than lexifying concepts such as Dog or individuals such as FredJones, because whereas the latter can be referred to with words or phrases, the former need some additional arguments before they can be uniquely picked out from a group of other relations. For example, the term "hit" does not help us pick out the relation between an individual hitting and the thing hit, but "X hits Y" does because here we have the verb which represents the action and we also have positions for the two arguments, the hitter and the hittee. So in lexifying relations, we will need to provide Cyc with examples of sentences which represent how the various arguments are related to one another.

Example

Let's use as our example the relation of liking between two people, where one person likes and the other is liked. For our example, we will use the Cyc predicate likesAsFriend. Let's assume that we want to be able to translate things like "Fred fancies Sally" using the predicate likesAsFriend.

Bring up likesAsFriend in your browser, and click on [Lexify]. In the "Phrase" box, type "fancy." Now, you will be shown any other concepts that the word "fancy" already refers to. Click "OK", and on the next page, you will be asked for an example sentence to illustrate the use of "fancy" you are entering. A good example sentence here would be "John fancies Martha"; type this into the box.

On the following pages, you will be asked to map terms in your example sentence onto argument positions of the predicate likesAsFriend. First, you would select "John" as argument 1; then, select "Martha" as argument 2. That is all you need to do. Cyc will make the appropriate lexical assertions.


Go to Top