A Brief List of the Applications

Ten applications are presented in this section in fairly specific, concrete form; i.e., a thumbnail ``scenario of use.'' In later sections, when we treat each application in turn, we will suggest additional tasks in that same category, and explain how and why Cyc could be used.

(1) Advice services
Over a network (e.g., CompuServe), offer a service that helps people select which type of new car to buy.

(2) Directed marketing
Use a person's buying history to infer their hobbies, interests, occupation, physical needs and preferences, etc. From that model, decide which products to try to sell them, and what ``argument'' to use to convince them they should buy the product.

(3) Online brokering of goods/services
Using a common vocabulary and shared fundamental knowledge about transactions, enable a large set of buyers/sellers to find each other and negotiate deals - online and (semi-)automatically.

(4) Data base cleaning
Relate a data base's fields and keys to Cyc predicates and terms, and use Cyc's common sense constraints to detect possible errors and inconsistencies in that data base.

(5) Data base integration
Use that same sort of ``articulation'' approach to have several heterogeneous data bases all relate their contents to one central knowledge base. Use this to (1) detect and resolve contradictions among data bases, and (2) handle queries that require accessing - and integrating the results from - multiple DB's.

(6) Corporate knowledge assets management
Represent a company's know-how, policies, important documents, programs, and data bases - all in a form that a program can effectively reason with. Use as an online policy manual, or as a sort of ``smart Yellow Pages'' for employees.

(7) Smart spreadsheets
Explain the ``meaning'' of the rows and columns, by tying them to a large corpus of general knowledge about occupations, human capabilities, goals, household objects, etc. Highlight abnormal (though not illegal) values, and suggest places where there seems to be a missing equation or constraint.

(8) Smarter interfaces
Use an online model of the ``meaning'' of what the application program is doing, what the user's goals are, what constraints they're under (e.g., deadlines), etc., to adjust the interface. E.g., reorder questions on a `form'; rearrange windows; highlight or cross out various parts; guess at values to fill in certain `blanks;' and so on.

(9) Machine translation of technical documents
Provide a set of flexible templates into which 95%of some class of technical documents (e.g., installation instructions for a laser printer) can be captured. Then, using currently-available natural language software, process such documents, converting them into a set of filled template instances. Conversely, the `first writing' of a manual could itself be done by filling out such templates. Translations into different languages could then be produced by composing the translations of each template, and each (disambiguated) term used. Difficult portions get flagged for checking by hand.

(10) Enriched `Artificial Reality' (AR)
Simulated `worlds' in which physical objects behave ``realistically,'' and machine-run personae behave ``intelligently'' - particularly when one interacts with those objects and personae in ways that were not specifically preconceived. In addition, complex, changing images must somehow be transmitted in real time, among a large set of interacting human users, at compression ratios exceeding one million to one.

In the next ten sections, we focus on each of the applications listed above. We go into more detail of what the application would be, why and how Cyc would be useful, list some alternative specific usage scenarios, and try to characterize what the ``general opportunity'' really is.


Tue Oct 11 15:19:00 CDT 1994