For Posterity!
A copy of my 1994 Computer Science Final Year Project
In this project I demonstrate the way computers can be used to predict the intention of a user based on knowledge of the user’s previous behaviour when using the system and the context of the current use. The idea of this prediction is the completion of a word or sentence (although as I will discuss these techniques are applicable across many areas of computer use) in such a way that the user has to spend less time communicating with the computer system or reduce errors.
This prediction, based on partial information of the whole phrase being presented, is some- thing that people do all the time, often without realizing it. Take, for example, the following phrase: “I went to the st...” If we look at this sentence without any information about grammar or the context of the sentence it would be very difficult to guess what the full phrase should be. If rules of grammar and syntax are added to aid with the prediction the final word can be nar- rowed down to be a choice from place nouns starting with “st.” Even with this information one cannot give a very good guess as to where the speaker went without some idea of context. In this case if the speaker was relating a story of a train journey the chance that the complete word is “station” would be greater than for “stables.”
On any computer system, the greatest bottleneck for the system is the user input - relying on the speed of the user’s communication to dictate the total operating time of the commands. Although now rather a cliche, it is important to reduce the time spent waiting for the user input. Technological advances have led to the proportion of time spent idle whilst the user enters commands, compared to processing time between inputs to have grown, since proces- sors have increased in speed whilst user interaction has remained very much the same throughout.
In this project I intend to put forward a method for the reduction of the time the user spends communicating with the computer through the prediction of the user’s intention. New meth- ods of human-computer communication have been developed over the last decade bringing with them promises of simpler and speedier communication methods. Rather than removing the need for the ideas I develop here, I suggest that these techniques can be applied to these developments. Methods of communication such as Character and Speech Recognition both suffer from setbacks brought about through the way in which they try to build up a complete input from the individual characters or phonemes making up the whole. I suggest the princi- ples for prediction discussed in this report can be applied to these communication methods to allow prediction of the correct input when a conflict between two or more possible matches exists.
Read more in the "Gary Dissertation.PDF"