Context, Usage, and Categorization
Context is the way in which a word is conceptualized within a sentence - not by the listener, and not necessarily by the speaker, but by the words that accompany a word. Other words create the "setting" in which a word is used, and the setting imposes meaning on a word. For example, the following sentences impose very different meanings on the word moon:
"The moon shone brilliantly over the horizon."
"Johnny will moon everyone at the end of the show."
Just as the speaker may conceptualize a word one way and the hearer another, the sentence itself imposes a concept on the word.
Word usage not only determines how a word becomes defined in the dictionary, usage is a continuum that continually migrates word meanings to altered and new meanings through misuse (misperceptions of meaning) and metaphor. Everyone who uses a word may not understand the word in the same way. Jump into any conversation and ask for a definition of a word recently used, and you will get multiple meanings.
For example, if everyone on the East Coast uses the word logoplastycism to mean, "to take words and combine them into one new word," and people on the West Coast hear the word without a definition and come to believe it means, "to extract individual letters into a new word," then the word has gained a new definition through misperception and usage. If people in the Midwest begin thinking of logoplastycism as an eloquent game, using the word metaphorically to mean in an abstract sense, "Origami with words," and then they use it repeatedly in that way, as a result through metaphor and usage the word acquires yet another definition.
Words also migrate into other areas. For example, as science and technology present us with new things, everyday words begin to describe these new things. For example, the word window now describes a window on your computer - that is, the visual display of an application.
So while a word may acquire a definition, the word is not static. It is a living thing which grows with the experience of a culture. It participates in the culture's experience as a vehicle for meaning. It not only transfers the original meaning from person to person through usage based on experience, it is the basis for metaphors for new abstract meaning to describe new experience, and is the basis for meaning migration to describe new things through experience that is familiar (patterns).
Have you ever had trouble finding your reading glasses, and then realize you misplaced them on top of your head? If you think it is difficult to find misplaced glasses, try finding glasses in the Yellow Pages. You won't find them under glasses. Look up eye glasses and you only get a referent to optical goods. Optical goods lists all kinds of places where you can get glasses, contacts, and eye exams. But if you are actually wanting an eye exam, you may be misled by the limited number of entries in this section.
If you look up optometrist because you want an eye exam, you will find many entries, but not be referred to optical goods, even though both categories cover roughly the same territory. It is often faster to call 411 than to stumble through the sometimes perplexing way in which the Yellow Pages creators have chosen to categorize things.
Because people classify things according to their experience, categorization fails as a way to create common categories by which people can find things. At best categorization is done by "logic" or by tabulating popular choices which in turn are individually dictated by experiences that are common to many. Regardless of which category you put something in, many people won't be able to find it. Their experience - their way of thinking - is different. Radio means "entertaining music" to one while to another it means "electronic equipment and electromagnetic waves."
The problem with classification is demonstrated very well in computer help-file design. Computer help files, whether printed or online, are supposed to provide a usable source of information about programs. When users search through the contents, they should be able to find the information they need quickly. But tables of contents are often created for leading people through a step by step introductory approach. Is this what the user wants or is it the experts perception of what the TOC should do?
When a program user arrives at the contents with a technical question, the table may have several classifications that are candidates for providing answers, with no clear indication which is the right one. If a user can't quickly find the information, he gets frustrated, and the help-file becomes a tool of torture which makes him dissatisfied with the application.
Indexes are more helpful, but present the problem of endless lists, and even indexes require some classification. The Find function is more helpful, but even this often lists relevant and irrelevant topics together, and sometimes under multiple obscure classifications.
Context sensitive help is even more helpful, since it relates directly to the task at hand, but it often doesn't give an overview so that the user can understand what he needs to do. And context sensitive help may not be available at the time the question arises. For example if the user needs to change a setting to make the program work properly, but has no idea where to go to do that, context sensitive help is no help.
Add to these the problem that help files are written by those with expert experience for those with minimal experience. Those with minimal experience can only guess at what the experts know and how they classify things. To make it all even more vexing, everyone has a different favored route to finding information in a help file: contents, index, find. There is no right place for everyone.
The enormity of the problem with finding information is demonstrated by the Internet. Internet search engines scan all available Web pages for information and record referential information within their systems. When a visitor searches on a word or words, the search engine quickly looks at its own references and provides those millions of possible referents to Web pages, ordered by what the search engine believes is the most relevant.
For the search engine, radio means every reference to radio - it has no idea whether you mean electromagnetic waves or entertainment. Even if you search for the pair of words radio waves, it doesn't know if you mean a rock group, a cultural symbol, or an electromagnetic disturbance.
The search engine doesn't know a number of things that it should know, about the user's word meaning and classification, to be effective at doing the search. It doesn't have a category in which to limit the search, or a word definition, nor does it know the context in which the word is being used, or the context of the referent Web pages.
Just as importantly, the search engine doesn't have a brain packed with experience. Artificial intelligence currently can infer relevance from word associations (a surface reflection of context) and even word prevalence, but it can't infer relevance from experience. Because of this, natural language query applications often give bizarre results that make them look inept and useless. With the amount of information on the Internet doubling every couple of years, the problem is growing rapidly.
Stevan Harnad has addressed this issue, with a possible resolution through grounding symbols in sensorimotor experience, in his 2002 paper, Symbol grounding and the origin of language
What might work, however, is relevance engines that create "ontologies" that show relationships between words. In a sense, this classifies words - not by human categorization, but independently of individual differences and in such a way that the word is located through its relationships with other words. How these word relationships are defined, and I believe the ultimate success of the system, depends on the theoretical approach used as a basis for the system. Language relationships and experiential relationships may be different things.
The next evolution in search technology could involve asking for a category of experience to indicate word meaning, however this would require that search indexes of collected material also be categorized by categories of experience.11>