[ Up to Morphology and the analysis of word forms.]
A morpheme is the smallest unit of meaning. Below the mophological level there are just sounds (or letters).
Examples: 'un-' 'comfort' '-able'
Note that not all languages have 'word paradigms'. Languages that use the roman alphabet enforce the word paradigm by specifically emphasizing word boundaries. Chinese is written as a steam of morphemes, and there is less than perfect aggreement among native speakers how sentences should be partitioned into 'words'.
Bound morphemes aren't found in isolation.
They usually have no parts of speech.
Examples: 'un-, '-ing', 'ab-' '-late' 'amok?'
Prefixes/suffixes go on either end of the root.
Infixes get put in-between: 'absofreakinglutely' (some languages make more use of this than English)
Roots are usually free, but may be bound, e.g. -late, -ceive, umbrage?
Derivational affixes operate directly on the meaning of the root,, eg:un- -able -ate. They 'operate on' their roots to build new meanings.
Inflectional affixes serve a grammatical function, eg: '-ed', '-ing', '-s', '-en'. Every language requires the speaker to track certain features of the things and events they are talking about. In English the speaker is required to use inflectional markers to indicate person, plurality, tense, etc.
Also known as a distinction between content and function words.
Most function words come from a closed class of words whose usages contrast one another, and whose 'meaning' may be difficult to pin down, eg: 'is', 'in' 'of'...
The illustration at right maps the frequency of Chinese characters in an 800,000-word corpus to the log of their frequencies. Generally speaking, the 'function words' fall into the almost-vertical column on the far-left; content words vary in frequency, filling in the rest of the curve.
[ Up to Morphology and the analysis of word forms.]
[ Up to How computers relate to morphemes.]
Identify newly encountered words. Many 'new' words are coined by combining productive morphemes.
Extract roots for comparison of content. If you're doing a search for web pages, and your query includes the word 'idea', you probably also want to include pages with the word 'ideas' as well.
Determine parts of speech.
Save (memory) space. You can use morphological rules to generate the various forms of, say, regular verbs, which saves you the space taken by duplicating each form seperately. In cases where substantial semantic knowledge is being encoded, it would be silly to have two completely separate entries for 'idea' and 'ideas'. What you need instead is a representation that allows you to express singular vs. plural.
[ Up to Morphology and the analysis of word forms.]
Trivial example: '-ing' (problem: 'earring')
Phonological rules also come into play.
Less trivial: 'in-' (-> 'im' before bilabials)
Less trivial still: 'drink, drank, drunk'.
In the case of -er and -est, there are rules which are somewhat systematic, but some words can take these suffixes, and other words cannot (you have to use more/most). Even native speakers of English differ in their perceptions on this issue.
[ Up to Typically straightforward rules can be used to good effect.]
Some of the rules that describe how morphemes combine can operate virtually without restriction, eg: '-ing'
Others only operate on a very narrow class of other morphemes, eg: '-en' to indicate plural.
This is related to the idea of 'collocation', e.g. why can't 'strong' and 'powerful' be used interchangeably?