ERIC Number: ED022148
Record Type: RIE
Publication Date: 1968-Jul
Reference Count: 0
A Procedure for Morphological Analysis.
Chapin, Paul G.; Norton, Lewis M.
A procedure, designated "MORPH," has been developed for the automatic morphological analysis of complex English words. Each word is reduced to a stem in canonical or dictionary form, plus affixes, inflectional and derivational, represented as morphemes or as syntactic features of the stem. The procedure includes the task of analyzing as many nested levels of affixation as a word may contain so that each morpheme of the input string will have a distinct representation. The overall strategy includes--(a) a set of analysis rules, (b) a set of morpheme-combinatorial rules, and (c) a set of redundancy rules. Spurious analyses are dealt with by certain modifications in these sets of rules. The procedure has been implemented on the IBM System 360 in TREET as part of an experimental text-processing system in which it provides the input to a transformational syntactic analysis procedure. The present sets of rules are quite small, removing only the most common inflectional suffixes (-s, -ed, -ing). Using these rules on a small vocabulary, MORPH has performed accurately at an average speed of 0.7 seconds per word. It was discovered in experimenting with this program that proper names require special treatment to avoid spurious analysis and that there is far greater homography with English prefixes than with suffixes, to the extent that the value of prefixational analysis is questionable. (JD)
Descriptors: Computational Linguistics, Computer Programs, English, Language Patterns, Morphemes, Morphology (Languages), Plurals, Spelling, Structural Analysis, Suffixes, Surface Structure, Verbs
The MITRE Corporation, Box 208, Bedford, Massachusetts 01730 (Information System Language Studies No. 18).
Publication Type: N/A
Education Level: N/A
Authoring Institution: N/A
Identifiers: TREET Programming System