On correction of semantic errors in natural language texts with a dictionary of literal paronyms

Abstract

Due to the open nature of the Web, search engines must include means of meaningful processing of incorrect texts, including automatic error detection and correction. One of wide-spread types of errors in Internet texts are malapropisms, i.e., semantic errors replacing a word by another existing word similar in letter composition and/or sound but semantically incompatible with the context. Methods for detection and correction of malapropisms have been proposed recently. Any such method relies on a generator of correction candidates-paronyms, i.e., real words similar to the suspicious one encountered in the text and having the same grammatical properties. Literal paronyms are words at the distant of few editing operations from a given word. We argue that a dictionary of literal paronyms should be compiled beforehand and that its units should be grammeme names. For Spanish, such grammemes are (1) singulars and plurals of nouns; (2) adjectives plus participles; (3) verbs in infinitive; (4) gerunds plus adverbs; (5) personal verb forms. Basing on existing Spanish electronic dictionaries, we have compiled a dictionary of one-letter-distant literal paronyms. The size of the dictionary is few tens thousand entries, an entry averaging approximately three paronyms. We calculate the gain in number of candidate search operations achievable through the proposed dictionary and give illustrative examples of correcting one-letter malapropisms using our dictionary.