Specific' words as a reason for not detecting errors by computerised spell check

This article describes problems of computerised spell checking of Russian-language texts. Microsoft Word ® 2016™ text editor's built-in spell checking engine is being investigated and evaluated. It is worth mentioning that many flaws and gaps of previous MS Word versions have been patched in MS Word 2016. Nevertheless, computerised analysis of word concord - in phrases and in standalone word combinations - raises even more questions, especially when compared with previous Orfo™-based spellers. Even detection of spelling errors (as a most developed analysis area) and prompting of possible corrections are still far from being perfect. When speller underlines a word that is very recently borrowed from foreign languages (marking as not recognised), users may think this lexical item is just not present in the system vocabulary. Users may overlook their misprints in many cases, especially if there is a compound or just a long word. The article contains multiple examples of words quite frequently used in modern phrases and not known by MS Word 2016 system vocabulary, which should not be detected as mistakes but skipped without remarks. At the same time, there is no reason to keep rare and low-frequent short lexical units which coincide with beginnings and endings of more commonly used words in the system vocabulary, because they may appear when a word is unintentionally split by space. The author provides a list of such short lexical units in the article, as well as a list of rare words considered by MS Word as correct despite of a significant chance of an error in writing more commonly used words. The article contains examples of specially constructed phrases with errors: interchange of letters in a word, hyphaeresis or gemination, word split or concatenation. All such words resulting from errors are present or generated within the system vocabulary. Word forms do not concord here; however, MS Word 2016 is unable to detect syntax errors of this type. Similar phrases can also be used for testing spell checkers of other MS Word versions, not only previous but also newer ones. It is advisable to remove some 'specific' rare words from internal system vocabularies or deactivate them for the time being, until a spell checker will be more informative about the contextual areas where the words can be used. The author also proposes to enrich the spell checker's system vocabulary with lists of the most probable misspellings of words and phrases along with correct options to be chosen. Evidently, the most popular spelling errors that are omitted by MS Word 2016 are not infinite and thus may be collected. A collection of such common errors paired with normative writing should contribute to more efficient work of computerised spell checking systems.

Keywords

Authors

References