Detecting and Correcting Morpho-syntactic Errors in Real Texts

Detecting and Correcting Morpho-syntactic Errors in Real Texts Theo Vosse* Nijmegen Institute for Cognition and Information University of Nijmegen and Cognitive Technology Foundation P.O. Box 9104 6500 HE Nijmegen, The Netherlands e-mail: vosse@nici.kun.nl Abstract This paper presents a system which detects and corrects morpho-syntactic errors in Dutch texts. It includes a spelling corrector and a shift-reduce parser for Augmented Context-free Grammars. The spelling corrector is based on trigram and triphone analysis. The parser is an extension of the well-known Tomita algorithm (Tomita, 1986). The parser interacts with the spelling corrector and handles certain types of structural errors. Both modules have been integrated with a compound analyzer and a dictionary of 275,000 word forms into a program for stand-alone proof-reading of Dutch texts on a large scale. The system is in its final testing phase and will be commercially available as from 1992. 1. Introduction One of the most widely used applications of natural language processing is spell, grammar and style checking. Although most probably semantic analysis is required to obtain entirely satisfactory results, it is never used -- for obvious reasons. Even worse, most language checkers today even restrain from syntactic analysis. This denies them the possibility to find morpho-syntactic errors, which form a large and fre- quently occurring class of spelling errors. One of the best known systems for English, which does perform syntactic analysis, is Critique (Richardson, 1988). In order to detect and correct morpho-syntactic errors a system needs (I) modules for word-level spell checking and correction, (2) a parser which contains a comprehensive grammar and an efficient parsing algorithm, and (3) a mechanism to detect and correct grammatical errors as well as to assist in correcting spelling errors. I will first define the domain of morpho-syntactic errors and motivate the *The author's current address is: Experimental Psychology Unit, Leiden University, P.O. Box 9555, 2300 RB Leiden, The Netherlands. need for a parser. After a brief overview of the sys- tem and a discussion of the word-level modules, I will describe the grammar formalism, the parser, its mechanism for error detection, and a pre-processor for word lattices. Finally, after looking at the integra- tion of the modules and at some useful heuristics, I will give a summary of the results obtained by a non-interactive Dutch grammar-driven spell checker. 2. Morpho-syntactic Errors This paper is concerned with three types of errors: typographical errors (typing errors or OCR scanning errors), orthographical errors (erroneous translitera- tions of phonemes to graphemes) and, most impor- tantly, morpho-syntactic errors (resulting from mis- application of morphological inflection and syntactic rules). Simple spell checkers are only able to spot er- rors leading to non-words; errors involving legally spelled words go unnoticed. These morpho-syntacfic errors occur quite frequently in Dutch texts, though, and are considered serious because they are seen as resulting from insufficient language competence rather than from incidental mistakes, such as typo- graphical errors. Therefore they constitute an inter- esting area for grammar checking in office and lan- guage teaching applications. I will now present …

Content truncated to 500 words