Automating rule generation for grammar checkers

In this paper, I describe several approaches to automatic or semiautomatic development of symbolic rules for grammar checkers from the information contained in corpora. The rules obtained this way are an important addition to manually-created rules that seem to dominate in rulebased checkers. However, the manual process of creation of rules is costly, time-consuming and error-prone. It seems therefore advisable to use machine-learning algorithms to create the rules automatically or semiautomatically. The results obtained seem to corroborate our initial hypothesis that symbolic machine learning algorithms can be useful for acquiring new rules for grammar checking. It turns out, however, that for practical uses, error corpora cannot be the sole source of information used in grammar checking. We suggest therefore that only by using different approaches, grammar-checkers, or more generally, computer-aided proofreading tools, will be able to cover most frequent and severe mistakes and avoid false alarms that seem to distract users. In what follows, I will show how Transformation-Based Learning (TBL) algorithms may be used to acquire rules. Before doing that, I will discuss the pros and cons of three approaches to creating rules and show the need to make use of them all in a successful grammar-checking tool. The results obtained seem to suggest that the machine-learning approach is actually fruitful, and I will point to some future work related to the reported research.