Content development using N-gram model in custom writing style

Abstract

Amateur writers usually find it difficult and often make errors while building up content when they are doing so in a style different from their own writing style. This causes loss of interest by the readers and sometimes even misinterpretations of actual thoughts desired to be conveyed by author. This work attempts to embark upon this problem statement by ranking the best available choices of words fitting the style of writing that the author would like to adopt. Our methodology allows authors to choose from amongst default, formal and literature style of writing. Also, authors can infer words and traces from his own past writings by developing a custom corpus of his own write-ups. A spell checker and N-gram based statistical model along with a corpus based technique is proposed to achieve above objectives. Rank 4 N-gram along with backoff smoothing provided optimum results for our work. To showcase the effectiveness of this method, we have tested it on real time data and performance evaluation fetched satisfactory results.