Two Simple Prediction Algorithms to Facilitate Text Production

TWO SIMPLE PREDICTION ALGORITHMS TO FACILITATE TEXT PRODUCTION Lois Boggess P.O. Drawer CS Mississippi State University Mississippi State, MS 39762 ABSTRACT Several simple prediction schemes are presented for systems intended to facili- tate text production for handicapped individuals. The schemes are based on single-subject language models, where the system is self-adapting to the past language use of the subject. Sentence position, the immediately preceding one or two words, and initial letters of the desired word are cues which may be used by the systems. INTRODUCTION For some years we have been investi- gating the use of a sizeable sample of a particular individual's language habits in predicting future language use for that individual. The research has taken two directions. One of these, the HWYE (Hear What You Expect) system, builds a large lan- guage model of the past language history of the individual, with special emphasis on the most frequent words of that person, and the result is used in speech recognition. In studying the language model developed by the HWYE system, several simple predictive schemes were noted which are capable of anticipating, during the generation of a sentence, a small set of words from which the next desired word can be selected. The two schemes described here are used for text generation (not speech recognition) in a format that could be of use to a physi- cally handicapped person; hence the schemes have no right context available. One of the schemes does use left context, and the other uses only sentence position as "context'. Both are implemented on IBM-PC systems with minimal memory requirements. MOTIVATION One hundred English words account for 47 per cent of the Brown corpus (about one million words of American English text taken from a wide range of sources). It seems reasonable to suppose that a single individual might in fact require fewer words to account for a large proportion of generated text. From our work on the HWYE system it was known that 75 words accounted for half of all the text of Vanity Fair, a 300,000 word Victorian English novel by Thackeray (which incorporated a fairly involved syntax, much embedded quotation, and passages in dialect and in French) [English and Boggess, 1986]. We further found that 50 words accounted for half of all the verbiage in a 20,000 word set of sentences provided by an individual who collaborated with us. This latter corpus, called the Sherri data, is a set of texts provided by a speech-handicapped individual who uses a typewriter to communicate, even with her family; it is conversational in nature, as can be seen in Figure 1. Most of the work reported in this paper gives special attention to the set of words required to account for half of all the verbiage of a given individual. We refer to this set as the set of high-frequency words. 33 You said something about a magazine that <namel> had about computers that I might like to borrow. I would some time. I think …

Content truncated to 500 words