“Grammaticality judgment in a word completion task”.
Renaud, A., Shein, F., Tsang, V.
View Full Text Show AbstractIn this paper, we present findings from a human judgement task we conducted on the effectiveness of syntax filtering in a word completion task. Human participants were asked to review a series of incomplete sentences and identify which words from accompanying lists extend the expressions in a grammatically appropriate way. The accompanying word lists were generated by two word completion systems (our own plus a third-party commercial system) where the ungrammatical items were filtered out. Overall, participants agreed more, to a statistically significant degree, with the syntax-filtered systems than with baseline. However, further analysis suggests that syntax filtering alone does not necessarily improve the overall acceptability and usability of the word completion output. Given that word completion is typically employed in applications to aid writing, unlike other NLP tasks, accounting for the role of writer vs. reader becomes critical. Evaluating word completion and, more generally, applications for alternative and augmentative communication (AAC) will be discussed.
“An Overview on the Existing Language Models for Prediction Systems as Writing Assistant Tools”.
Ghayoomi, M., Momtazi, S.
View Full Text Show AbstractThe prediction task in national language processing means to guess the missing letter, word, phrase, or sentence that likely follow in a given segment of a text. Since 1980s many systems with different methods were developed for different languages. In this paper an overview of the existing prediction methods that have been used for more than two decades are described and a general classiﬁcation of the approaches is presented. The three main categories of the classiﬁcation are statistical modeling, knowledge-based modeling, and heuristic modeling (adaptive).
“Multi-word Prediction for Legal English Context: A Study of Abbreviated Codes for Legal English Text Production”.
Väyrynen, P.A., Noponen, K.H., Seppänen, T.K..
View Full Text Show AbstractHere, we investigate using strings of expressions longer than a single orthographic word in English word prediction in the legal English domain. The goal of the kind of prediction strategy, called multi-word prediction, is to speed up performance of humans in text production by means of word prediction. Accuracy of two prediction techniques was preliminarily estimated on a simulation without using human subjects with the lexicon of 7,009 multi-word units of legal English. The results show that the average of 70% of characters can be saved for the units in the lexicon in the best-case performance. An improvement in performance actually gained with a real text mainly depends on length and token frequency of units predicted. We also show how the length of multiword units predicted appear to be related to the code lengths used in their prediction and how this finding can be utilized to practical advantage in multi-word prediction.
“Word Prediction Software for Students with Writing Difficulties”.
View Full Text Show AbstractThis investigation examined the effects of currently available word prediction software programs that support phonetic/inventive spelling on the quality of journal writing by six students with severe writing and/or spelling difficulties in grades three through six during a month-long summer writing program. A changing conditions single-subject research design was used and replicated across the participants. Using a daily writing prompt, students alternated between Co:Writer, WordQ, and WriteAssist word prediction programs. The results provided evidence for the effectiveness of various word prediction programs over word processing, and demonstrated improvements in spelling accuracy across conditions. Relative gains in the total number of words and composition rate were modest for the majority of the participants and should be interpreted with caution due to several methodological issues. The social validity interviews revealed that all students enjoyed the word prediction programs and found them beneficial. Study limitations and recommendations for future research are discussed.
“Measuring the outcomes of word cueing technology”.
Tam, C., Archer, J., Mays, J., Skidmore, G.
View Full Text Show AbstractMeasurement of assistive technology outcomes is complex because many factors (e.g., environment and model of service delivery) influence the successful use of the technology. Using the example of measuring the outcomes of word cueing technology, this paper presents an approach for measuring assistive technology outcomes. The Canadian Occupational Performance Measure (COPM) was administered to 29 children with physical and learning disabilities, between the ages of 3.9 and 19 years. Participants were provided with WordQ, a software program designed to assist the development of writing skills. Follow-up data were collected through telephone interviews. Results. The COPM findings supported the effectiveness of WordQ Version 1 to enhance written productivity, with a mean performance change score of 3.5 (SD = 1.5).The COPM was an effective tool for measuring clients’ perceived outcome of word cueing technology. Telephone interview was considered a successful method for collecting outcome data. Practice Implications. A mix of tools and methodologies should be used to gain a comprehensive understanding of the impact of assistive technology.
“Testing the efficacy of part-of-speech information in word completion”.
Fazly, A., Hirst, G.
View Full Text Show AbstractWe investigate the effect of incorporating syntactic information into a word completion algorithm. We introduce two new algorithms that combine part of-speech tag trigrams with word bigrams, and evaluate them with a test bench constructed for the purpose. The results show a small but statistically significant improvement in keystroke savings for one of our algorithms over baselines that use only word n-grams.