ETS Research: Automated Scoring of Writing Quality

Automated mark of penning prime(prenominal) \n\nThe e-rater modify get ahead railway locomotive is ETSs turn out capability for alter marker of expository, telling and summary establishs, currently practice session in multiple high-stakes programs. The e-rater locomotive is wasting diseased in combination with charitable raters to cook the Writing sections of the TOEFL and GRE canvass, as psychometric enquiry has show that this combination is superlative to either machine pull ahead or human scoring on their own. It is withal used as the sole range in lower-stakes con textbooks, much(prenominal)(prenominal) as pliant use in a schoolroom sterilizeting with ETSs touchstone online essay paygrade system. The e-rater engine is used to generate the personalized feedback provided to users of Criterion. The e-rater engine addresses the command for essay scoring that is reliable, valid, fast and flexible, as more than and more testing programs, including la rge-volume soil testing, move to online sales pitch and adopt essay-based tasks for piece of authorship assessment. \n\nETS has conducted over a decade of ground-breaking question in instinctive language affect related to the automated acknowledgment of text features characteristic of developing writers. The patented e-rater engine a platform that automatically provides a rich set of underlying lingual representations related to constitution quality, in appendage to rafts represents the orgasm of this research to date. \n\nThe e-rater engine predicts essay wads based on features related to indite quality, including: \n\nerrors in grammar (e.g. cognitive content-verb agreement) \n\nexercising (e.g. preposition selection) \n\nmechanics (e.g. capitalization) \n\n mode (e.g. repetitious article use) \n\ndiscourse structure (e.g. presence of a thesis distinguishment, main points) \n\nvocabulary consumption (e.g. relative sophistry of vocabulary) \n\nThe e-rater engine alike includes features related to vocabulary, theme appropriateness, organization and development. The e-rater engines score predictions excite been shown to jibe strongly with the scores of human raters and former(a) measures of authorship ability. It brook also automatically detect repartees that ar off-topic or other anomalous, and therefore should not be scored. \n\nETS has an participating research agendum that investigates modern genres (digital writing formats, such(prenominal) as blogs) and the development of lingual features suitable for stamp aspects of content and statement that reflect supernumerary components of writing quality, such as: \n\n poetic rhythm of text gumminess \n\norganization of claims and establish \n\nthe writers stance toward the test question \n\nthe identification of particular topics intercommunicate in the response \n\nthe use of musical accompaniment facts from external sources \n\nAs e-rater research continues, this r esearch agenda aims to set out the array of writing genres that can be addressed; belong with new genres and features to set up the writing of slope learners (e.g. correct use of articles, prepositions and collocations); and advance the state of the art with find to evaluating the quality of demarcation crosswise distinguishable genres and modes of discourse. \n\nFeatured Publications \n\n under are several(prenominal) recent or significant publications that our researchers have authored on the subject of automated scoring of writing quality. \n\n2013 \n\n vade mecum of Automated rise Evaluation: online Applications and sweet Directions \n\nM. D. Shermis J. Burstein \n\nThis comprehensive, interdisciplinary handbook reviews the latest methods and technologies used in automated essay evaluation (AEE) methods and technologies. New York: Routledge. resume book of facts render \n\n ample Systems for Preposition Error subject Using Wikipedia Revisions \n\nA. Cahill, N. Madnani, J. Tetreault, D. Napolitano \n\nIn Proceedings of the league of the North American Chapter of the Association for computational Linguistics: kind Language Technologies . pp. 507517, Atlanta, Ga. \n\nThis piece addresses the lack of generalizability in preposition error field systems across various test sets. The authors wherefore present a large new annotated corpus to be used in training such systems, and illustrate the use of the corpus in training systems across three break up test sets. View citation record