A new sentence generator providing material for maximum reading speed measurement
| Autor: | Damien Paille, Jean-Luc Perrin, Thierry Baccino |
|---|---|
| Rok vydání: | 2014 |
| Předmět: |
Adult
Male Computer science media_common.quotation_subject Speech recognition Experimental and Cognitive Psychology Ontology (information science) computer.software_genre Young Adult Arts and Humanities (miscellaneous) Truth value Reading (process) Common knowledge Developmental and Educational Psychology Psychophysics Humans Learning Control (linguistics) General Psychology media_common Language Psycholinguistics business.industry Middle Aged Knowledge Reading Research Design Female Psychology (miscellaneous) Artificial intelligence business Comprehension computer Natural language processing Sentence Algorithms Generator (mathematics) |
| Zdroj: | Behavior research methods. 47(4) |
| ISSN: | 1554-3528 |
| Popis: | A new method is proposed to generate text material for assessing maximum reading speed of adult readers. The described procedure allows one to generate a vast number of equivalent short sentences. These sentences can be displayed for different durations in order to determine the reader's maximum speed using a psychophysical threshold algorithm. Each sentence is built so that it is either true or false according to common knowledge. The actual reading is verified by asking the reader to determine the truth value of each sentence. We based our design on the generator described by Crossland et al. and upgraded it. The new generator handles concepts distributed in an ontology, which allows an easy determination of the sentences' truth value and control of lexical and psycholinguistic parameters. In this way many equivalent sentence can be generated and displayed to perform the measurement. Maximum reading speed scores obtained with pseudo-randomly chosen sentences from the generator were strongly correlated with maximum reading speed scores obtained with traditional MNREAD sentences (r = .836). Furthermore, the large number of sentences that can be generated makes it possible to perform repeated measurements, since the possibility of a reader learning individual sentences is eliminated. Researchers interested in within-reader performance variability could use the proposed method for this purpose. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full text from SpringerLink