Automating Partial Credit Scoring in Sentence Transformation
Gespeichert in:
| Veröffentlicht in: | PQDT - Global (2004) |
|---|---|
| 1. Verfasser: | |
| Veröffentlicht: |
ProQuest Dissertations & Theses
|
| Schlagworte: | |
| Online-Zugang: | Citation/Abstract Full Text - PDF |
| Tags: |
Keine Tags, Fügen Sie das erste Tag hinzu!
|
| Abstract: | Many computer assisted language tests with automatic marking capability offer only correct/incorrect grading of short text responses. This approach does not only not reflect how human markers tend to grade short answers, but it also means that useful information about the candidate’s language ability and knowledge is lost.This thesis investigates the issue of automating partial credit scoring in computer assisted language testing, for the item type ‘sentence transformation’ in a computer assisted test of German grammar. After a brief introduction to pertinent fields, i.e. testing and the criteria applied in the development of a language test in particular, as well as computer assisted language testing, the thesis takes a closer look at adaptive testing and the underlying principles of item response theory, examining the possibility of applying Masters’ partial credit model in this project.This is followed by a historical overview of the role of grammar in language teaching and testing and the recent discussion on focus on form vs. focus on forms. The role of grammar in language acquisition in discussed with reference to a study into the acquisition sequence of the German tense system.Subsequently the area of German perfect tense is outlined and an item bank for a pilot test of German perfect tense is developed. This is done with special focus on the problem of verb selection as exemplified by an analysis of basic word lists and the GCSE syllabus for 2003/4.The pilot item bank was field trialled using a test created in Question Mark Designer at schools and universities and the resulting experimental corpus is converted into Microsoft Access database tables which are subsequently analysed to yield item parameters for difficulty and discriminability to verify the suitability of the verb selection process.A partial credit tagging and scoring system is then developed and applied to the responses, its efficiency and accuracy are subsequently tested using the filter and query functions. The concluding section outlines the possibilities of the application of this approach to other areas of grammar, the web, its integration into national tests, the development of an interface between a computer assisted test and a database. |
|---|---|
| ISBN: | 9798379405113 |
| Quelle: | ProQuest Dissertations & Theses Global |