] Placement Testing and Audio Quiz-making with Open Source Software Pacific Association of CALL International Conference National University of Singapore, 3 December 2004 Don Hinkelman Sapporo Gakuin University University of Melbourne Case Study: 1. 2. 3. 4. Sapporo Gakuin University 2000 first and second year students. Two year required general English program. Students assigned alphabetically by family name. Mixed ability levels in each class. Six full-time teachers agreed in a staff workshop that unstreamed class levels was their greatest problem - bright students not challenged, lost interest - weak students over-challenged, lost interest Mixed levels - “Like herding cats” QuickTime™ and a Sorenson Video decompressor are needed to see this picture. Case Study: Sapporo Gakuin University 2. Commercial testing software was expensive and would take 1-2 years to get budget approval. We had doubts about its usefulness. We wanted a system where we could control and improve the the test items. - Calus Testing Service 3000 yen/student/year 1000 freshmen/year. Total budget = 3,000,000 yen/year (USD31,000) Case Study: Sapporo Gakuin University 3. We wanted to adapt our current progress test system (50 questions--30 listening, 20 reading) for placement of incoming freshmen. But paper test marking services 300,000 yen /year We wanted a “testing machine” that would not give an audio-based test, but also analyze the results so we could improve the test questions. Placement Testing Machine version 3.1 50 User friendly questions Reusable stainless steel interface User adaptable hood harness Wrist and leg straps not necessary I am level 3! Questions Case Study: Sapporo Gakuin University 4. Tool Selection We had experience administering computer-based m/c tests in large English classes using an open source CMS with quiz-making tools (Moodle). - 10 computer rooms available in beginning of term. Why not do a computer-based placement test Rooms had no headphones. But quizzes could not do audio. How would we do listening questions? Case Study: Sapporo Gakuin University 5. Programming We went to local colleague-volunteer and the lead developer of Moodle and asked them to modify Moodle to handle listening questions. - One colleague contributed 110,000 yen of personal research money for programming. - Sent money to moodle.com to supervise process. - Local colleague fell through--no programming - Dropped all bells and whistles requirements - moodle.com came through with core program: auto-embedded flash buttons with media filter \ ] Core Requirements for a placement test programming Play/Stop Button within each Question: It should be intuitive for student to use No Popup Player: Avoid confusion for student Number of Plays Setting: It should be easy to create an online course from existing learner resources, Simple Login for Student: It should provide a simple screen. provide bilingual menus, help, and input. No Result Displayed at End \ ] Actual Accepted Interface for a placement test programming OK Play/Stop Button within each Question: It should be intuitive for student to use OK No Popup Player: Avoid confusion for student CUT Number of Plays Setting: We just set a start and end time for the whole test. We let students play as many times as needed, but warned them to finish in time allowed. CUT Simple Login for Student: We used standard Moodle login. Teachers turned on machines, logged into main school system. Then we assigned seats, and put unique username and password on a paper on the keyboard of each machine. CUT No Result Displayed at End Course Mangement System Evaluating Tests: Practicality Reliability Validity Washback Four Criteria Types of Tests Criterion Referenced Proficiency Tests Achievement Tests Norm Referenced * Placement Tests [Implications for design, validity] Content can be more arbitrary Construct based more on listening, less on reading Validity Content ● Related to high school curriculum in Japan ● General topics of common interest in Hokkaido Construct ● Greater emphasis on listening 60% ● Less emphasis on reading 40% ● No speaking, writing Washback ● Effect on teachers – Informal feedback: ● ● enthusiastic comments felt student group cohesion better, Reliability Item Facility - 50 questions X O X 0~30% 30~40% 40~60% 60~70% 70~100% Item Discrimination X <1.0 1.0~1.3 Distractor Analysis O >1.3 Reliability Results Item Facili ty Performance Number of Items Level Good 12 Acceptable 16 Poor: too difficult 6 Poor: too easy 16 Item Discrimi nation Performance Level Good Acceptable Poor Number of Items 20 18 12 Too easy 92% correct High IF Item Q-38 (page 3 in handout) Watching the news always _____ me sad A. puts B. changes C. results D. makes Too hard 5% correct Low IF Item Q-55 (page 3 in handout) What does the author feel about human relationships in the future? A. B. C. D. He thinks they will be more convenient He thinks they will be more controlled by technology He does not express an opinion on the matter He thinks they will contain a large amount of information Random Answers Low ID = 1.3 Poorer students answer correctly as often as better students Item Q-48 (page 3 in handout) Which word means the same as “task”? (in the second paragraph) A. B. C. D. tool job ability industry Practicality ● ● ● ● Organizing and Achieving Consensus Exam Proctoring and Management Hardware Test Software Test Results Server 1 Mac OS X, 933 mhz, single Result: 150 students, server stopped, switched to paper test 80 students, very slow, successful Server 2 Mac OS X, 1.2 ghz dual Result: 100 students, server stopped Finder problems, hardware freezes Server 3 Linux Fedora 2, 3.2ghz single Result: 100 students, successful 250 students at one time planned for 2005 2 servers, plus two backup servers Hardware Specifications/Cost Server 3 Linux Fedora 2, Pentium 4--3.2ghz single processor Built from parts, server engineer made LAMP in free time. Parts Motherboard Processor Memory, 2gb Hard disk, 120gb DVD-R drive Case Labor Assembly Configuration yen 13,000 30,000 45,000 8,000 6,000 8,000 USD 120 270 410 75 55 70 10,000 40,000 90 360 Total Cost: 160,000 yen 1,450 USD RAID with second hard disk will automatically clone whole server +30,000 yen Time Costs Test Rooms 4 x 60 computers Conclusion ● Future: – – 80 students >> 1000 students in one day Two tests ● ● ● Higher Level: English/Humanities majors Lower Level: Economics/Law/Computer majors Goal: – – – April 2005 Gradually improve test items 45% problem items >> 10% problem items Save $30,000 commercial testing fees/year Superior placement test norming for our student population Thank you Contact: Testing Operation: Don Hinkelman, Tim Grose <hinkel@sgu.ac.jp> <grose@sgu.ac.jp> Statistical Report Module: Tom Robb <tom@robb.net>