Achieving Error-Free Scoring
SOLUTIONS FOR FIXING THE COLLEGE BOARD SAT SCORING ERRORS
The 2006 SAT test scoring and grading errors by the tabulation company engaged by the College Board highlights the need for upgrading current scanning technology. The test assessment and grading errors were attributed to moisture induced paper expansion and the inability to resolve lightly marked “bubble marks”.
The inability to resolve lightly marked “bubble marks” and errors due to paper expansion or shrinkage is a chronic industry problem. Combined with paper alignment errors, approximately 1.5% of the SAT tests were tallied incorrectly. The largest scoring and grading error rate was 450 points or 19% on a test of 2400 points or 25% for a 1600 point test.
Interestingly, the same method and error rates also appeared in tabulation of votes in US elections when paper ballots are used. In fact, the Federal Voting System Standards of 2002 specifically mandate less than 1 error in 500,000 marks when the paper ballot is marked correctly (i.e. properly filled “bubble marks”). Systems using discrete sensor/OMR technology struggle to meet such requirements.
AVANTE International Technology, Inc., though, has developed a solution tallying ballots using pixel count and document imaging technology. The AVANTE system achieve less than one error in 1.5 million marks! The same technology has been used successfully for tabulation of surveys and questionnaires besides tallying ballots.
The following compares AVANTE’s Test and Survey Scoring System and the conventional OMR and discrete sensor mark-sense systems:
COMPARATIVE ANALYSIS OF DISCRETE SENSOR/OMR AND PIXEL-IMAGING (AVANTE Patented US 6,892,944; 7,077,313 and other patents pending) | ||
DISCRETE SENSOR-OMR | IMAGING + PIXEL COUNT | |
Recognition of filled “bubble mark” | Yes | Yes |
Recognition of lightly filled “bubble mark” | Mostly no (Need high threshold) | Yes |
Independence on timing tracks | No | Yes |
Ability to resolve registration (fiducial) marks | No | Yes |
Ability to resolve wrinkled and creased papers | No | Yes |
Ability to resolve paper shrinkage/expansion | No | Yes |
Recognition of “þ” and “ý” marks | Mostly No or marginal. | Yes |
Resolving barcode (type or individual form) | Yes | Yes |
Resolving multiple and different pages | Yes | Yes |
Resolving random orientation of pages | No | Yes |
“Separating” written answers from “bubbles” | No | Yes |
Automatic self-check for accuracy | No | Yes |
Ability to retrieve individual tests for recount | No | Yes |
Speed (Per Scanner) | Up to 10,000 pages/hr-scanner | Up to 6,000 pages/hr-scanner |
Accuracy | > 1/1,000 (2006 SAT case: >1/100) | <1/1,500,000 (Federal ITA tested) |
The discrete Optical Mark Recognition (OMR) sensor technology relies on completed documents passing through the scanner “exactly” straight. The width marking positions must line up “correctly” and timing must be exact for reading the length of the paper. With the use of advanced document imaging technology, AVANTE uses fiducial markings to scale for paper orientation and expansion/shrinkage variations. The quantitative use of counting pixels in each “bubble mark” also provides automatic self-checking for possible light marks, smears or errors. The following table is a summary of the differences between traditional discrete sensor OMRs and AVANTE’s imaging with quantitative pixel counting technology.
Comparative Analysis of the Traditional Discrete Sensor Based “OMR” Tabulation System And Pixel Based Optical Imaging Tabulation System | ||
Discrete Sensor Based “OMR” System | Pixel Based Optical Imaging System (AVANTE patented innovation) | |
Ease of Audit |
|
|
Costs of forms and readers |
|
|
Multiple page forms |
|
|
Stacking and alignment |
|
|
Flexibility of forms |
|
|
Accuracy |
|
|