Math Image Description project

1. Math Image Description project

2. Who we are • Rob Wall Emerson, Dawn Anderson • Western Michigan University • Yue-Ting Siu, doctoral student at UC, Berkeley • Under contract from MeTRC (Mathematics eText Research Center) at University of Oregon • Mark Horney • In partnership with DIAGRAM

3. What we are looking at • How best to describe the image components found in typical math textbooks (not the math equations) • What types of images require what level of description? • Are there types of images whose content cannot be adequately conveyed by any description? • We are not looking technological solutions but try to present material in a way as close as possible to a common student experience

4. Our approach • Play audio files of portions of math texts, containing images, to students who are blind from grades 5, 8, and 11. • Files have varying levels of description. • None (“image”) • Little (“image of a graph”) • Standard • Extended • Students are assessed on capture of content and ease of capture of audible material.

5. Some specifics • We categorized images in representative math texts from grades 5, 8, and 11. • Identified 21 exhaustive and mutually exclusive image categories • 4 “meta categories” represent context for the images • introducing concepts • guided example • short question • real world manipulative

6. More specifics • Word documents were created that mirrored the physical page in layout and coloring. • Each file contains images and ancillary text to provide context. • Math content that was not image related was translated into MathML and the entire file spoken using JAWS.

7. Initial data • Chicago: sample of grade 5 and grade 11 students • Texas: 21 grade 8 and grade 11 students • Some images need no description (icons, borders) • Some image categories need limited description (cartoon characters, question specific images) • For some images more description is counter productive (tables, line graphs) • A major trend seems to be that many image categories would benefit from a multi-modal presentation of content • Have an audio version with description and for image related content, also have a braille version of the “description” and a tactile image.

8. Data collection • Data trips being planned for Tennessee, Arizona, New Mexico, and Space Camp in Georgia • Data collection will continue through next school year with a target of 100 students being enrolled • We welcome any schools wanting to talk about their students being involved

9. Category frequency • First tier categories appear either on nearly every page of a text or several times on a page within certain areas of a text. Second tier images are more specific and appear occasionally, usually to serve a specific purpose. Third tier images appear infrequently. • The most commonly occurring image categories (from most to least) in the first tier were: • 1. Side images (background picture, graphic unrelated to question, organizational banners, headers, icons, extra features notation) • 2. balloon/sidebar • 3. question specific image • 4. shapes/2D or 3D representation • 5. table • 6. scatterplot/line graph • 7. number line • 8. ray/line diagram

10. Category frequency continued • The most commonly occurring image categories (from most to least) in the second tier were: • 1. screen shot • 2. flow chart • 3. equation • 4. pattern/series • 5. bar graph • 6. directions/illustrations of a physical task • 7. models (used to indicate similarity) • 8. calculator stuff • 9. maps • The most commonly occurring image categories (from most to least) in the third tier were: • 1. picture in a picture • 2. procedural aid • 3. organizational chart • 4. pie chart