| Images on the Web present a major
| |
| | our goal to assign proper descriptions to
|
| accessibility issue for the visually
| |
| | arbitrary images. A "proper" description
|
| impaired, mainly because the majority of
| |
| | is correct if it makes sense with respect
|
| them do not have proper captions. This
| |
| | to the image, and sufficient if it gives
|
| paper addresses the problem of attaching
| |
| | enough information about its contents.
|
| proper explanatory text descriptions to
| |
| | Rather than designing a computer vision
|
| arbitrary images on the Web. To this end,
| |
| | algorithm that generates natural
|
| we introduce Phetch, an enjoyable
| |
| | languagedescriptions for arbitrary images
|
| computer game that collectsexplanatory
| |
| | (a feat still far from attainable), we
|
| descriptions of images. People play the
| |
| | opt for harnessing humans. It is common
|
| game because it is fun, and as a side
| |
| | knowledge that humans have little
|
| effect of game play wecollect valuable
| |
| | difficulty in describing the contents of
|
| information. Given any image from the
| |
| | images, although typically they do not
|
| World Wide Web, Phetch can output a
| |
| | find this task particularly engaging. On
|
| correct annotationfor it. The collected
| |
| | the other hand, many people would spend a
|
| data can be applied towards significantly
| |
| | considerable amount of time involved in
|
| improving Web accessibility. In addition
| |
| | an activity they consider "fun." Thus,
|
| toimproving accessibility, Phetch is an
| |
| | like the ESP Game, we achieve our goal by
|
| example of a new class of games that
| |
| | working around the problem, and creating
|
| provide entertainment in exchange
| |
| | a fun game that produces the data we aim
|
| forhuman processing power. In essence, we
| |
| | to collect.
|
| solve a typical computer vision problem
| |
| | We therefore introduce Phetch, a game
|
| with HCI tools alone.
| |
| | which, as a side effect, generates
|
| The Web is not built for the blind. Only
| |
| | explanatory sentences for randomly chosen
|
| a small fraction of major corporate
| |
| | images. As with the ESP Game, we show
|
| websites are fully accessible to the
| |
| | that if our game is played as much as
|
| disabled, let alone those of smaller
| |
| | other popular online games, we can assign
|
| organizations or individuals . However,
| |
| | captions to all images on the Web in a
|
| millions of blind people surf the Web
| |
| | matter of months. Using the output of the
|
| every day, and Internet use by those with
| |
| | game, we mention how to build a system to
|
| disabilities grows at twice the rate of
| |
| | improve the accessibility of the Web.
|
| the non-disabled .
| |
| | Design of a Useful Game
|
| One of the major accessibility problems
| |
| | A traditional algorithm is a series of
|
| is the lack of descriptive captions for
| |
| | steps that may be taken to solve a
|
| images. Visually impaired individuals
| |
| | problem. We consider Phetch as a kind of
|
| commonly surf the Web using "screen
| |
| | algorithm. Analogous to one, Phetch has
|
| readers," programs that convert the text
| |
| | well-defined input and output: an
|
| of a webpage into synthesized speech.
| |
| | arbitrary image from the Web and its
|
| Although screen readers are helpful, they
| |
| | proper description, respectively. Because
|
| cannot determine the contents of images
| |
| | it is designed as a game, Phetch needs to
|
| on the Web that do not have descriptive
| |
| | be proven enjoyable. We do so by showing
|
| captions. Unfortunately the vast majority
| |
| | usage statistics of a oneweek trial
|
| of images are not accompanied by proper
| |
| | period. Because it is designed to collect
|
| captions and therefore are inaccessible
| |
| | a specific kind of data, Phetch's output
|
| to the blind (as we show below, less than
| |
| | needs to be proven both correct and
|
| 25% of the images on the Web have an HTML
| |
| | sufficient. We prove this through a
|
| ALT caption). Today, it is the
| |
| | specifically designed experiment.
|
| responsibility of Web designers to
| |
| | Roberto is the owner of Online
|
| caption images. We want to take this
| |
| | Multiplayer Games.
|
| responsibility off their hands. We set
| |
| |
|
