NotesNicolas
Group reports
Game group
Use Goodiff content in text-based narrative game. Game arbitrarily gives permission or denies to a character, like the TOS's. Use the actions semantics (examine, take, E, W, S, N) and confront them to TOS's.
Examples: you cannot pick up the TOS on the table because it is too heavy, or you are too tired to read it. Or cannot take content because it has been taken down in the meantime.
Idea: a time machine in the game gives possibility to go in different versions of the TOS, different views on the same paragraph.
Sidenotes: documentary GET LAMP, early text games gave the feeling of infinite text while having very limited storage.
Proposal: expand the game during a pyweek. http://pyweek.org
Historical Context group
Track main phrases that changes across commits. But what exactly is changing? Can the Markow algorithm help? It could give statistical information on changes.
Interesting to detect high frequency of changes. How to trace the related changes in law? How to analyze what changed? Word frequency?
Proposals:
- Evolution in size of the text of the TOS's
- Evolution of tag clouds over time: the top ten words across commits. Compare tag clouds.
Simple Stats group
Highlight/compare the incongruity of the legal jargon.
Ambiguity of a word like "ambiguity". Express it visually, like an action game. What 'control' means in a policy and what it means in another.
How to find interesting terms and know they are present in a sufficient number of policies? Look for word that appear the most? Select from there words that have multiple meanings. Example of visual comparison: http://www.neoformix.com
A word describing an action could also feed an action game.
Illustrated TOS: select an image to illustrate a word, a rebus puzzle for TOS.
See also: Wordnet python module. synsets provide multiple meanings of a word: "give me other forms of this word". Beware of the context, it is far from being neutral (ie. the meanings of woman). Danger when it becomes a universal thesaurus.
Timeline group
Example TOS Video Google -> how to extract the 'real' text of the page? (get rid of navigation information).
Use white space ('white' text) to show what disappears and keep a change of the text still growing (sentences are replaced, augmented). Feeling of pages with growing holes.
Use the button "I agree" to go from one page to the other.
Note: the diff algorithm
Between two documents, select the common longest sequence of words, remove from both texts, place a marker. redo until there remains nothing common between the two texts.