This is in a similar vein to another of my earlier posts about what a fully integrated website platform might look like. That was in turn inspired by the format of lernu! and various imageboards, message boards, and data repositories.
First some background on these things, forums are places you go to post stuff and have discussions. You may contrast them with chats, which are much faster and spool out ridiculous amounts of data...tho the most popular forums and message boards are about as fast as the fastest chat. Chats are not private like instant messages between two people. Twitter is very similar to the chats and message boards, with lots of short texts, but its more like blogs in how people connect to you and interact; tweets are like a soapbox. In a message board, the newest message bumps the entire topic to the top (unless the respondent wishes to see the topic not be pushed: see SAGE(ru) a Japanese term), but how long it stays up there depends on how active the topic is. Unlike a forum, message boards show part or all of each topic in-line on the main page, while in a forum, each topic is linked to from a main page with a title and an optional subtitle. IMDB forums are neat in that they show the tree structure of the conversations i.e. who responded to who; they do that by indentation.
Now all of these have the same linking structure, i.e. clicking a link will bring you to a certain kind of page. In a Wikipedia, that means entry/category/talk/edit/etc. pages. What if each link were treated as a kind of object you could interact with by clicking on it and pulling up a menu...Maybe you could shift+click it, or alt+click it, to do things faster, like a videogame. Right now it's the individual pages that are affiliated with links, and according to their structural template (e.g. Wikipedia entry pages have talk, edit, etc. pages) but what if it was the data on the pages themselves that needed specific entry pages? This is the idea I got because every esperanto word on a page at the lernu forums can be clicked on to find its definition, in addition to the possibility of using that same data as URL links. (I got this idea from the drop-down menu format used in certain new adventure games where the text is alive and the only way to interact with objects. Clicking on a hyperlinked object pulls up a number of applicable action verbs or METHODS you can use on it, in addition to various objects you can do those actions to it with. See Quest Adventure games.)
The 3rd thing that I thought was really important was the need for more audio online. The only advantage text has over audio is that it's faster to skim and search. Well, what if every audio object had a text? There is a minimal level formatting that would be useful, so that instead of skimming, we just click a mouse and skip to the relevant portion. What would that require? Very little. Dragon already does a darn good job inserting punctuation into documents on its own now without needing to even end your sentences by saying period. This is a much easier task, associating word/clause/sentence boundaries with spaces/commas/periods/etc. Thats not much more than some basic thing you cna do in Praat (altho I'm not sure you can just click on words and play audio, see Praat). Unless we want to go whole hog and just speak to our PCs and let people sort out what was actually said when the computer malfunctions and mistranscribes something. Where did I get this idea? From Dragon Naturally Speaking. You can actually go into Dragon, and check what you really said and whether it matches what the computer wrote. I can see people getting used to that and not wasting time writing anymore. Opera has a good voice synthesizer but this would be putting speaking before writing, which is really the way it should be. Speaking is easier, faster and carries more information.
So that's several layers of information--dictionary, audio, links. Lets add another. What if there were several tropes layers? Tropes are patterns. One of the best expamples is in intonation, inflection, how we say things. Intonation is very poorly studied and readers really have to read between the lines to get it right. What if we had a crowd sourcing effort? It would be much easier with an audio corpus, and for the first time this seems possible because everyone has audio recording functionality in even the most basic computer models nowadays. Essentially, a text would be analyzed according to its audio and annotated with tags at the syllable, prosodic word, intonational phrase, etc. levels. There is no specificity whether this folksonomy should be holistic, like emotions e.g. "Angry", or based on paragons e.g. "Y'know", or if it would break up and be more fine-grained and take into account gradient phenomena. At the very least, it should be able to sort out like from dislike, and provide good audio. (That's another thing, audio should be marked for how natural it sounds, as in atypical or poorly enunciated.)
The basic structure of the platform should take into account older vs. newer discussions, but ongoing discussions should be allowed to survive. I suggest sorting topics by their busy-ness, and the stretch of time
No comments:
Post a Comment