Board ArchivesSite FeaturesActiveworlds SupportHistoric Archives |
richard wojcik // User Search
richard wojcik // User SearchEjection identification - Put to TOP of listJun 14, 1998, 2:13pm
I hate to be discouraging here, because I'm actually in the language processing
business. Filtering software is an interesting challenge, but it isn't trivial. What is trivial, as others have pointed out, is figuring out a way to outsmart the fuucking filters. This is much more of a social problem than a technical one, and we already have some tools for dealing with nuisances--mute and eject. Nobody can tell when they are being muted, so why would a silent filtering method be useful? Are there cases where you want to prevent yourself from hearing obscenities, but still listen to the rest of the content? Mute the b*st*rd! If you created a filter that was clever enough to detect all possible spelling permutations of obscene words, then you would probably have it blowing holes in text all over the place. It just isn't as easy as you think. I once proposed an alternative to the eject mechanism--public mute. Individual muting is only partially helpful, since not everyone agrees to mute nuisances. What I've observed in such cases is that you end up hearing people that you don't mute in argument with people that you do. So the miscreant gets to annoy you by the proxy of your friends' behavior. Ejection is the only solution right now, but it forcibly excludes the person. There are cases where you want to quiet someone down--e.g. public meetings--but not necessarily exclude that person from the world. Actually, a total public mute would be very helpful during public events, since chat tends to get so chaotic when there are a lot of avatars present. Anyway, I don't think that linguistic filtering is going to be all that helpful in dealing with public nuisances. [View Quote] > Would you be willing to copy this over to wishlist, maybe even > worldbuilders where your most excellent idea would be more readily seen by > the more technologically "endowed" people? Especially try to talk to > Roland. USER control is absolutely the most sensible and amenable means of > dealing with what we, each individual, finds offensive. > > In article <3582A01C.5D8E at constant.com>, marvin at constant.com says... VoiceJun 7, 1998, 1:10pm
I have advocated something a bit simpler in the past. I like the
text-to-speech (tts) capability of Powwow. With a flick of a switch, the text in the chat window is sent to a tts program. The program mispronounces words like crazy, but you can follow the conversation without having to look at the screen. If implemented in AW, tts would allow you to keep the browser minimized or hidden behind other windows while working on other things. You would hear it when someone tried to initiate a chat you. Also, tts, unlike true voice, would not add anything to bandwidth. It would run totally on the client side. The Powwow tts is not very good, but it is functional. I wonder if it would be possible to use Festival, which is a very good tts program. One can get Festival with the free speech processing Toolkit from the Oregon Graduate Institute's Center for the Study of Language Understanding (CSLU) at http://www.cse.ogi.edu/CSLU/. [View Quote] Tutoring, Flash Cards, Quizzes, etc.Jun 19, 1998, 12:50am
Last week, a few of us were speculating on how best to create interactive true/false
and multiple choice questions with signs. I had proposed the use of two signs for a "flash card" effect--one in back of the other. After a little more thought and testing, I arrived at the following technique. Sorry if you've thought of this already, but I hadn't tried to do this kind of thing before. I have a "wish" at the below, so I am cross-posting this to wishlist. The effect you want to create is that of a single sign with a question on it. You click on it, and the answer pops up in the same sign. In fact, you can create the illusion of a single sign in which the message changes several times by clicking on it or some other object (e.g. true/false or multiple choice answers). I don't know any way to implement this with a single sign. The best you can do is change the text once by having an 'activate sign "answer"' command in the Action box and the question text in the Description box. The text changes on a single click, but you have to recreate the object to get the original text back (e.g. by leaving and coming back). The trick is to have two or more signs (or picture objects) occupy the same space. All but one of the objects is created with 'visible off'. All have distinct names, say 'name a1','name a2', etc. All boxes have an 'activate visible off, visible <next sign> on' command. I have implemented some examples of the technique in AW at 1052N 198W 150. In theory, you can use this method to create a long series of signs, with true/false or multiple-choice branches. You are limited by the amount of text permitted in the cell, but you can just teleport the student-avatar to a new location to continue the quiz. I'm not sure how useful this technique would really be for teaching purposes. After all, why not just do the quiz on a web page? But the illusion of changing text and picture content might come in handy for some purposes. Here's my wish: It would work much more conveniently if we could change the syntax of the 'sign' command to allow control of named sign and picture objects. If you could attach an 'activate sign <name> "Correct!"' to a separate object, then you wouldn't need to put multiple sign objects with 'visible on/off' actions in the same space to achieve this effect. |