Again WikiTrivia is a proof of concept. The unrelated question problem could be resolved in a couple of ways:
1) We could use better category search terms. Here is a list of search terms for each category: http://github.com/alexksikes/wikitrivia/blob/a6a1956621ace2010310a8b2e9f2decf2f045c7c/app/models/categories.py 2) We could use something like open calais, as you suggested, to verify a question matches a given category. If it does not then we could fetch another question from the category using 1) and repeat. 3) We could make use of user feedback. Users could flag unrelated questions. 4) We could download all of Wikipedia and simply use their categories. You mentioned some questions are impossible to answer: 1) It might be because there are too difficult. In which case a heuristic might be to only take the top questions from the yahoo search. We could play around with difficulty levels this well. The higher ranked, the more popular and therefore the easier the question might be. 2) It might be because the question comes from a discussion page or is a stub or is a year to be guessed... In that case using the '-' operator on the yahoo query for certain keywords such 'stub' would do the trick. Note that this is already in place but needs improvements. 3) Other options include accumulating statistics and weed out questions that users were never able to answer... or making use of user feedback so that at the end of each question a little survey would appear: very difficult, difficult, medium, piece of cake, impossible to answer. Alex 2009/1/16 adelevie <[email protected]>: > > Running all or some of wikipedia ahead of time would not be such a bad > idea, imo. My overall point was that the user experience was poor and > could use some work--just offering constructive criticism. > > On Jan 15, 4:58 am, Alex Ksikes <[email protected]> wrote: >> The blank to be guessed is simply the title of the fetched wikipedia >> article. However it is true that words such as "of, the, they etc .." >> should not be blanked in the question. So for example "king of france" >> would be blanked but "of" should not be blanked every where else in >> the question... That problem could be resolved by simply using a stop >> list. I still fail to see the use of heavy weight NLP methods here. >> >> We use a yahoo query in order to restrict questions to specific >> categories. For example the yahoo search terms for the category >> "Movies" could be tv, movies, cinema, actor, actress etc ... >> restricted to the site wikipedia.org. >> >> Here is a the complete set of search terms for each >> category:http://github.com/alexksikes/wikitrivia/blob/a6a1956621ace2010310a8b2... >> >> Maybe we could use something like open calais to guess the category of >> a random wikipedia article. However that would require we pre-fetch a >> lot of articles before hand.. so we may as well download all of >> wikipedia and use their categories instead ... >> >> 2009/1/14 adelevie <[email protected]>: >> >> >> >> > I was imagining open calais to be used as a tool for selecting which >> > words/phrases to turn into blanks for questions. For example, with >> > open calais, someone would choose to do questions about programming >> > languages, for example. Open Calais knows which words are programming >> > languages. >> >> > On Jan 14, 12:43 pm, Alex Ksikes <[email protected]> wrote: >> >> 2009/1/13 adelevie <[email protected]>: >> >> >> > Great idea, although thus far, it is poorly executed. There needs to >> >> > be some way to further limit the topics, as some of these are so >> >> > random, they are impossible to answer. Furthermore, sometimes I would >> >> > be asked to fill in the blank of a discussion page on wikipedia. Maybe >> >> > you should take a look at opencalais to aid with natural language >> >> > processing. Good luck. >> >> >> Don't need something like opencalais for that. You could use the page >> >> rank of the wikipedia article (or yahoo site explorer) to get a feel >> >> of how difficult a question might be. Also you could always show >> >> articles that have an image as a hint to filter out the the bad >> >> questions. >> >> >> There are many improvements for sure and WikiTrivia should be taken >> >> much more as a proof of concept. It'd be nice for example to play with >> >> different languages. Feel free to have a look at the code and improve >> >> on it. >> >> >> Thank you for the feedback. >> >> >> > On Jan 11, 11:04 am, Alex K <[email protected]> wrote: >> >> >> Hello, >> >> >> >> WikiTrivia, the trivia game freshly generated from Wikipedia articles >> >> >> has been entirely rewritten in webpy and the source code has been made >> >> >> available. >> >> >> >>http://wikitrivia.ksikes.net/http://github.com/alexksikes/wikitrivia/... >> >> >> >> Please feel free to let me know what you think, >> >> >> >> Alex >> >> > > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "web.py" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [email protected] For more options, visit this group at http://groups.google.com/group/webpy?hl=en -~----------~----~----~----~------~----~------~--~---
