[alicebot-style] again,object originated aiml qustion style

[tom]

hi all,
    i am sorry,the last letter's coding is GB2312,some of you may haven't seen it.this time i chaged it to ascii.
    thanks for Kim's response.
    here is some more of my argument.
    first,to Kim,you say "You will have a lot more "what is" questions than "bush" questions, so I think number 2 is better and will lead to a more optimal storage of information inside of the graphmaster." .what i want to say is ,there are a lot of interrogatives like "how","whether",etc. surposing we have 10 interrogatives and 1000 objects,we can have 1000 links for each interrogative or we can have 10 links for each objects, the occupation of space of the two ways are the same.you say "But how do you solve questions, where there is more than one level?",en,that is a hard question ,i haven't think that more,but i am sure that if the frame work is good, the complex ones will be solved after the simple ones.What i say next is something more i am thinking about this,may be useful.
    The first is using the concept of inherit.also start with a sample. we meet a sentence "the earth is round",the question we may store should not be "earth" "whatis" "round" but should be "earth" "whatis" "shape","shape" is the supper class of "round", and that is the attribute of the earth we want to get.we can save the super class of a object in its "whatis" pattern ,like  <category><pattern>round whatis</pattern> <template>shape</template></category>.Because we can not write a category like     <category><pattern>earth whatis <srai>shape whatis</srai></pattern> <template>round</template></category>,we have to get the supper class of "shape" before we write the category.that need some more thinking.
    The second i want to say is fuzzy match.Currently one NodeMaster can only save one word,so the searching is a precise match.That is why AIML is not good at long sentence handling.i think we can modify the program to make the NodeMaster to save a string which can contain more words,so we can make a fuzzy match.here is a example. "why you came to school late this morning?",we can save "you" "why" "come to school late morning".if we ask "why you are late this morning for school?",if we use the old saving model,we could not get the answer, but if we use the new one ,we can ask "you" "why" "late morning school",the third level of NodeMaster "late morning school" can have a high similarity with " come to school late morning",so,we can get the answer.
    that is all, hoping for your replies.our company get some new work, maybe i can not afford much time for this,but i will continue my thinking.

best wishes

tom
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.alicebot.org/pipermail/alicebot-style/attachments/20061211/b6633c9c/attachment.html