XPost: comp.ai.nat-lang   
   From: ptelegone@wanadoo.fr   
      
   "Erik Norvelle"  a écrit dans le message de news:   
   4032468a$1@news.unimelb.edu.au...   
   > Greetings,   
   >   
   > Here's my problem, and it's a doozy:  How can I evaluate   
   > (computationally) the similarity between the following sentences:   
   >   
   > 1) The unexamined life is not worth living   
   > 2) Only the examined life is worth living   
   > 3) It's only worthwhile to live an examined life   
   > 4) A life that is not examined is not worthy of being lived   
   >   
   > The reason I need to be able to evaluate this kind of "conceptual   
   > equivalence" is that I am designing a study-aid program for   
   > philosophy, which depends on students' being able to express "in their   
   > own words" the concept that a philosopher has expressed in response to   
   > a particular issue.  I'm trying to avoid multiple-choice type   
   > interfaces, as well as strict memorization.   
   >   
   > For instance, in response to the question "Why did Socrates practice   
   > philosophy?", the student should be able to give any of the   
   > above-cited answers and still be given credit.  I'd like to be able to   
   > evaluate the "degree of closeness" between what the philosopher   
   > actually said and what the student says, and give credit   
   > appropriately.   
   >   
   > Ideally, I'm looking for a toolkit that I can integrate into my own   
   > programs; it doesn't matter if the maturity of the toolkit is poor.   
   > The language can be anything that I can integrate with Java using the   
   > JNI (i.e. C, C++, etc). In particular, I am looking for open-source   
   > projects, or projects sponsored by public institutions, rather than   
   > commercial tools.  However, I would appreciate  hearing about   
   > commercial products if nothing else is available.   
   >   
   > As a former linguistics student, I'm aware that I may be on a wild   
   > goose chase, due to the difficulty of the subject matter.  I'm hoping   
   > that progress has been made in this area since my last immersion in   
   > this stuff (some 13-odd years ago).   
   >   
   > I will certainly check this newsgroup for updates, but if you'd be so   
   > kind, I'd also appreciate carbon-copies to my email address, which is   
   > "erik at norvelle dot net"   
   >   
   > Thanks very much,   
   > Erik Norvelle   
   > Master's Student/Philosophy   
   > Universidad de Navarra   
   > Pamplona, Espana   
      
   Hello,   
      
   Well, I have a tool that could be adapted,I think, for this kind of work.   
   This tool is Qa and it is a chatbot that I have developed to do   
   questions-answering. It is able to see the similarity (or more precisely the   
   dependance) between two facts, provided that a rule of dependance matching   
   these 2 facts has been given.   
      
   For instance, if the user first says "John is the son of Michael", when he   
   will enter "Michael is the father of John", the system will answer "I know":   
   it has found that the two facts are similar. It does that by a using a   
   deduction rule that has been specified to it.   
      
   I think this feature could be adapted to your need with some little work. Of   
   course, the deduction rule for the moment is hard coded in the tool, but it   
   could be externalize easily.   
   Qa will be a commercial tool. For the moment, there is a beta version (that   
   can be tested online or downloaded). Only simple turns of phrase are   
   supported for the moment.   
      
   The web site is www.interronet.com   
      
   Regards   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)