<html><body><div style="color:#000; background-color:#fff; font-family:HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif;font-size:14pt"><div class="" style=""><span class="" style="">Hi,</span></div><div class="" style="color: rgb(0, 0, 0); font-size: 18.399999618530273px; font-family: HelveticaNeue, 'Helvetica Neue', Helvetica, Arial, 'Lucida Grande', sans-serif; background-color: transparent; font-style: normal;"><span class="" style="">John said:</span></div><div style="color: rgb(0, 0, 0); font-size: 18.399999618530273px; font-family: HelveticaNeue, 'Helvetica Neue', Helvetica, Arial, 'Lucida Grande', sans-serif; background-color: transparent; font-style: normal;" class=""><span style="font-family: monospace; font-size: 12.800000190734863px;" class="">"For empirical analysis, however, the ontology must be derived or</span><br style="font-family: monospace; font-size: 12.800000190734863px;" class=""><span style="font-family:
monospace; font-size: 12.800000190734863px;" class="">discovered from the documents. The primary reason for the failure</span><br style="font-family: monospace; font-size: 12.800000190734863px;" class=""><span style="font-family: monospace; font-size: 12.800000190734863px;" class="">of so many projects in formal linguistics is that they start with</span><br style="font-family: monospace; font-size: 12.800000190734863px;" class=""><span style="font-family: monospace; font-size: 12.800000190734863px;" class="">an a priori ontology of the subject matter."</span></div><div style="color: rgb(0, 0, 0); font-size: 18.399999618530273px; font-family: HelveticaNeue, 'Helvetica Neue', Helvetica, Arial, 'Lucida Grande', sans-serif; background-color: transparent; font-style: normal;" class=""><span class="" style="">I see two problems with this statement. The first: when a human being reads a text he is using the background knowledge to interpret the text.
Therefore a system should also have coded background knowledge before extracting relevant facts from text. The second : there is no system capable of learning an ontology from text. The best systems are supervised ones. They are learning relations and rules from text with some level of precision and they are relying on annotated data. Moreover the output of such systems should be manually cleaned because it has a high number of errors.</span></div> <div class="qtdSeparateBR"><br><br></div><div class="yahoo_quoted" style="display: block;"> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif; font-size: 14pt;" class=""> <div style="font-family: HelveticaNeue, Helvetica Neue, Helvetica, Arial, Lucida Grande, sans-serif; font-size: 12pt;" class=""> <div dir="ltr" class="" style=""> <font size="2" face="Arial" class="" style=""> On Friday, August 8, 2014 3:48 AM, John F Sowa <sowa@bestweb.net> wrote:<br
class="" style=""> </font> </div> <br class="" style=""><br class="" style=""> <div class="" style="">Mike,<br class="" style=""><br class="" style="">I'm not arguing for prescriptive linguistics. On the contrary,<br class="" style="">note my strong endorsement of the point by Sue A. & Adam K.:<br class="" style="">"I don't believe in word senses."<br class="" style=""><br class="" style="">JFS<br class="" style="">>> That is an important point: Using the word 'ontology' as<br class="" style="">>> a catchall term for any kind of list makes it meaningless.<br class="" style=""><br class="" style="">MM<br class="" style="">> it's amazing to me to see a bunch of *corpus* linguists arguing<br class="" style="">> about what a word *should* mean... But the discussion here has<br class="" style="">> been bordering on prescriptive linguistics, if you ask me.<br class="" style=""><br class="" style="">Thank you for
raising that point, because it clarifies the issues.<br class="" style="">In short, WordNet is *descriptive*, and the ontologies used in<br class="" style="">AI and knowledge representation are most often *prescriptive*.<br class="" style="">Unless you add an emulsifier, they mix like oil and vinegar.<br class="" style=""><br class="" style="">Note the often quoted definition by Tom Gruber: "An ontology<br class="" style="">is a formalization of a conceptualization." I don't like that<br class="" style="">definition, and I don't recommend it -- because it is strongly<br class="" style="">biased toward prescriptive ontologies.<br class="" style=""><br class="" style="">For certain applications, such as software design, a prescriptive<br class="" style="">approach is obligatory: the chief engineer must choose the design<br class="" style="">principles and enforce them consistently.<br class="" style=""><br class="" style="">For empirical
analysis, however, the ontology must be derived or<br class="" style="">discovered from the documents. The primary reason for the failure<br class="" style="">of so many projects in formal linguistics is that they start with<br class="" style="">an a priori ontology of the subject matter. Then they take pride<br class="" style="">in following Frege's principle of using the syntax to guide the way<br class="" style="">their a priori ontology is combined to generate an interpretation.<br class="" style="">That method is guaranteed to fail.<br class="" style=""><br class="" style="">I keep telling them to read Wittgenstein or Adam K's article.<br class="" style="">But they never get the point: To understand a document in<br class="" style="">any natural language, you must *discover* its underlying<br class="" style="">ontology. Nobody (human or computer) can understand a document<br class="" style="">if they start by trying to
impose an a priori ontology on it.<br class="" style=""><br class="" style="">John<br class="" style=""><br class="" style="">_______________________________________________<br class="" style="">UNSUBSCRIBE from this page: <a href="http://mailman.uib.no/options/corpora" target="_blank" class="" style="">http://mailman.uib.no/options/corpora</a><br class="" style="">Corpora mailing list<br class="" style=""><a ymailto="mailto:Corpora@uib.no" href="mailto:Corpora@uib.no" class="" style="">Corpora@uib.no</a><br class="" style=""><a href="http://mailman.uib.no/listinfo/corpora" target="_blank" class="" style="">http://mailman.uib.no/listinfo/corpora</a><br class="" style=""><br class="" style=""></div> </div> </div> </div> </div></body></html>