Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!jarthur!uunet!tdatirv!sarima From: sarima@tdatirv.UUCP (Stanley Friesen) Newsgroups: comp.ai.philosophy Subject: Re: Chinese room -- Empirical tests Message-ID: <78@tdatirv.UUCP> Date: 13 Dec 90 23:54:13 GMT References: <74@tdatirv.UUCP> Reply-To: sarima@tdatirv.UUCP (Stanley Friesen) Distribution: comp Organization: Teradata Corp., Irvine Lines: 51 In article ht@cogsci.ed.ac.uk (Henry Thompson) writes: >In article <74@tdatirv.UUCP> I write: >> In article ht@cogsci.ed.ac.uk (Henry Thompson) writes: >> >1) Could you learn Chinese (or any other natural language) as a second >> >language using ONLY a monolingual Chinese (or ...) dictionary? >> No way ... >> ... Also, the examples of the usage in >> a dictionary are generally insufficient to derive the full grammar of the >> language. >For what it's worth, Harnad disagrees, and, unusually, I concur. We >simply observe that what we have been set is not actually a language >learning task in the first instance, but rather a code breaking task. >With the vast amount of material available in a monolingual >dictionary, with a guarantee of no change of code or other tricks, it >should actually be a perfectly doable job by the standards of the WWII >codebreakers. I am not sure this really addresses my reservations about deriving the grammar. Certainly deriving the vocabulary is essentially a code breaking task (though somewhat harder than normal due to incomplete matching between vocabularies). However, almost all real codes have essentially the same grammar as the base language, and code breaking techniques are likely to be inadequate for dealing with, say, the difference in structure between English and Japanese. Ideed, this is likely the reason that Japan and Germany were unable to break our military "code" during WWII, it wasn't really a code, it was Navaho. > Once the dictionary had >been translated, a bilingual dictionary could be constructed and >another tedious, but again perfectly possible in principle, task is >all that remains. Fine, now how do you figure out the rules governing all the complex grammatical constructs, such as conjugations, non-finite subordinate clauses, concord, and so on down the line. For instance, in my Russian example, I would have had little chance with the dictionary if the text had had very many verbs. Why? Because the conjugation of verbs in Russian is so complicated that the commonly encountered finite forms often differ greatly from the dictionary form, making locating them in the dictionary nearly impossible. [I was stuck for several weeks by the 3rd singular indicative present imperfect form of the verb meaning 'have', I only found it by pure luck (I noticed it in a table of irregular verbs). (Of course this is an extreme case, but still illustrative of the problem, in particular the inflected forms of Russian verbs often differ in the initial consonant+vowel)] Vocabulary is easy, grammar is a beast. -- --------------- uunet!tdatirv!sarima (Stanley Friesen)