Guests’ Page

Ian Parker on How Important Grammar is to Language Acquisition

28 September 2008

Ian, retd physicist seen here with wife Margaret Parker

Have you heard about the Honeywell Buzzword Generator? People are inclined to use a lot of words like “The Theory of Knowledge” to sound clever, without really knowing what they mean. I have a number of grave misgivings with the whole idea as applied to teaching.
 
 Emphatically grammar is needed in Machine Translation systems. Google Translate uses pure statistics and this shows. Google’s Arabic translation is absolutely appalling. It makes no reference to parsing. To do a machine translation you need first of all to parse. Once you have a parsed version of a language, translating it is relatively easy. You simply look the words up in a dictionary, get the correct inflection. OK you do need to then use context to resolve ambiguities but 90% of the battle is over with a parse.
 
 However to say that machine intelligence prompts us to a particular view of grammar although persuasive is not in itself compelling. The human brain is capable of doing a lot of things far better than a computer. Pattern recognition is one of them. One does not either send the Indian cricket team to NORAD to learn how AI performs missile interception. No, cricketers perform that task instinctively. My difficulty with a “theory of knowledge” is that TOK refers to mathematics and mathematical machines. It assumes implicitly that humans perform tasks in the same way as machines. What are the instincts involved in language? Is there an instinct that enables us to pick it up faster than other pieces of knowledge? There might be, but the literature is ambiguous on the subject to say the least.
 
 The fact of the matter is that people who are bilingual have the grammar of both languages instilled in their subconscious. You certainly do need grammar to be able to understand a language. The question really is do you get that understanding by simply listening/reading another language, or do you get it by explicitly learning grammar. The only real answer would be to to a control experiment…

The question to me is this. If you were going to the British or Indian embassy in Damascus, would you learn phrases or would you learn prefixes, suffixes and their compatibilities with different stem types? I think if you are going as a tourist you learn phrases, if you are a diplomat or a business person you should learn about prefixes, suffixes and grammar in general. As I have said it depends on your level and perhaps how well you remember.
 
 To me there is one other piece of evidence. People can learn the Koran by heart and not be able to put together meaningful sentences in Arabic. If you really could learn a language in this way, one would at least expect someone to know Arabic after learning the Koran by heart and looking at a bilingual text.

No you need to understand a structure. All the empirical evidence is in favour of this point of view.

When I had recently written that we need to teach grammar, Ian had commented that he agrees. So I had requested him to write why HE thinks so.  What you read above is the matter he sent in to be posted here.

Ian says: I am a retired scientist with a strong interest in Artificial Intelligence. My wife is a theologian specializing in Church History. I speak French and German fluently. I have some knowledge of Spanish and have used this for my examples on the resolution of ambiguity.Von Neumann proposed a theory of formal languages. It should be pointed out that all of Von Neumann’s languages were context independent. Natural language is very much context dependent. I discuss the provision of a context free language. In fact I believe that the achieving of Artificial Intelligence will be in the form of a context independent operating a parallel operating system. I believe that it is possible to build a Von Neumann machine and research should be done on this, along with research on textual AI.

Ian’s website is here