Despite tremendous scientific progress over the past fifty or so years, there still seems to be a long way to go before we can reach a comprehensive explanation of human spoken language behaviour and can create a speech technology with performance approaching or exceeding that of a human being. It is my belief that progress is hampered by the fragmentation of spoken language research across many different disciplines, coupled with a failure to create an integrated view of the fundamental mechanisms that underpin one organism’s ability to interact with another.
In fact I would argue that "spoken language is the most sophisticated behaviour of the most complex organism in the known universe", and that we have grossly underestimated the amazing ability that human being's have evolved for communicating with each other. As a consequence, we have completely failed to realise the significance of spoken language as a topic of fundamental scientific investigation that could provide a unique window into the intricate workings of the human mind. Spoken language should sit alongside particle physics (the science of the infinitessimally small) and cosmology (the science of the infinitely large) as one of the most important scientific research topics of the current era. It is only the illusion of the ease with which we acquire and use spoken language that blinds us to its truly fantastic nature.
Subscribe to:
Post Comments (Atom)
10 comments:
But isn't language but the expression of cognition? And don't all animals with eyes and ears have the ability to demonstrate cognitive understanding? Me thinks that the study of language (the protocol) in itself would be far less rewarding than the study of cognition (the real understanding).
In fact my quote was inspired by remarks made by Richard Dawkins (in 'The Blind Watchmaker') and Gopnik, Meltzoff and Kuhl (in 'The Scientist in the Crib'). For sure, cognition is the core behaviour of a living organism, creating and maintaining coherent structural representations of both its internal and external worlds, whilst simultaneously managing the interactions with the external environments (which, of course, may include other organisms). However, I think it is time to step beyond the view that language is simply a "protocol" for interfacing cognition to the outside world. I think it is more profitable to acknowledge the intimate relationship between cognition and language - indeed, to use your words, to view language as an "expression of cognition" that has evolved to provide human beings with a rich capacity for abstract thought and imagination and to able to synchronise such behaviours within and between individuals.
Do you partly subscribe to a Wharfian view, then? That is to say that our thought is influenced by our language?
I can see what you're saying about their being an intimate relationship between language and cognition but I find it hard to completely abondon the idea that language is not only a 'protocol' but an imperfect one at that. How often do we know exactly what we wish to express but are not eloquent enough to express it as well as we would like? We sometimes are incapable of this because of lack of linguistic experience but more often than not because the limitations of the language we are communicating in only allows an approximation of the concepts we wish to convey.
Now imagine if we could directly communicate our thoughts and feelings without the need to translate them first into an imperfect protocol. A computer system able to understand at this level would not only be hands free and eyes free but also 'lips and vocal chords free' ;-) However, I fear that not even Star Trek had one of those devices.
Anyway, seriously, I think the really inviting thing about representing cognitive understanding is that cognitive understanding is what is truly universal to all people of all nations independent of their language and that if we are ever to reach a perfect language independent machine translation architecture it would make use of a cognitive represenation of understanding of the source text as the starting point for generation of the source text.
This is why I feel that the study of cognitive understanding is as important, if not more so, than the study of language. Although, I agree with your sentiments that language is not given enough attention by the scientific community. Language is a window into the workings of our minds.
Not only do I sign up to a Whorfian view, but I would go further and suggest that the same mechanism underpins both cognition and language. Language (as a communicative behaviour) is an inevitable consequence of how an organism such as ourselves interacts with the world (and other organisms within it).
Of course that part of language which is specific to a community can be regarded as a localised protocol, but this is only the surface realisation of the behaviour - language itself runs much deeper and is universal.
hi,
during your Future of Spoken Language Processing seminar (English Language and Linguistics)yesterday, you stressed the importance of the intimate relationship between cognition and language and how linguists need to recognise that language is an expression of cognition rather than simply a protocol for interfacing cognition to the outside world.
Which linguists are you talking about? Such criticism seems apt for those linguists who subscribe to a reductionist (Chomskyan) account of language. But your comments made me wonder whether you were aware of the work of linguists such as Langacker and Lakoff (to name just two) who are/were/have been formulating theories over the past 20+ years which do attempt to capture this intimate relationship between cognition and language.(in fact there is a whole movement ( with its own journal "Cognitive Linguistics") which is home to a variety of linguists dedicated to formulating a cognitively realistic linguistic theory).
Of particular note (re: yesterday's seminar), George Lakoff has worked with Vittorio Gallese (of mirror neurons fame) trying to develop a Neural Theory of Language. Unfortunately, there wasn't enough time for you to talk about mirror neurons in any depth, so I've no idea if Lakoff and Gallese's work is of any interest to you.
It may also be possible that I misunderstood your comments, and that you are infact aware of such linguists but feel that they also fall short in accounting for the intimate relationship between cognition and language.
PS. I really enjoyed your seminar; it's a shame you didn't have time to talk about mirror neurons and also to explain in some detail how your model works.
Hi,
Regarding my talk, I don't believe that I said that linguists view langage as a communications protocol separate fron cognition. What I meant to say is that some individuals take that view. In fact I was referring to the discussion with 'anonymous' above about the distinction (or not) between language and thought.
Indeed I've been following some of the research relating to a neural theory of language, and I was an invited speaker at a recent workshop in Italy (organised by some of the mirror neuron people) on the topic "Is a Neural Theory of Language Possible" - see http://web.unife.it/progetti/neurolab/lecce/ However, I'd missed the Lakoff and Gallese work, so thanks for the pointer.
Just to make things clearer for people who may read this discussion and not really understand what we are talking about, the reference to a 'Whorfian' view is a reference to 'textbook psychology' where the debate is: 'Thought is influenced by language' versus 'Language is influenced by thought'. The Whorfian view is that thought is influenced by language. An extreme Whorfist believes that human thought is completely constrained and shaped by language and the classic example used by Whorf, the pschologist who formalised the school of thought, is that of the Eskimos who had several words for different kinds of snow and that their cognitive conception of snow is shaped by their language i.e. they do not see snow, they see one of several types of snow.
The Whorfian view, was for a long time, the mainstream one but has been long discredited. Many languages contain a rich vocabulary for snow. In English we can talk about snow, sleet, hail, slush and if you are a snowboarder 'powder' is the gift of the gods. But whether your first language has a rich vocabulary in that area or not it does not affect your innate ability to cognitively percieve the difference between different kinds of snow and to learn new labels or even make them up e.g. 'powder' for snowboarders.
It is much easier to demonstrate the opposite view that 'language is influenced by thought'. An English person's idea of 'bread' is probably a load of Warburton's or some other preconfectioned and ready sliced chemically concoctive imitation of what 'bread' really is. The French will typically have another cognitive idea of 'bread' as will the Italians. Our ideas of what 'bread' is are the culmination of our cognitive experiences with objects of type 'bread' and continues to change as the breadth of our experiences grows.
Unlike language, cognition is common to all animals and a sparrow's lack of language is no impedence to its cognitively learning that young human's are inherently bad intentioned and that it would be a good idea to fly away if one approaches.
However, both extreme views, 'language is influenced by thought' and 'thought is influenced by language' are ridiculous when taken to the extreme. The reality is that the linguistic and cognitive development of humans is more likely to be a mixture of the two with the former being of most importance in early stages of development.
Anyway, back to the discussion. I agree that language/cognition is far more complex than micro or macro physics. The complexity of the brain is frighteningly immense when compared with the simplicity of the workings of an atom or the predicable forces of gravity that hold the galaxies together and make them dance around a bit. They're child's play, really, before the magnitude and depth of the study of the inner workings of the human brain, or the brain of any other animal for that matter.
However, I feel the need to reiterate the fact that language is merely an imperfect protocol, but a tiny window into the depths of our cognition. Our cognitive ability to understand events in the physical universe is limitless. Our linguistic ability to express those events has much poorer performance.
I dare say that if bees were the most complex organism of the planet they would make boast that their intelligent little dances that tell the other bees where the honey dust is puts them above the other species in their ability to express cognitive events.
If our linguistic abilities were so advanced we wouldn't so often hear other people or even ourselves making statements of the kind 'Words cannot describe...', "I'm speechless!" or "How can I explain it to you?".
The harsh reality is that our language is an approximation of our cognitive experiences and when we say "Can I have a cup of tea?" we hope that the listener's cognitive association of the phrase is similar to our own or we end up making amusing questions like "Where is the milk?" as a not very travelled friend of mine once said when we were in the Ukraine. I'll never forget the look on my Ukrainian friends faces when I explained to them that the English put milk in their tea. You'd think I'd just told them he likes sand in his mayonaise (forgive the spelling).
I'm glad you mentioned abstractness. Because it is a delightful example of how our powers of cognition far outreach our linguistic abilities - especially in the realm of learning.
If I were to take a 3 year old child and tell him that in Ukrainian the object, when a personal pronoun, commonly comes before the verb in Ukrainian he would probably look at me like I was some kind of martian speaking in a funny kind of Jupiter dialect that only Plutonians are familiar with. Even if I were to sit with him and explain what an object is and a personal pronoun he probably wouldn't understand unless I illustrated with concrete examples.
Even once he had finally understood what I am talking about just 'understanding' wouldn't be enough for the child to go away and put the rule into practice and become a fluent Ukrainian speaker.
However, if I were to put him in the Ukraine and to teach him with examples like 'ya tebe lublu (I love you)' or 'ya tebe chekayu (I'll wait for you)' his innate learning circuitry would take the concrete examples and generalise (abstractise) the rule for future application. And so from a cognitive point of view it seems clear to me that there are levels of understanding (perhaps this is what you are referring to when you mention surface and deep). That on the surface the child understands that words which express cognitive experiences. At a deeper level he cognitively understands these concepts as he is able to translate them into some universal cognitive language. And at yet a deeper still level he is able to abstractise patterns from these examples of good sentences for future application even though he lacks the linguistic ability to express such generalisations or even to understand explanations of them.
I know these examples may seem contrived as they are limited to those of a young child but we see similar patterns of behaviour even in fully developed adults. In the Machine Learning classes, for example, algorithms are explained in abstract terms, at first. And at this point everbody seems to stare with glazed eyes indicative of lack of understanding. But as the abstract is given meaning by walking through the algorithms with concrete examples people slowly start to nod and have facial expressions that say 'Ah! Now I understand.'
The human cognitive ability to abstractise from concrete examples is, therefore, far more powerful than their ability to comprehend the abstract from direct abstract explanations. The psychological reasons for which are deeply routed in our innate and universal cognitive abilities and the order of their development.
I hope my extended waffle, has in places made some sense to somebody.
Cheerio!
Hi, we lead our own research in using language as the primary entity of the robotic control paradygm implementations. This leads us potentially to the system that can be programmed and queried with language. We do use semantic spaces fields and artificial neural networks as connectors in these fields. Can we cooperate, say, in European Framework Programme 7 or in other ways?
I am always interested in potential collaborations. See my web page at http://www.dcs.shef.ac.uk/~roger/ for details of how to contact me direct.
I found Watanabe's book "Pattern Recognition: Human and Mechanical" (1985) very illuminating, although it does not deal primarily with speech. Some more formal parts are a bit clumsy to me, but the discussion is always interesting and imo deserves a reading.
Post a Comment