Outsiders' views of the value of linguistics

Thu Oct 21 09:31:18 UTC 2010

Dear Fritz, Matti and all,

This is a very interesting question and as a person standing on both 
sides (outside and inside linguistics), I would like to comment on 
some matters.

In computational linguistics and especially in natural language 
processing research as an application orinted area of artificial 
intelligence, it has been commonplace to think that linguists 
("proper") often focus on too specific phenomena that do not help in 
increasing the overall performance of a system. This may been seen as 
the background for Jelinek's comment.

In addition to this lack of coverage issue, the representation of 
linguistic knowledge and its use in building natural language 
processing computational systems is a complex issue. The relationship 
between implicit linguistic skills and explicit representations are 
not at all as straightforward as some rule-based representation 
oriented scholars such as Noam Chomsky have tried to suggest (for 
discussion on the methodology of implicit and explicit representation, 
please see "Modeling communities of experts: Conceptual grounding of 
expertise", www.cis.hut.fi/tho/online-papers/TKK-ICS-R24.pdf - 
suggestions for journals that would be interested in this 
multidisciplinary topic are also welcome).

Charles Osgood's comment reflects a distinction that can also be seen 
between (autonomous) general linguistics and cognitive linguistics.
We who try to create models of (real or artificial) systems that learn 
and use language need to think basically of all languages: What are
the cognitive mechanisms that give rise to the ability to learn and
use language? One needs to be able to model systems that learn 
language from the input. A central research question then is
what needs to be there in a cognitive system for the learning to be 
possible. Proponents of the poverty of stimulus argument etc. have 
suggested that not so much can be learned but there is also a lot 
of opposing evident.

In our field, an interesting recent example is the development of 
Morfessor system that models certain aspects of morphology of 
basically any language through unsupervised learning and information 
theoretical principles (see http://www.cis.hut.fi/projects/morpho/ for 
details). Another, older result is the experiment in which 
we were able to exhibit emergence of linguistically motivated 
categories through the use of the self-organizing map method (the 
most popular computational model of cortical organization).
(http://www.cis.hut.fi/tho/online-papers/honkela_pulkki_kohonen_icann95_grimm.pdf)

Some potentially interesting issues related to computational modeling 
of human and social sciences are covered in a keynote paper
presented last June 
(http://www.cis.hut.fi/tho/online-papers/honkela_mashs10_final.pdf).
One quote from the paper that might be interesting from the 
point of view of Fritz' original question:

   "Computational linguistics is an area in which computers have been
   used for a relatively long time as a research tool. Linguistics can
   be considered to particularly interesting from the point of view of
   scientific practice and scientific representation because language
   is a central means for representing and communicating scientific
   results."

Summa summarum, research on language is extremely important in its all 
flavors. Research on chemisty and biology is important for the health. 
Research on language (in its widest sense including functional 
linguistic, cognitive linguistics, sociolinguistics, computational 
linguistics, etc.) is of crucial imporance on how we are able deal 
with any aspect or problem in the world. We all are so immersed in 
language that especially those who are not inside this field do not 
necessarily appreciate the importance of this field of inquiry.

Best regards,
Timo

On Thu, 21 Oct 2010, Matti Miestamo wrote:

> Dear Fritz,
>
> what about Greenberg & al. (1978: v) quoting a comment by psychologist Charles Osgood:
>
> "... while linguistics had an admirable and well worked out method, 
> it was being applied merely to the description of individual 
> languages. Could the linguists present tell him anything about 
> **all** languages? That would be of the highest interest to 
> psychologists.”
>
> Reference:
> Greenberg, Joseph H., Charles A. Ferguson & Edith A. Moravcsik. 1978. Preface.
> In Joseph H. Greenberg (ed.), Universals of Human Language, vol. 1,
> Method and Theory, v–xi. Stanford: Stanford University Press.
>
> Best wishes,
> Matti
>
> --
> Matti Miestamo
> http://www.ling.helsinki.fi/~matmies/
>
>
>
> On Oct 20, 2010, at 20:12 PM, Frederick J Newmeyer wrote:
>
>> Hello,
>>
>> For a survey article that I'm writing, I plan to assemble quotes from people outside the field of linguistics on what they see as the value, or lack of value, of work done in linguistics. So I would like to cite published quotes from psychologists, anthropologists, literary specialists, etc. on their views about the value/relevance of linguistics for their particular concerns and its value/relevance in general. Can anybody help me out by pointing me to relevant quotes?
>>
>> Let me give one example of the sort of thing that I am looking for. The late computational linguist Fred Jelinek reportedly wrote: 'Whenever I ﬁre a linguist our system performance improves'.
>>
>> Thanks. I'll summarize.
>>
>> Best wishes,
>>
>> --fritz
>>
>> fjn at u.washington.edu
>>
>>
>>
>> Frederick J. Newmeyer
>> Professor Emeritus, University of Washington
>> Adjunct Professor, University of British Columbia and Simon Fraser University
>> [for my postal address, please contact me by e-mail]
>>
>
>

--
Timo Honkela, Chief Research Scientist, PhD, Docent
Adaptive Informatics Research Center
Aalto University School of Science and Technology
P.O.Box 5400, FI-02015 TKK, Finland

timo.honkela at tkk.fi,  http://www.cis.hut.fi/tho/