LINGUIST List 14.1745

Fri Jun 20 2003

FYI: Transcription/Speech Processing; Lang Evolution

Editor for this issue: Marie Klopfenstein <marielinguistlist.org>

Directory

1. Magali Jeanmaire, ELRA news - ESTER campaign
2. lastmanthere, Chimps, genomes, and language

****************************************************************************
 Call for Participation:
****************************************************************************

Introduction:
- -----------

The aim of the ESTER campaign (l'Evaluation des Syst�mes de
 Transcription enrichie d'Emissions Radiophoniques / Evaluation of
automatic broadcast news transcription systems) is to promote the
evaluation of speech processing systems for the French language, to
establish a permanent evaluation infrastructure and to disseminate, as
widely as possible, the information and the resources acquired in the
campaign. The ESTER campaign forms part of the EVALDA project and is
 funded by the French Ministry of Research in the context of its
Technolangue programme.

In the context of the ESTER campaign, the transcriptions are to be
annotated with additional and associated information such as speaker
turns, named entities etc. On the one hand, such an enriched
transcription aims to provide an orthographic transcription of an
audio signal and on the other hand, a structured representation of an
audio document enabling information extraction from audio files. The
aim of evaluating the quality of the associated annotations, along
with the evaluation of the orthographic transcriptions, is to
establish a reference or benchmark of present performance levels of
each component of an indexation system whilst also providing an idea
of overall system performance.

Participation:
- ------------

Research and development centres, public or private, wishing to take
part in the ESTER campaign are invited to make themselves known to the
campaign organisers (see below) in order to register as
participants. On registering as a participant and signing an end-user
contract, you will receive the data available for Phase 1 of the
campaign.

The contract will be sent to you after declaring your interest in the
campaign.

Those participating in the ESTER campaign will receive the
training/development/test data free of charge (providing results
obtained during the evaluation campaign are returned). The data
provisionally available for distribution are listed below:

- Le Monde: 1997- 2002
- MLCC: Transcriptions of debates from the European Parliament
(1992-1994)
- France Inter: 25 hours of transcribed broadcast news
- Radio France International (RFI): 15 hours of transcribed broadcast
news

Participants that return results for the obligatory evaluation tasks
are permitted to keep the data mentioned above, for no additional
cost.

For more information on the project and the timetable, please consult
the project website: http://www.afcp-parole.org/ester.

Organisers
- -----------

The ESTER campaign forms part of the EVALDA project under the aegis of
the Association Francophone de la Communication Parl�e (AFCP) with the
support of the D�l�gation G�n�rale de l'Armement (DGA) and ELDA
(Evaluation and Language resources Distribution Agency).

Contacts :
- ----------

Guillaume Gravier, IRISA (ggravieririsa.fr)
Jean-Fran�ois Bonastre, AFCP (Jean-francois.bonastrelia.univ-avignon.fr)
Edouard Geoffrois, DGA/CTA (edouard.geoffroisetca.fr)
Kevin McTait, ELDA (mctaitelda.fr)

The chimp genome
Carl Gierstorfer

http://www.spectator.co.uk/article.php3?table=old&section=current&issue=2003-06-14&id=3201

Now let us take another example the evolution of language. It is
surely central to the evolution of culture. The fact that language
must have a genetic basis had already been recognised by the American
linguist Noam Chomsky back in the 1960s.

For Chomsky, it simply seemed unrealistic to assume that a child of
four years of age can string together hundreds of words using complex
rules of grammar without there being a genetic factor at work. His
so-called "poverty of stimulus argument" is nicely illustrated by the
rise of Creole languages among the descendants of slaves. Initially,
different peoples speaking different tongues were forced to
communicate by sign language. Yet the children of the slaves already
began to invent their own language, using a primitive grammar. Over
the generations, without any teaching or dictionaries, the pidgin
dialects arose, and finally the Creole languages. They are languages
in their own right, based on English or French but with their own
complex grammar.

Accounts of the descent of man never explain why such as chimpanzees
remained static, while a subset, man, emerged in a long range
directional emergence. To say that this is the result of a special
adaptational context is vacuous, what context.

While it is hardly surprising that we should discover a genetic basis
to language, we are still in the dark as to how that genetic basis
evolved. Darwinian accounts are in a strange limbo where the
mutational explanation retreated in the face of the discovery of
developmental genetics (although textbooks never say thsi), and the
selectionist account is then applied to the surely improbable
emergence of developmental sequences.

The tacit assumption that selectionist evolution has done all this in
the time frame of several hundred thousand years stretches credulity,
even in the age of hox genes and developmental genetics.

One of the ambiguities of information theory is the question of
meaning, which is more than information. Can we really assume without
proof, first that a developmental process as complex as seen in
children evolved at random as blind chance to produce the ultra-rich
meaning communication of man?

The account of the descent of man by Darwinists, as the case of
linguistics makes clear, has cast a spell on biologists and the result
is a myth, not yet science.

The reason for these remarks is to reiterate a demonstration
historically of a developmental sequence that isn't genetic, the eonic
effect, (see link) and to ask if the emergence of language as blind
mutational advance, evolution without meaning, is not contradicted by
the macroevolution visible in history and whose evolutionary
interaction reaches the level of complex linguistic art forms?

One thinks also of a work such as W. MacNeill's Keeping Together in
Time claiming that dance and song are built into the evolutionary
process.

The point being that a holistic account transcending the genetic needs
to be tabled if only to the degree that we are under no obligation to
take current theories as established with sufficient evidence.

John Landon
Website for
World History and the Eonic Effect
http://eonix.8m.com