LINGUIST List 6.719

Wed 24 May 1995

Sum: Can NLP help maintaining language diversity?

Editor for this issue: <>

Directory

1. T M Ellison, Summary: Can NLP help maintaining language diversity?

In an earlier article, I made the following proposition:

) P=``Natural language processing (NLP) tools can assist in slowing, if not
) halting, the slide of individual languages towards extinction.''

and asked some questions about it:

) Q1. Is P true? Can (any, some or all) NLP tools help keep languages
) alive? If not, is there any role for technology in maintaining
) language diversity?
)
) Questions 2 and 3 are predicated on P being true.
)
) Q2. Are there any NLP tools which have had a positive impact on the
) survival of a language?
)
) Q3. Which new tools (feel free to make them up, within reason) would
) be of greatest assistance?

There were four replies, which I summarise. I'd like to thank the
respondents. The closest thing to a common theme one could draw from
the replies was that basic information technology, word-processing and
communications were the most vital to language maintenance.

(1) Kevin Donnelly writes that using self-constructed NLP tools helped
him to learn and use Irish Gaelic. The tool consisted of
spell-checking and thesaurus instantly available to check spelling or
gender or inflection.

(2) Thierry van Steenberghe: for many languages is too late for NLP
because narrow group of speakers who often would not be likely to use
NLP tools even if they were available. Any language with no NLP tools
will come into risk of extinction. Comparison with introduction of
printing. Important language tools: basic word-processing (eg fonts)
and support by operating systems and telecommunications.

Tools to help languages avoid threat from lack of NLP:

- Operating systems and telecom protocols with correct [ISO-10646 or
at least Unicode compliant] languages support;
- Resources as dictionaries of all types, including bi/multi-lingual,
morphological and synactic modules, corpora; note that this a
condition towards the development of other NLP tools.
- Localised applications: too few apps are readily available in other
languages than their original [often US] version. Specially true for
non mainstream business apps.
- Multilingual applications: instead of unilingual localised versions
[at best, now], ML apps would allow to change language 'on the fly' or
at least at the installation time.

(3) Stavros Macrakis
Basic text processing tools. `Unicode/ISO 10646 is a first step at the
character code level, but is not sufficient in the absence of suitable
input and output methods. Email and netnews can support dispersed
speaker groups. No serious evidence for this: except spelling
checkers reduce quality of spelling and proofreading.

(4) Bruce Connell writes
P is false. NLP tools assume some standard dialect of a
language. Their use may endanger other dialects. Thus lanugage
diversity is decreased. Fieldwork to record these should be done
first.

------
Regards and thanks for your attention,
Mark Ellison