Upload
leo-valdez
View
14
Download
0
Embed Size (px)
DESCRIPTION
What Linguists Want. (we think) Helen Aristar Dry & Anthony Aristar LINGUIST List & E-MELD. Language Documentation Used. Research: Historical / comparative Ling Typology Language description Phonology & phonetics Syntax Psycholinguistics Discourse Analysis Anthropological linguistics - PowerPoint PPT Presentation
Citation preview
What Linguists Want
(we think)
Helen Aristar Dry & Anthony Aristar
LINGUIST List & E-MELD
Language Documentation Used
• Research:• Historical / comparative Ling• Typology• Language description • Phonology & phonetics• Syntax• Psycholinguistics• Discourse Analysis• Anthropological linguistics• Ethnomusicology
• Teaching of all of the above
So they want• Access
• Central index of available material that supports flexible searching
• Ability to preview material• Clear indication of access rights • Fast permissions (24-hour turnaround)
• Stability• Cited versions of resources still available• Assembled sub-corpora available for a
specified period of time, e.g., for the duration of a course
• Ease of use • Single interface — things work the
same way in different archives (hard to misunderestimate the
technical skill of academics)• Registration that persists—i.e., they
don’t have to keep filling out registration forms
These desiderata addressed in Scenarios 4 and 5
And they would like• Ability to manipulate the data
• To annotate corpus & share annotations with co-researchers
• To track their own annotations & additions (as opposed to those of others)
• To use a concordance program or other text processing program on the corpus
• To extract relevant portions of texts and create a sub-sub-corpus; to share this sub-corpus with co-researchers or students
They would REALLY like
• Ability to identify resources by searching for linguistic structures, e.g.• Morphosyntactic categories (classifiers)• Morphosyntactic features (paucal)• Phonetic features (nasalization)*• Supersegmentals (tone)*
• E.g. to search, not just the metadata, but the annotations and transcriptions of the archived material.
*transcriptions, not sound — though search by sound would be even better
Structures central to:
• Research:• Historical / comparative Ling• Typology• Language description • Phonology & phonetics• Syntax
• Teaching of all of the above
Want to answer Qs like:
• Do all IE languages have a contrast between voiced and unvoiced consonants?
• Which languages have a distinction between trial and paucal number?
• Where can I find examples of voiceless nasals (e.g., for a phonology problem)?
Need to search for…
• Morphemes representing morphosyntactic categories and features
• Phonetic segments • Co-occurrences of segments,
categories, & features
Need to search by
• Language families and subgroups• Feature classes (e.g. “stops”, not
[ b ] )• Morphosyntactic concepts (not just
terminology, as this varies)
Requires enhanced
•Documentation •Meta-information •Search tools
Documentation
•Complete & transparent phonetic transcription
•Detailed & transparent morphosyntactic annotation
•Unambiguous language identification & classification
Meta-Information
• Unambiguous language identification system (language codes)
• Language classification system, organizing languages into families and subgroups
• Structured (graphic) taxonomy of phonetic features
Meta-Information• Structured taxonomy of
morphosyntactic categories and features (concepts and definitions)
• Lists of morphosyntactic terminology in use by various groups
• Mapping of the different terminology sets to the concepts and definitions
Search tools that can
• Interpret meta-information • Use it to construct intelligent
searches• Search
• Annotation & Transcription• OR Language profiles• OR Annotation indexes
What we have
• New Documentation• Audio / video recordings w/ translation• Phonetic transcription• Little morphosyntactic annotation (sometimes)
• Legacy documentation • Detailed morphosyntactic annotation• Complete phonetic transcription• Non-transparent (idiosyncratic) markup• Inaccessible format (e.g., paper)
What we have
• Meta-information• Ontology of morphosyntactic
concepts (GOLD —and others?)• Terminology sets (DatCat Registry)• Ontology of phonetic features• Language codes & associated family
trees (Ethnologue based)
What we have
• Search• Prototype search of phonetic transcription
using ontology of phonetic features, e.g. “Find all voiceless stops.”
• Steps toward search of morphosyntactic features:
• Language profiles which give the morphosyntactic categories and features used in a language (in XML)
• Conversion path for • mapping idiosyncratic markup to the GOLD ontology
(metaschemas + XSLT)• Converting GOLD compliant markup into RDF for
searching via semantic web
What we have: Tools
• For ontology-based morphosyntactic annotation • OntoElan (MPI’s Elan + ontology-
based terminology mapper)• OntoGloss (ontology-aware stand-off
annotation of web documents)
• For creating language profiles• FIELD
What we need
•Comprehensive, integrated system that supports this kind of searching
•“Architecture, not just tools”