Towards Linking Linguistic Resources in the Humanities

Alexander O'Connor
David Lewis, Felix Sasaki

This research is supported by:
Science Foundation Ireland (Grant 12/CE/I2267) as part of CNGL Centre for Global Intelligent Content at Trinity College Dublin.
And European Commission under the 7^th Framework Programme (CENDARI, FALCON)

Cannot make assumptions about the frame of reference.

Sources for training data

• Pre-existing Corpus Content
• Google
• Twitter
• Gutenberg, Hathi
• Wikipedia
• Newspaper

Significant pre-processing required!

Hidden Knowledge

• Questions requiring interpretation of the text
• Questions not answered by a single document
• References to other resources

Some example pit-falls from Cendari

• Language, locale, dialect, register, variant
• Short text
• References to entities (people, places, events, 'things')
• Markup ≠ Content ≠ Query
• Calendar (span and representation)
• Automatic tools
• Polyglotism

Can we put some historical documents into a package that is linked and marked up for the purpose of:

• training
• translating
• mining
• reading

Different content—Different markup

How to make it traceable?

What TEI has

• Variety (music, poetry, practice)
• Editorial/Revision
• Linguistic
• Adoption

What TEI is Missing

• Strong localisation support
• Support outside the humanities
• RDF

Vocabularies

Workflows

To Sum Up

1. Source of richer content, with richer tasks
2. Need to integrate with the overall effort to produce reusable, traceable, marked up materials for training
3. Particularly need to handle dialects, polyglot documents and secondary documents
4. Standards exist, workflows are coalescing, but the last steps need to be joined

Data = ƒ(content, markup, quality, provenance)

Get the corpus I want now, but able to get more later