Overview
********

Selkie is a set of applications and a software
library to support **language digitization**, which is to say,
**computational documentary linguistics**.
By *documentary linguistics* I mean the combination of language
documentation and language description. Conventionally, the product of
the former is a corpus, and the product of the latter is a grammar and
lexicon. The products of language digitization are electronic versions
of corpus, lexicon, and grammar, integrated with each other and
supporting additional computational functionality, such as automated
interpretation.

Selkie is experimental code, not a finished product. Much of it is
under active development and is likely to change in the
future. Similarly, this documentation is still in draft form.
I am making it publicly available to give easier access to
students and anyone else who may be interested.

The two major pieces of Selkie are a **corpus editor** (an application for
documentary linguistics), and a natural-language processing (NLP) pipeline. The pipeline is not
currently integrated into the editor, though that is intended.

Installation
============

Selkie is written in Python, and is installed in the usual way::

    $ pip install selkie

The components of Selkie
========================

**Corpus Editor.**
A central piece of Selkie is an application for creating, editing, and
viewing linguistic corpora. The currently working version is called
CLD ("Computational Language Documentation").

**Next Editor.** The corpus editor is being redesigned. The new
version is still fragmentary.

**The NLP pipeline.**
Viewed at the highest level, a language is a relation between sentences and
meanings. Accordingly, a central piece of functionality is an
implementation of that relation, in the form of an NLP
pipeline that translates natural
language sentences to a formal semantic representation. The pipeline
is currently unidirectional, but the inverse, generating natural
language from a representation of the meaning, is under
development. The pipeline is encapsulated in a conversational agent
application (selkie.bot), which includes a knowledge base and
automated reasoner as a stand-in for the rest of cognition.
The bot is intended as a testing ground for a grammar as
digital language description.

**Dataset interfaces.**
Convenience interfaces to a variety of third-party datasets
are provided, to make it easier to access them and use them for
language description.

**Supporting Packages.**
The supporting packages contain lowlevel functionality that is used across the
components of Selkie.

Credits
=======

The following icons in selkie/data/seal are from
gnome-icon-theme-3.12.0:

 * cog.png
 * gnome/16x16/categories/applications-system.png

The files in selkie/data/census were downloaded from http://www.census.gov/genealogy/names/,
though that is now a dead link. (One may instead use
http://web.archive.org/web/19970617171355/http://www.census.gov/genealogy/names/.)

The files in selkie/data/iso were downloaded from:

 * https://www.loc.gov/standards/iso639-2/ascii_8bits.html (ISO 639-2)
 * https://iso639-3.sil.org/ (ISO 639-3)
 * https://www.loc.gov/standards/iso639-5/index.html (ISO 639-5)

The files in selkie/data/conll/2006/universal-pos-tags were
downloaded from https://github.com/slavpetrov/universal-pos-tags.