Note: For a full report on this tagger's accuracy, see "CLTK POS tagging cross–validation".
This repository contains treebanks for Ancient Greek from the Ancient Greek Dependency Treebank, version 1.7.
The file make_pos_training_set.py generates the training set pos_training_set.txt, which encodes part of speech data like so:
κάτοιδα/v1sria--- νυκτέρων/a-p---ng- ὁμήγυριν/n-s---fa- ,/u-------- καὶ/c-------- τοὺς/l-p---ma- φέροντας/t-pppama- χεῖμα/n-s---na- καὶ/c-------- θέρος/n-s---na- βροτοῖς/n-p---md- λαμπροὺς/a-p---ma- δυνάστας/n-p---ma- ,/--------- ἐμπρέποντας/t-pppama- αἰθέρι/n-s---md- [ἀστέρας/n-p---ma- ,/--------- ὅταν/c-------- φθίνωσιν/v3ppsa--- ,/--------- ἀντολάς/n-p---fa- τε/g-------- τῶν]/p-p---mg- ./---------pos_training_set.txt is made for the purpose of being used with NLTK's TaggedCorpusReader.
Example custom training:
In [1]: from nltk.corpus.reader import TaggedCorpusReader
In [2]: from nltk.tag import UnigramTagger
In [3]: from nltk.tokenize import wordpunct_tokenize
In [4]: reader = TaggedCorpusReader('.', r'.*\.pos')
In [5]: train_sents = reader.tagged_sents()
In [6]: tagger = UnigramTagger(train_sents)
In [7]: untagged_text = """θεοὺς μὲν αἰτῶ τῶνδ᾽ ἀπαλλαγὴν πόνων φρουρᾶς ἐτείας μῆκος, ἣν κοιμώμενος στέγαις Ἀτρειδῶν ἄγκαθεν, κυνὸς δίκην, ἄστρων κάτοιδα νυκτέρων ὁμήγυριν, καὶ τοὺς φέροντας χεῖμα καὶ θέρος βροτοῖς λαμπροὺς δυνάστας, ἐμπρέποντας αἰθέρι ἀστέρας, ὅταν φθίνωσιν, ἀντολάς τε τῶν."""
In [8]: untagged_tokens = wordpunct_tokenize(untagged_text)
In [9]: tagger.tag(untagged_tokens)
Out[9]:
[('θεοὺς', 'N-P---MA-'),
('μὲν', 'G--------'),
('αἰτῶ', 'V1SPIA---'),
('τῶνδ', None),
('᾽', None),
('ἀπαλλαγὴν', 'N-S---FA-'),
('πόνων', 'N-P---MG-'),
('φρουρᾶς', 'N-S---FG-'),
('ἐτείας', 'A-S---FG-'),
('μῆκος', 'N-S---NA-'),
(',', 'U--------'),
('ἣν', 'P-S---FA-'),
('κοιμώμενος', 'T-SPPEMN-'),
('στέγαις', 'N-P---FD-'),
('Ἀτρειδῶν', 'N-P---MG-'),
('ἄγκαθεν', 'D--------'),
(',', 'U--------'),
('κυνὸς', 'N-S---FG-'),
('δίκην', 'N-S---FA-'),
(',', 'U--------'),
('ἄστρων', 'N-P---NG-'),
('κάτοιδα', 'V1SRIA---'),
('νυκτέρων', 'A-P---NG-'),
('ὁμήγυριν', 'N-S---FA-'),
(',', 'U--------'),
('καὶ', 'C--------'),
('τοὺς', 'P-P---MA-'),
('φέροντας', 'T-PPPAMA-'),
('χεῖμα', 'N-S---NA-'),
('καὶ', 'C--------'),
('θέρος', 'N-S---NA-'),
('βροτοῖς', 'N-P---MD-'),
('λαμπροὺς', 'A-P---MA-'),
('δυνάστας', 'N-P---MA-'),
(',', 'U--------'),
('ἐμπρέποντας', 'T-PPPAMA-'),
('αἰθέρι', 'N-S---MD-'),
('ἀστέρας', None),
(',', 'U--------'),
('ὅταν', 'C--------'),
('φθίνωσιν', 'V3PPSA---'),
(',', 'U--------'),
('ἀντολάς', 'N-P---FA-'),
('τε', 'G--------'),
('τῶν', 'L-P---MG-'),
('.', 'U--------')]
In [10]: tagger.evaluate(train_sents)
Out[10]: 0.91961233400652131. Preamble
1.1 Source
The Ancient Greek Dependency Treebank is available at:
http://nlp.perseus.tufts.edu/syntax/treebank/
1.2 License
AGDT 1.7 is licensed under a Creative Commons Attribution-
NonCommercial-ShareAlike 2.5 License:
http://creativecommons.org/licenses/by-nc-sa/2.5
2. Documentation
2.1 Data Format
The data given in this treebank is provided as an XML document. Each
word contains six required attributes:
id: This is a unique identifier, and corresponds to the word's linear
position in the sentence. The first word in a sentence is given
id 1.
cid: This is a canonical identifier for the word within the larger corpus.
form: The token form of the word, in Beta Code.
lemma: The base lemma from which the word is derived, in Beta Code.
head: The id of the word's parent. If a word depends on the sentence
root, its head is 0.
relation: The syntactic relation between the word and its parent. A
catalogue of syntactic tags can be found in the syntactic guidelines
described below.
postag: The morphological analysis for the word. This field is 9
characters long, and corresponds to the following morphological
features:
1: part of speech
n noun
v verb
t participle
a adjective
d adverb
l article
g particle
c conjunction
r preposition
p pronoun
m numeral
i interjection
e exclamation
u punctuation
2: person
1 first person
2 second person
3 third person
3: number
s singular
p plural
d dual
4: tense
p present
i imperfect
r perfect
l pluperfect
t future perfect
f future
a aorist
5: mood
i indicative
s subjunctive
o optative
n infinitive
m imperative
p participle
6: voice
a active
p passive
m middle
e medio-passive
7: gender
m masculine
f feminine
n neuter
8: case
n nominative
g genitive
d dative
a accusative
v vocative
l locative
9: degree
c comparative
s superlative
---
For example, the postag for the noun "a)/ndra" is "n-s---ma-",
which corresponds to the following features:
1: n noun
2: -
3: s singular
4: -
5: -
6: -
7: m masculine
8: a accusative
9: -
2.2 Text
AGDT 1.7 is comprised of 18 texts, in the following distribution:
Hesiod, Shield of Heracles: 3,834 words
Hesiod, Theogony: 8,106 words
Hesiod, Works and Days: 6,941 words
Homer, Iliad: 128,102 words
Homer, Odyssey: 104,467 words
Aeschylus, Agamemnon: 9,806 words
Aeschylus, Eumenides: 6,380 words
Aeschylus, Libation Bearers: 6,566 words
Aeschylus, Persians: 6,270 words
Aeschylus, Prometheus Bound: 7,058 words
Aeschylus, Seven Against Thebes: 6,232 words
Aeschylus, Suppliants: 5,949 words
Sophocles, Ajax: 9,474 words
Sophocles, Antigone: 8,751 words
Sophocles, Electra: 10,489 words
Sophocles, Oedipus Tyrannus: 11,185 words
Sophocles, Trachiniae: 8,822 words
Plato, Euthyphro: 6,097 words
The editions of these texts are as follows:
Hesiod. The Homeric Hymns and Homerica with an English Translation by
Hugh G. Evelyn-White. Works and Days. Cambridge, MA.,Harvard University
Press; London, William Heinemann Ltd. 1914.
Homer. The Odyssey with an English Translation by A. T. Murray.
Cambridge, MA., Harvard University Press; London, William Heinemann,
Ltd. 1919.
Homer. Homeri Opera in five volumes. Oxford, Oxford University Press.
1920.
Aeschylus. Aeschylus, with an English translation by Herbert Weir Smyth,
Ph. D. in two volumes. Cambridge. Cambridge, Mass., Harvard University
Press; London, William Heinemann, Ltd. 1926. Including edits proposed by
by Francesco Mambrini for Libation Bearers, Persians, Prometheus Bound,
Seven Against Thebes, and Suppliant Women. 2012.
Sophocles. Sophocles. Vol 1: Oedipus the king. Oedipus at Colonus. Antigone.
With an English translation by F. Storr. The Loeb classical library, 20.
Francis Storr. London; New York. William Heinemann Ltd.; The Macmillan Company. 1912.
Including edits proposed by Francesco Mambrini. 2012.
Sophocles. Sophocles. Vol 2: Ajax. Electra. Trachiniae. Philoctetes With an
English translation by F. Storr. The Loeb classical library, 21. Francis Storr.
London; New York. William Heinemann Ltd.; The Macmillan Company. 1913.
Including edits proposed by Francesco Mambrini. 2012.
Plato. Platonis Opera, ed. John Burnet. Oxford University Press. 1903.
The following document_ids in the treebank correspond to the following
works:
Perseus:text:1999.01.0127 Hesiod (Shield of Heracles)
Perseus:text:1999.01.0129 Hesiod (Theogony)
Perseus:text:1999.01.0131 Hesiod (Works and Days)
Perseus:text:1999.01.0133 Homer (Iliad)
Perseus:text:1999.01.0135 Homer (Odyssey)
Perseus:text:1999.01.0003 Aeschylus (Agamemnon)
Perseus:text:1999.01.0005 Aeschylus (Eumenides)
Perseus:text:9999.01.0006 Aeschylus (Libation Bearers)
Perseus:text:9999.01.0008 Aeschylus (Prometheus Bound)
Perseus:text:9999.01.0009 Aeschylus (Persians)
Perseus:text:9999.01.0007 Aeschylus (Seven Against Thebes)
Perseus:text:9999.01.0010 Aeschylus (Supplian Women)
Perseus:text:1999.01.0183 Sophocles (Ajax)
Perseus:text:9999.01.0013 Sophocles (Antigone)
Perseus:text:9999.01.0012 Sophocles (Electra)
Perseus:text:1999.01.0191 Sophocles (Oedipus Tyrannus)
Perseus:text:9999.01.0011 Sophocles (Trachinae)
urn:cts:greekLang:tlg0059.tlg001.perseus-grc1 Plato (Euthyphro)
2.3 Annotation Standards
This release of the treebank has been annotated according to the
guidelines specified in version 1.1 of the "Guidelines for the Syntactic
Annotation of the Ancient Greek Dependency Treebank (1.1)," found in
docs/guidelines.pdf
2.4 Authorship
Aside from the scholarly treebanks of Aeschylus (ed., Francesco Mambrini),
Plato (ed., Giuseppe G.A. Celano) and Sophocles (ed., Dan Libatique and
ed., Francesco Mambrini), each sentence in the Ancient Greek Dependency Treebank
is built from the efforts of two independent annotators (marked "primary" in the data)
reconciled by a third (marked "secondary"). We would like to recognize the contribution
of the following individuals toward its creation and thank them for their
commitment to the advancement of Classical scholarship:
Jennifer Adams, James Artz, Jennifer Curtin, James C. D'Amico, W. B. Dolan,
Calliopi Dourou, Scott J. Dube, C. Dan Earley, J. F. Gentile, Francis Hartel,
Connor Hayden, Kenny Hickman, Tovah Keynton, Michael Kinney, Florin Leonte,
Alex Lessie, Daniel Lim Libatique, Brian Livingston, Viet Luong, Meg Luthin,
George Matthews, Molly Miller, Jack Mitchell, Skylar Neil, Robin Ngo, Jessica
Nord, Anthony D. Yates, Sam Zukoff and the Tufts University LAT-181 class
(Spring 2008).