spaCy/docs/redesign/spec.jade

130 lines
4.2 KiB
Plaintext

mixin columns(...names)
tr
each name in names
th= name
mixin row(...cells)
tr
each cell in cells
td= cell
details
summary: h4 Overview
p.
This document describes the target annotations spaCy is trained to predict.
This is currently a work in progress. Please ask questions on the issue tracker,
so that the answers can be integrated here to improve the documentation.
details
summary: h4 Tokenization
p Tokenization standards are based on the OntoNotes 5 corpus.
p.
The tokenizer differs from most by including tokens for significant
whitespace. Any sequence of whitespace characters beyond a single space
(' ') is included as a token. For instance:
pre.language-python
code
| from spacy.en import English
| nlp = English(parse=False)
| tokens = nlp('Some\nspaces and\ttab characters')
| print([t.orth_ for t in tokens])
p Which produces:
pre.language-python
code
| ['Some', '\n', 'spaces', ' ', 'and', '\t', 'tab', 'characters']
p.
The whitespace tokens are useful for much the same reason punctuation is
– it's often an important delimiter in the text. By preserving
it in the token output, we are able to maintain a simple alignment
between the tokens and the original string, and we ensure that no
information is lost during processing.
details
summary: h4 Sentence boundary detection
p.
Sentence boundaries are calculated from the syntactic parse tree, so
features such as punctuation and capitalisation play an important but
non-decisive role in determining the sentence boundaries. Usually this
means that the sentence boundaries will at least coincide with clause
boundaries, even given poorly punctuated text.
details
summary: h4 Part-of-speech Tagging
p.
The part-of-speech tagger uses the OntoNotes 5 version of the Penn Treebank
tag set. We also map the tags to the simpler Google Universal POS Tag set.
p.
Details here: https://github.com/honnibal/spaCy/blob/master/spacy/en/pos.pyx#L124
details
summary: h4 Lemmatization
p.
A "lemma" is the uninflected form of a word. In English, this means:
ul
li Adjectives: The form like "happy", not "happier" or "happiest"
li Adverbs: The form like "badly", not "worse" or "worst"
li Nouns: The form like "dog", not "dogs"; like "child", not "children"
li Verbs: The form like "write", not "writes", "writing", "wrote" or "written"
p.
The lemmatization data is taken from WordNet. However, we also add a
special case for pronouns: all pronouns are lemmatized to the special
token -PRON-.
details
summary: h4 Syntactic Dependency Parsing
p.
The parser is trained on data produced by the ClearNLP converter. Details
of the annotation scheme can be found here: http://www.mathcs.emory.edu/~choi/doc/clear-dependency-2012.pdf
details
summary: h4 Named Entity Recognition
table
thead
+columns("Entity Type", "Description")
tbody
+row("PERSON", "People, including fictional.")
+row("NORP", "Nationalities or religious or political groups.")
+row("FACILITY", "Buildings, airports, highways, bridges, etc.")
+row("ORG", "Companies, agencies, institutions, etc.")
+row("GPE", "Countries, cities, states.")
+row("LOC", "Non-GPE locations, mountain ranges, bodies of water.")
+row("PRODUCT", "Vehicles, weapons, foods, etc. (Not services")
+row("EVENT", "Named hurricanes, battles, wars, sports events, etc.")
+row("WORK_OF_ART", "Titles of books, songs, etc.")
+row("LAW", "Named documents made into laws")
+row("LANGUAGE", "Any named language")
p The following values are also annotated in a style similar to names:
table
thead
+columns("Entity Type", "Description")
tbody
+row("DATE", "Absolute or relative dates or periods")
+row("TIME", "Times smaller than a day")
+row("PERCENT", 'Percentage (including “%”)')
+row("MONEY", "Monetary values, including unit")
+row("QUANTITY", "Measurements, as of weight or distance")
+row("ORDINAL", 'first", "second"')
+row("CARDINAL", "Numerals that do not fall under another type")