How to programmatically get a list of all dependency parse tags of a loaded model?

[fhamborg]

The info is statically described in the documentation, i.e., https://spacy.io/api/annotation#dependency-parsing, but what if one needs to get the list of all dependency parse tags programmatically, e.g., in case different models are employed in a system? Is there an internal structure (that probably comes with / is set when loading a model) that can be used?

[adrianeboyd]

For the core components, you can access .labels:

import spacy
nlp = spacy.load("en_core_web_sm")
assert nlp.parser.labels == ('ROOT', 'acl', 'acomp', 'advcl', 'advmod', 'agent', 'amod', 'appos', 'attr', 'aux', 'auxpass', 'case', 'cc', 'ccomp', 'compound', 'conj', 'csubj', 'csubjpass', 'dative', 'dep', 'det', 'dobj', 'expl', 'intj', 'mark', 'meta', 'neg', 'nmod', 'npadvmod', 'nsubj', 'nsubjpass', 'nummod', 'oprd', 'parataxis', 'pcomp', 'pobj', 'poss', 'preconj', 'predet', 'prep', 'prt', 'punct', 'quantmod', 'relcl', 'xcomp')
assert nlp.get_pipe("parser").labels == ('ROOT', 'acl', 'acomp', 'advcl', 'advmod', 'agent', 'amod', 'appos', 'attr', 'aux', 'auxpass', 'case', 'cc', 'ccomp', 'compound', 'conj', 'csubj', 'csubjpass', 'dative', 'dep', 'det', 'dobj', 'expl', 'intj', 'mark', 'meta', 'neg', 'nmod', 'npadvmod', 'nsubj', 'nsubjpass', 'nummod', 'oprd', 'parataxis', 'pcomp', 'pobj', 'poss', 'preconj', 'predet', 'prep', 'prt', 'punct', 'quantmod', 'relcl', 'xcomp')

Likewise nlp.tagger or nlp.get_pipe("tagger") and nlp.entity or nlp.get_pipe("ner").

[fhamborg]

Strange, when I run this code I get an error:

import spacy

nlp = spacy.load("en_core_web_lg")
print(nlp.parser.labels)

will result in

/Users/felix/anaconda3/envs/ctsacuda/bin/python /Users/felix/IdeaProjects/newstsc/dependencyparser.py
Traceback (most recent call last):
  File "/Users/felix/IdeaProjects/newstsc/dependencyparser.py", line 16, in <module>
    print(nlp.parser.labels)
  File "pipes.pyx", line 1022, in genexpr
  File "pipes.pyx", line 1022, in genexpr
IndexError: list index out of range

Process finished with exit code 1

[adrianeboyd]

Hmm, I can't reproduce this. Could you provide more information about your environment (spacy info --markdown) and spacy model versions (spacy validate)?

[fhamborg]

Sure, please see below:

Info about spaCy

• spaCy version: 2.1.0
• Platform: Darwin-19.3.0-x86_64-i386-64bit
• Python version: 3.7.6

(ctsacuda) ➜  ~ spacy validate
✔ Loaded compatibility table

====================== Installed models (spaCy v2.1.0) ======================
ℹ spaCy installation:
/Users/felix/anaconda3/envs/ctsacuda/lib/python3.7/site-packages/spacy

TYPE      NAME             MODEL            VERSION
package   en-core-web-lg   en_core_web_lg   2.1.0   ✔

[adrianeboyd]

That looks like a bug (#4275) that didn't get fixed until v2.2.0. Unless you have a particular reason for using v2.1, I'd recommending upgrading to a more recent version (v2.2+).

If you specifically need v2.1 (and even then I'd recommend upgrading to at least v2.1.8), you have to work around that bug, but the information is still available:

labels = set()
for move in nlp.parser.move_names:
    if "-" in move:
        move = move.split("-")[1]
        if "||" in move:
            move = move.split("||")[1]
        labels.add(move)

assert labels == {'agent', 'acl', 'intj', 'predet', 'expl', 'prt', 'meta', 'attr', 'pcomp', 'parataxis', 'compound', 'auxpass', 'mark', 'prep', 'dative', 'xcomp', 'npadvmod', 'advcl', 'preconj', 'pobj', 'csubj', 'dep', 'punct', 'case', 'ROOT', 'quantmod', 'cc', 'det', 'subtok', 'neg', 'nmod', 'amod', 'advmod', 'nsubjpass', 'appos', 'conj', 'dobj', 'acomp', 'nsubj', 'csubjpass', 'nummod', 'ccomp', 'oprd', 'poss', 'relcl', 'aux'}

[lock[bot]]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.