Pattern matching with tregex in Stanzas Corenlp implementation doesn't seem to finde the right subtrees - stanford-nlp

Question

I am relatively new to NLP and at the moment I'm trying to extract different phrase scructures in german texts. For that I'm using the Stanford corenlp implementation of stanza with the tregex feature for pattern machting in trees.
So far I didn't have any problem an I was able to match simple patterns like "NPs" or "S > CS".
No I'm trying to match S nodes that are immediately dominated either by ROOT or by a CS node that is immediately dominated by ROOT. For that im using the pattern "S > (CS > TOP) | > TOP". But it seems that it doesn't work properly. I'm using the following code:
text = "Peter kommt und Paul geht."
def linguistic_units(_client, _text, _pattern):
matches = _client.tregex(_text,_pattern)
list = matches['sentences']
print('+++++Tree++++')
print(list[0])
for sentence in matches['sentences']:
for match_id in sentence:
print(sentence[match_id]['spanString'])
return count_units
with CoreNLPClient(properties='./corenlp/StanfordCoreNLP-german.properties',
annotators=['tokenize', 'ssplit', 'pos', 'lemma', 'ner', 'parse', 'depparse', 'coref'],
timeout=300000,
be_quiet=True,
endpoint='http://localhost:9001',
memory='16G') as client:
result = linguistic_units(client, text, 'S > (CS > ROOT) | > ROOT'
print(result)
In the example with the text "Peter kommt und Paul geht" the pattern I'm using should match the two phrases "Peter kommt" and "Paul geht", but it doesn't work.
Afterwards I had a look at the tree itselfe and the output of the parser was the following:
constituency parse of first sentence
child {
child {
child {
child {
child {
value: "Peter"
}
value: "PROPN"
}
child {
child {
value: "kommt"
}
value: "VERB"
}
value: "S"
}
child {
child {
value: "und"
}
value: "CCONJ"
}
child {
child {
child {
value: "Paul"
}
value: "PROPN"
}
child {
child {
value: "geht"
}
value: "VERB"
}
value: "S"
}
value: "CS"
}
child {
child {
value: "."
}
value: "PUNCT"
}
value: "NUR"
}
value: "ROOT"
score: 5466.83349609375
I now suspect that this is due to the ROOT node, since it is the last node of the tree. Should the ROOT node not be at the beginning of the tree?
Does anyone know what I am doing wrong?

Answer 1

A few comments:
1.) Assuming you are using a recent version of CoreNLP (4.0.0+), you need to use the mwt annotator with German. So your annotators list should be tokenize,ssplit,mwt,pos,parse
2.) Here is your sentence in PTB for clarity:
(ROOT
(NUR
(CS
(S (PROPN Peter) (VERB kommt))
(CCONJ und)
(S (PROPN Paul) (VERB geht)))))
As you can see the ROOT is the root node of the tree, so your pattern would not match in this sentence. I personally find the PTB format easier to see the tree structure and for writing Tregex patterns off of. You can get that via the json or text output formats (instead of the serialized object). In the client request set output_format="text"
3.) Here is the latest documentation on using the Stanza client: https://stanfordnlp.github.io/stanza/client_properties.html