⚠️ How to to upgrade to grewpy 0.7

Version 0.7 introduced some changes, which are described here. Feel free to post an issue here if you need help with the new version.

The modules Matchings and Observations are removed.

Clustering parameters in search and count (module Corpus)

The clustering mechanim for the functions search and count were previously accessible with two arguments named clustering_parameter and clustering_keys. The first is removed and only clustering_keys is now available.

Example

We look at the relation comp:cleft in SUD_French-GSD@2.16 and want to observe the POS of the governor and dependent of this relation. We suppose that a local folder named SUD_French-GSD@2.16 is available in the current directory (data can be downloaded from GitHub).

from grewpy import Corpus, Request
corpus = Corpus("SUD_French-GSD@2.16")
request = Request ("pattern { X -[comp:cleft]-> Y }")
print (corpus.count (request, clustering_keys=["X.upos", "Y.upos"]))

returns

{('AUX', 'VERB'): 71, ('AUX', 'SCONJ'): 117, ('AUX', 'AUX'): 24}

The output is a “flat” dictionary where keys are pair for each possible value of the two feature X.upos and Y.upos. More generally, if clustering_keys is a n-tuple, the output dictionary contains n-tuple keys.

It is possible to get a recursive dictionary with the flat=False argument. This corresponds to the default behavior before grewpy version 0.7 and with grew grep (see doc).

from grewpy import Corpus, Request
corpus = Corpus("SUD_French-GSD@2.16")
request = Request ("pattern { X -[comp:cleft]-> Y }")
print (corpus.count (request, flat=False, clustering_keys=["X.upos", "Y.upos"]))

{'AUX': {'VERB': 71, 'SCONJ': 117, 'AUX': 24}}