posextract

posextract offers grammatical information extraction methods designed for the analysis of historical and contemporary textual corpora. It traverses the syntactic dependency relations between parts-of-speech and returns sequences of words that share a grammatical relationship. See our article for more. You can also download posextract for pypi with pip.

Usage

• extract_triples to extract subject-verb-object (SVO) and subject-verb-adjective complement (SVA) triples
• extract_adj_noun_pairs to extract adjective-noun pairs
• extract_subj_verb_pairs to extract subject-verb pairs

Required Paramters:

• input can be the name of a csv file or an input string
• output name of the output file

Optional Paramters:

• --data_column specify the column to extract triples from.
• --id_column specify a unique ID field if csv file is given.
• --file-delimiter specify comma, pipe, or tab. Default is comma.
• --post-combine-adj combine triples (adjective predicate with object)
• --add-auxiliary extract future and past tense triples.
• --prep-phrase extract the . Default set to false.
• --no-compound-noun Extract just the subject or object (e.g. "Indian Government" is extracted as just "Government").
• --lemma specify whether to lemmatize parts-of-speech. Default is non-lemmatized.
• --verbose print

Examples

Interactive:

Extract grammatical triples.

from posextract import grammatical_triples

triples = grammatical_triples.extract(['Landlords may exercise oppression.', 'The soldiers were ill.'])

for triple in triples:
    print(triple)

# Output: Landlords exercise oppression, soldiers were ill

Extract grammatical triples using different options from default:

from posextract.util import TripleExtractorOptions

triples = grammatical_triples.extract(sent, TripleExtractorOptions(prep_phrase = True))

Or extract adjectives and the nouns they modify.

from posextract import adj_noun_pairs

adj_noun = adj_noun_pairs.extract()

Or extract subjects and their verbs.

from posextract import subj_verb_pairs

subj_verb = subj_verb_pairs.extract()

Over CLI:

posextract can extract grammatical triples from text:

python -m posextract.extract_triples "Landlords may exercise oppression." output.csv

# Output: Landlords exercise oppression

posextract can extract SVO/SVA relationships separately or it can combine the adjective as part of a SVO triple:

python -m posextract.extract_triples "The soldiers were terminally ill." output.csv --post-combine-adj

# Output: soldiers were terminally, soldiers were ill 

python -m posextract.extract_triples "The soldiers were terminally ill." output.csv --post-combine-adj

# Output: soldiers were terminally ill

If provided a .csv file:

python -m posextract.extract_triples --data_column sentence --id_column sentence_id input.csv output.csv

For More Information...

... see our Wiki:

• About Our Evaluation Data
• About the Syntactic Dependency Parser
• How to Use posextract on Databricks