spaCy/spacy/cli/converters/jsonl2json.py

34 lines
1.2 KiB
Python

# coding: utf8
from __future__ import unicode_literals
import ujson as json
from .._messages import Messages
from ...compat import json_dumps, path2str
from ...util import prints, get_lang_class
from ...gold import docs_to_json
def ner_jsonl2json(input_path, output_path, lang=None, n_sents=10, use_morphology=False):
if lang is None:
prints(Messages.M054, exits=True)
json_docs = []
input_tuples = list(read_jsonl(input_path))
nlp = get_lang_class(lang)()
for i, (raw_text, ents) in enumerate(input_tuples):
doc = nlp.make_doc(raw_text)
doc[0].is_sent_start = True
doc.ents = [doc.char_span(s, e, label=L) for s, e, L in ents['entities']]
json_docs.append(docs_to_json(i, [doc]))
output_filename = input_path.parts[-1].replace(".jsonl", ".json")
output_loc = output_path / output_filename
with (output_loc).open('w', encoding='utf8') as file_:
file_.write(json_dumps(json_docs))
prints(Messages.M033.format(n_docs=len(json_docs)),
title=Messages.M032.format(name=path2str(output_loc)))
def read_jsonl(input_path):
with input_path.open('r', encoding='utf8') as file_:
for line in file_:
yield json.loads(line)