Last active
January 27, 2021 07:13
-
-
Save liangfu/954c032adf25273bb15f5706ed2f3013 to your computer and use it in GitHub Desktop.
Convert origindata.txt in bosonnlp to CoNLL format for named entity recognition (NER).
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import random | |
random.seed(111) | |
def bosonnlp_to_bio2(origfile, trainfile, valfile): | |
val_ratio = 0.2 | |
traindata = [] | |
valdata = [] | |
with open(origfile, 'rt') as fp: | |
lines = fp.readlines() | |
random.shuffle(lines) | |
val_samples = int(len(lines) * val_ratio) | |
val_lines = lines[:val_samples] | |
train_lines = lines[val_samples:] | |
def transform(line): | |
# print(line) | |
it = 0 | |
document = "" | |
annotations = [] | |
while True: | |
start = line.find("{{", it) | |
end = line.find("}}", start) | |
next_start = line.find("{{", end) | |
if end < 0: | |
break | |
# print(start, end) | |
labeltext = line[start+2:end] | |
loc = labeltext.find(":") | |
label = labeltext[:loc] | |
text = labeltext[loc+1:] | |
# label, text = line[start+2:end].split(":") | |
prefix = line[it:start] | |
if next_start > 0: | |
suffix = line[end+2:next_start] | |
else: | |
suffix = line[end+2:] | |
tic = len(prefix) + len(document) | |
toc = len(prefix) + len(document) + len(text) | |
annotations.append([tic, toc, label]) | |
document += prefix + text + suffix | |
it = next_start | |
document = document.replace(' ', '_') | |
document = document.replace(',', ',') | |
document = document.replace('“', '"') | |
document = document.replace('”', '"') | |
document = document.replace(':', ':') | |
document = document.replace('(', '(') | |
document = document.replace(')', ')') | |
document = document.replace('\t', '_') | |
return annotations, document | |
documents = [] | |
def strip_suffix(tag): | |
if tag.lower().endswith("name"): | |
tag = tag[:-4] | |
return tag.upper().strip("-_") | |
with open(trainfile, 'w') as fp: | |
for idx, line in enumerate(train_lines): | |
annotations, document = transform(line) | |
# print(annotations) | |
# print(document) | |
document = document.strip() | |
documents.append(document) | |
count = 0 | |
for i, c in enumerate(document): | |
label = "O" | |
for a in annotations: | |
if i >= a[0] and i < a[1]: | |
label = "I-"+strip_suffix(a[2]) | |
fp.write("{} X X {}\n".format(c, label)) | |
# limit sequence length to 128 - 2 | |
if (count % 125) == 124 or c in ["。", ";"]: | |
fp.write("\n") | |
count = 0 | |
else: | |
count += 1 | |
fp.write("\n") | |
with open(valfile, 'w') as fp: | |
for idx, line in enumerate(val_lines): | |
annotations, document = transform(line) | |
# print(annotations) | |
print(document) | |
document = document.strip() | |
documents.append(document) | |
count = 0 | |
for i, c in enumerate(document): | |
label = "O" | |
for a in annotations: | |
if i >= a[0] and i < a[1]: | |
label = "I-"+strip_suffix(a[2]) | |
fp.write("{} X X {}\n".format(c, label)) | |
# limit sequence length to 128 - 2 | |
if (count % 125) == 124 or c in ["。", ";"]: | |
fp.write("\n") | |
count = 0 | |
else: | |
count += 1 | |
fp.write("\n") | |
with open("documents.txt", "w") as fp: | |
fp.write("\n".join(documents)) | |
bosonnlp_to_bio2('origindata.txt', 'train.txt', 'val.txt') |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment