qurator-spk/sbb_ner
1
0
Fork 0
mirror of https://github.com/qurator-spk/sbb_ner.git synced 2025-10-30 00:54:13 +01:00
Code Issues Releases Wiki Activity

correctly interpret output of NER model

Browse source
This commit is contained in:
Kai Labusch 2020-05-26 16:55:10 +02:00
parent fab5d460f2
commit fa72739312
1 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

7
qurator/sbb_ner/webapp/app.py
View file

@ -220,6 +220,9 @@ def ner(model_id=None):
for pos, (token, word_pred) in enumerate(zip(tokens, word_predictions)): for pos, (token, word_pred) in enumerate(zip(tokens, word_predictions)):
if word_pred == '[SEP]':
word_pred = 'O'
if not token.startswith('##') and token != '[UNK]': if not token.startswith('##') and token != '[UNK]':
if len(word) > 0: if len(word) > 0:
output_sentence.append({'word': word, 'prediction': last_prediction}) output_sentence.append({'word': word, 'prediction': last_prediction})
@ -244,9 +247,6 @@ def ner(model_id=None):
if not token.startswith('##') and word_pred == 'X': if not token.startswith('##') and word_pred == 'X':
word_pred = 'O' word_pred = 'O'
if word_pred == '[SEP]':
word_pred = 'O'
token = token[2:] if token.startswith('##') else token token = token[2:] if token.startswith('##') else token
word += token word += token
@ -267,6 +267,7 @@ def ner(model_id=None):
logger.warning('Input and output different!!! \n\n\nInput: {}\n\nOutput: {}\n'. logger.warning('Input and output different!!! \n\n\nInput: {}\n\nOutput: {}\n'.
format("".join(input_sentence).replace(" ", ""), format("".join(input_sentence).replace(" ", ""),
"".join([pred['word'] for pred in output_sentence]))) "".join([pred['word'] for pred in output_sentence])))
torch.cuda.empty_cache()
return jsonify(output) return jsonify(output)