def do_predict(args):
paddle.set_device(args.device)
wordtag = Taskflow("text2knowledge",
model="wordtag",
batch_size=args.batch_size,
max_seq_length=args.max_seq_len)
txts = ["《孤女》是2010年九州出版社出版的小说，作者是余兼羽。", "热梅茶是一道以梅子为主要原料制作的茶饮"]
res = wordtag(txts)
print(res)
plt.figure(figsize=(10, 5))
plt.bar(range(len(cnt)), cnt.values(),
tick_label=list(cnt.keys())) # , orientation="horizontal"
# for a, b in zip(x_list, y_list):
# plt.text(a, b + 0.05, '%.0f' % b, ha='center', va='bottom', fontsize=10)
plt.xticks(rotation=45)
plt.savefig('./img.png')
plt.show()
if __name__ == '__main__':
# 模型路径下必须含有model和params文件
ocr = PaddleOCR(
det_model_dir='./PaddleOCR/output/ch_db_mv3_inference/inference',
use_angle_cls=True)
lac = Taskflow("pos_tagging")
enti_list = []
pdfFolder = './ResearchReport'
for p in os.listdir(pdfFolder):
if os.path.isdir(os.path.join(pdfFolder, p)):
print('Processing folder:', p)
imgPath = pdfFolder + '/' + p
res_list = []
for i in os.listdir(imgPath):
img_path = os.path.join(imgPath, i)
result = ocr.ocr(img_path, cls=True)
res_list.append(result)
enti = LAC(lac, res_list)
enti_list += enti
def __init__(self, model="csc-ernie-1.0"):
self.text_correction = Taskflow("text_correction", model=model)
import paddlenlp
from paddlenlp import Taskflow
# from paddlenlp.transformers import SkepTokenizer,SkepForTokenClassification
#
# modle=SkepForTokenClassification.from_pretrained(pretrained_model_name_or_path="skep_ernie_1.0_large_ch")
#
#
# tokenizer = paddlenlp.transformers.SkepTokenizer.from_pretrained(
# "skep_ernie_1.0_large_ch")
senta = Taskflow("sentiment_analysis",model="skep_ernie_1.0_large_ch")
data=senta("每一次的失败，都要总结原因，才能赚钱")
print(data)
if os.path.isfile(args.audio_file):
audios = [args.audio_file]
elif os.path.isdir(args.audio_file):
audios = [x for x in os.listdir(args.audio_file)]
audios = [os.path.join(args.audio_file, x) for x in audios]
else:
raise Exception('%s is neither valid path nor file!' % args.audio_file)
audios = [x for x in audios if x.endswith('.wav')]
if len(audios) == 0:
raise Exception('No valid .wav file! Please check %s.' %
args.audio_file)
if args.uie_model is None:
parser = Taskflow('information_extraction', schema=args.schema)
else:
parser = Taskflow('information_extraction',
schema=args.schema,
task_path=args.uie_model)
with open(args.save_file, 'w') as fp:
for audio_file in tqdm(audios):
# automatic speech recognition
text = mandarin_asr_api(args.api_key, args.secret_key, audio_file)
# extract entities according to schema
result = parser(text)
fp.write(text + '\n')
fp.write(json.dumps(result, ensure_ascii=False) + '\n\n')
print(text)
pprint.pprint(result)