def predict_sentiment_file():
model_path = "/home/kuan.li/msnlp/output_models/sentiment"
pretrain_type = "bert"
task_name = "text_classification"
data_dir = "/home/kuan.li/msnlp/data/sentiment"
test_file = os.path.join(data_dir, "test.tsv")
output_file = os.path.join(data_dir, "test_predict.tsv")
test_set = pd.read_table(test_file, sep='\t')
texts = {"text": test_set.text.tolist()}
# texts = {
# "text": [
# "已经被你们坑了两次了",
# "草泥马逼的30次都没提现成功",
# "还光了吧我",
# ]
# }
predictor = NLPPredictor(task_name, model_path=model_path, pretrain_type=pretrain_type)
pres = predictor.predict_rest(texts)
test_set["predict_label"] = pres["data"]
test_set.to_csv(path_or_buf=output_file, sep='\t', header=False, index=False)
print("file saved.")
num_correct = 0
for v in test_set.values:
if v[1] == v[2]:
num_correct = num_correct + 1
print("Accuracy:{}".format(float(num_correct)/len(test_set)))
print("total lines:{}".format(len(test_set)))
def predict_xma():
model_path = "/home/dingnan.jin/projects/nlpsdk/testout/xma"
pretrain_type = "bert"
task_name = "text_classification"
texts = {
"text": [
"几点到几点上班",
"贷20000利息多少",
"我不用了，怎么解绑",
"立即还款里显示我没借过钱 我怎么还",]
}
predictor = NLPPredictor(task_name, model_path=model_path, pretrain_type=pretrain_type)
pres = predictor.predict_rest(texts)
print(pres)
def predict_tnews():
model_path = "/home/kuan.li/msnlp/output_models/tnews"
pretrain_type = "bert"
task_name = "text_classification"
texts = {
"text": [
"以色列暗杀叙利亚7名“国家英雄”的行为该如何定性？",
"小学生高校食堂写作业：不会的问大学生，他们也做错，只能靠自己",
"芜湖有哪些上市公司？",
"赵本山在东北的地位如何？最近怎么没有消息了呢？", ]
}
predictor = NLPPredictor(task_name, model_path=model_path, pretrain_type=pretrain_type)
pres = predictor.predict_rest(texts)
print(pres)
def predict_sst():
model_path = "/home/dingnan.jin/projects/nlpsdk/testout/SST-2"
pretrain_type = "bert"
task_name = "text_classification"
texts = {
"text": [
" it 's just incredibly dull",
"it 's a charming and often affecting journey . ",
"unflinchingly bleak and desperate",
"allows us to hope that nolan is poised to embark a major career as a commercial yet inventive filmmaker .",]
}
predictor = NLPPredictor(task_name, model_path=model_path, pretrain_type=pretrain_type)
pres = predictor.predict_rest(texts)
print(pres)