Algorithm praise detector

用来检测教师鼓励行为的模型，模型存放在./model_zoo：

行为	模型文件名称	Precision	Recall	F1	Accuracy
鼓励	package_model_praise	0.86	0.91	0.82	0.84

Efficiency Performace

• test environment

  • Tesla P4, ubuntu 18.04, 4-cores CPU, 16G memory, GPU 7611MB(Not use), driver version= 418.67

  • Duration: 17 hours

• Performace

  • Average Memeory Usage : 5%

    

  • Average CPU Usage : 35%

    

  • speed

单位	value
字符/秒	26857

Quick Start

Train Model With Demo Dataset

> python train.py --dataset ./dataset.py --model_path /tmp

USAGE

• Basic version

# Top two lines are not necessary when do outer import
import sys
sys.path.append("../")

from algorithm_praise_detector.predictor import Predictor


model_path = "/workspace/tmp/package_model_praise"
model = Predictor(model_path)

# predict one piece of sentence
labels, probs, keywords = model.predict("你真的好棒啊!")
print(labels, probs)
print(keywords)

# predict list of sentences
input_list = ["你真的好棒！"]*3000
labels, probs, _ = model.predict(input_list)
assert len(input_list) == len(labels) == len(probs)

• Interface version

# Top two lines are not necessary when do outer import
import sys
sys.path.append("../")

from algorithm_praise_detector import model


model_path = "./model_zoo/package_model_praise"
model = model(model_path)
input_text = [
    {
        "text":"这句话呢，其我我靠实都是很棒告诉你很棒规则，他就看你能不能看到他这个给你的规定了。",
        "begin_time":1326750,
        "end_time":1332165
    }
]
print(model(input_text))
# """[
#     {
#     'keyword_list':
#                     [
#                         {
#                         'keyword': '很棒',
#                         'word_count': 2
#                         }
#                     ],
#     'sentence': '这句话呢，其我操我靠实都是很棒告诉你很棒规则，
#                     他就看你能不能看到他这个给你的规定了。',
#     'begin_time': 1326750,
#     'end_time': 1332165}
#     ]
# """