|
import json |
|
import argparse |
|
import sys |
|
import numpy as np |
|
import jieba |
|
from nltk.translate.bleu_score import sentence_bleu, SmoothingFunction |
|
from nltk import ngrams |
|
|
|
|
|
|
|
def load_file(filename): |
|
data = [] |
|
with open(filename, "r") as f: |
|
for line in f.readlines(): |
|
data.append(json.loads(line)) |
|
f.close() |
|
return data |
|
|
|
def proline(line): |
|
return " ".join([w for w in jieba.cut("".join(line.strip().split()))]) |
|
|
|
|
|
def compute(golden_file, pred_file, return_dict=True): |
|
golden_data = load_file(golden_file) |
|
pred_data = load_file(pred_file) |
|
|
|
if len(golden_data) != len(pred_data): |
|
raise RuntimeError("Wrong Predictions") |
|
|
|
num = 0 |
|
for g, p in zip(golden_data, pred_data): |
|
if isinstance(g["label"], str): |
|
l = int(g["label"].strip()) |
|
elif isinstance(g["label"], int): |
|
l = g["label"] |
|
else: |
|
raise Exception("Data type error") |
|
|
|
if isinstance(p["label"], str): |
|
p = int(p["label"].strip()) |
|
elif isinstance(p["label"], int): |
|
p = p["label"] |
|
else: |
|
raise Exception("Data type error") |
|
if l == p: |
|
num += 1 |
|
|
|
return {'accuracy': float(num)/len(golden_data)} |
|
|
|
def main(): |
|
argv = sys.argv |
|
print("预测结果:{}, 测试集: {}".format(argv[1], argv[2])) |
|
print(compute(argv[2], argv[1])) |
|
|
|
|
|
if __name__ == '__main__': |
|
main() |
|
|