实验二

1.tagger-master中的loader.py实现从文件中加载词向量。

pretrained = set([#关注的是这里，打开词嵌入文件。
line.rstrip().split()[0].strip()
for line in codecs.open(ext_emb_path, 'r', 'utf-8')
if len(ext_emb_path) > 0
])

 2.Tagger-master如何处理evaluate中写入文件，显示精度、召回率、F1值等：

y_preds = f_eval(*input).argmax(axis=1)
predictions.append(new_line)
with codecs.open(output_path, 'w', 'utf8') as f:
        f.write("
".join(predictions))

写入的score文件内容：

processed 121289 tokens with 9809 phrases; found: 9505 phrases; correct: 8296.
accuracy:  96.92%; precision:  87.28%; recall:  84.58%; FB1:  85.91
         Chemical: precision:  91.75%; recall:  87.95%; FB1:  89.81  5162
          Disease: precision:  81.97%; recall:  80.47%; FB1:  81.21  4343

再从文件中读取并返回：

eval_lines = [l.rstrip() for l in codecs.open(scores_path, 'r', 'utf8')]
# F1 on all entities
return float(eval_lines[1].strip().split()[-1])

3.Att-NER如何处理evaluate中写入文件，显示精度、召回率、F1值等：

y_preds = f_eval(*input).argmax(axis=1)
predictions.append(new_line)
#write to file 
with codecs.open(filename, 'w', 'utf8') as f:
    f.write("
".join(predictions))
#读取
return get_perf(filename)