spacy 3 NER Scorer() throws TypeError: score() takes 2 positional arguments but 3 were given

Question

运行尝试使用 Scorer

在我的测试集上获取分数时出现以下错误

TypeError: score() takes 2 positional arguments but 3 were given

import spacy 
from spacy.tokens import Span
from spacy import displacy
from spacy.training import *
from spacy.scorer import Scorer
from spacy.util import minibatch, compounding

def evaluate(ner_model, testing_data):
    scorer = Scorer()

    for input_, annot in testing_data:
        doc_gold_text = ner_model.make_doc(input_)
        example = Example.from_dict(doc_gold_text, {"entities": annot})
        pred_value = ner_model(input_)
        
    return scorer.score(pred_value, example)

print(evaluate(nlp_updated, testing_tagged))

其中 testing_tagged:

testing_tagged = [
    ("Who was Hamlet?", [(8,14,'PERSON')]),
    ("Have you ever met Rome?", [(18,22,'LOC')])
]

预期输出是 p、r 和 f 不为 0 的地方：

{'uas': 0.0, 'las': 0.0, 'las_per_type': {'': {'p': 0.0, 'r': 0.0, 'f': 0.0}}, 'ents_p': 0.0, 'ents_r': 0.0, 'ents_f': 0.0, 'ents_per_type': {'PERSON': {'p': 0.0, 'r': 0.0, 'f': 0.0}, 'LOC': {'p': 0.0, 'r': 0.0, 'f': 0.0}}, 'tags_acc': 0.0, 'token_acc': 100.0, 'textcat_score': 0.0, 'textcats_per_cat': {}}

我最初使用 GoldParse 函数而不是 Example.from_dict 来处理这个问题 - 但我升级到 Spacy 3.0.5，我不明白为什么会出现这个错误。

Answer 1

自 spaCy v3 起，scorer.score 仅采用示例列表。每个 Example 对象包含两个 Doc 对象：

一个 reference 带有黄金标准注释的文档，创建于来自给定 annot 字典的示例
一个 predicted 文档与预测。然后记分员会比较两者。

所以你想要这样的东西：

from spacy.training import Example
examples = []
for ...:
    example = Example.from_dict(text, {"entities": annot})
    example.predicted = ner_model(example.predicted)
    examples.append(example)
scorer.score(examples)

spacy 3 NER Scorer() throws TypeError: score() takes 2 positional arguments but 3 were given

spacy 3 NER Scorer() throws TypeError: score() takes 2 positional arguments but 3 were given

python

spacy-3