python - 训练自定义 NER 模型

Question

我一直在一些文本上训练我的 NER 模型，并尝试使用自定义实体查找其中的城市。

示例:-

    ('paragraph Designated Offices Party A New York Party B Delaware paragraph pricing source calculation Market Value shall generally accepted pricing source reasonably agreed parties paragraph Spot rate Spot Rate specified paragraph reasonably agreed parties',
  {'entities': [(37, 41, 'DesignatedBankLoc'),(54, 62, 'CounterpartyBankLoc')]})

我正在这里寻找 2 个实体 DesignatedBankLoc 和 CounterpartyBankLoc。单个文本也可以有多个实体。

目前我正在对 60 行数据进行训练，如下所示:

import spacy
import random
def train_spacy(data,iterations):
    TRAIN_DATA = data
    nlp = spacy.blank('en')  # create blank Language class
    # create the built-in pipeline components and add them to the pipeline
    # nlp.create_pipe works for built-ins that are registered with spaCy
    if 'ner' not in nlp.pipe_names:
        ner = nlp.create_pipe('ner')
        nlp.add_pipe(ner, last=True)


    # add labels
    for _, annotations in TRAIN_DATA:
         for ent in annotations.get('entities'):
            # print (ent[2])
            ner.add_label(ent[2])

    # get names of other pipes to disable them during training
    other_pipes = [pipe for pipe in nlp.pipe_names if pipe != 'ner']
    with nlp.disable_pipes(*other_pipes):  # only train NER
        optimizer = nlp.begin_training()
        for itn in range(iterations):
            print("Statring iteration " + str(itn))
            random.shuffle(TRAIN_DATA)
            losses = {}
            for text, annotations in TRAIN_DATA:
                nlp.update(
                    [text],  # batch of texts
                    [annotations],  # batch of annotations
                    drop=0.5,  # dropout - make it harder to memorise data
                    sgd=optimizer,  # callable to update weights
                    losses=losses)
            print(losses)
    return nlp


prdnlp = train_spacy(TRAIN_DATA, 100)

我的问题是:-

当输入不同/相同的文本模式包含经过训练的城市时，模型预测正确。即使文本模式相同/不同，但模型不会预测任何实体，但不会预测训练数据集中从未出现过的不同城市。

请告诉我为什么会发生这种情况，请让我理解它是如何获得训练的概念？

Answer 1

根据经验，您有 60 行数据并训练 100 次迭代。您过度拟合了实体的值(value)而不是它们的位置。

要检查这一点，请尝试在句子中的随机位置注入(inject)城市名称，看看会发生什么。如果算法对它们进行了标记，则您可能会过度拟合。

有两种解决方案:

• 为这些实体创建更多具有更多不同值的训练数据
• 测试不同次数的迭代