Python系列：NLP系列二：命名实体识别（NER）、用深度学习实现命名实体识别（NER）

import torch
import torch.nn as nn
from torch.nn import functional as F
 
class NERModel(nn.Module):
    def __init__(self, vocab, embeddings, tag_to_idx):
        super(NERModel, self).__init__()
        self.embedding = nn.Embedding(len(vocab), 64)
        self.embedding.weight = nn.Parameter(torch.tensor(embeddings, dtype=torch.float))
        self.embedding.weight.requires_grad = False
        self.lstm = nn.LSTM(64, 64, bidirectional=True)
        self.hidden2tag = nn.Linear(128, len(tag_to_idx))
 
    def forward(self, input_seq):
        embeds = self.embedding(input_seq)
        lstm_out, _ = self.lstm(embeds.view(len(input_seq), 1, -1))
        tag_space = self.hidden2tag(lstm_out.view(1, -1))
        tag_scores = F.log_softmax(tag_space, dim=1)
        return tag_scores
 
# 示例用法
vocab = {'hello': 1, 'world': 2}
embeddings = [[1, 2], [3, 4]]  # 假设的嵌入矩阵，实际应该从文件中加载
tag_to_idx = {'B-PER': 1, 'I-PER': 2, 'B-ORG': 3, 'I-ORG': 4}
model = NERModel(vocab, embeddings, tag_to_idx)
input_seq = torch.tensor([1, 2])  # 假设的输入序列
output = model(input_seq)
print(output)

这段代码定义了一个基于PyTorch的简单命名实体识别模型，它使用了嵌入层、双向长短期记忆单元(LSTM)和全连接层。示例用法展示了如何实例化模型并对输入序列进行处理。