modify tasks

ChineseNER/BiLSTM_crf.py

@@ -0,0 +1,47 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torch.nn import CrossEntropyLoss
+from torch.nn.utils.rnn import pad_packed_sequence, pack_padded_sequence
+
+
+class BiLSTM(nn.Module):
+    def __init__(self, config):
+        super(BiLSTM, self).__init__()
+        self.config = config
+        self.device = config.device
+
+        self.emb_dim = config.emb_dim
+        self.hidden_dim = config.hidden_dim
+        self.vocab_size = config.vocab_size
+        self.tag2id = config.tag2id
+        self.tagset_size = len(self.tag2id)
+
+        # map input tokens to unique embedding(vector)
+        self.emb = nn.Embedding(self.vocab_size, self.emb_dim)
+        # LSTM is a variant of Recurrent Neural Network(RNN)
+        self.lstm = nn.LSTM(self.emb_dim, self.hidden_dim, num_layers=1, batch_first=True, bidirectional=True)
+        # linear layer, predict the probability of each tag
+        self.hidden2tag = nn.Linear(2 * self.hidden_dim, self.tagset_size)
+
+        # Loss: compute the distance between our prediction and the gold tag
+        self.loss = CrossEntropyLoss()
+
+    def forward(self, sent, labels, lengths, mask):
+        embedded = self.emb(sent)
+
+        # The padded batch should be packed before LSTM
+        embedded = pack_padded_sequence(embedded, lengths, batch_first=True)
+        lstm_out, _ = self.lstm(embedded)
+        # The packed batch should be padded after LSTM
+        lstm_out, _ = pad_packed_sequence(lstm_out, batch_first=True)   # lstm_out: [batch_size, max_len, hidden_dim]
+        logits = self.hidden2tag(lstm_out)  # logits: [batch_size, max_len, tagset_size]
+
+        # Predict the tags
+        pred_tag = torch.argmax(logits, dim=-1)
+
+        # Compute loss. Pad token must be masked before computing the loss.
+        logits = logits.view(-1, self.tagset_size)[mask.view(-1) == 1.0]
+        loss = self.loss(logits, labels.view(-1))
+
+        return loss, pred_tag

ChineseNER/README.md

@@ -0,0 +1,70 @@
+# BiLSTM Baseline
+
+
+
+简单的 BiLSTM NER baseline，作为我们 Project 的一个起点
+
+代码已经做好注释，如有不明白之处请及时提问，如发现 bug 请及时报告hhh
+
+
+
+## 项目结构
+
+```shell
+├── data						# The directory of datasets
+│   └── renMinRiBao
+│       ├── tags.txt
+│       ├── test_data.txt
+│       ├── train_data.txt
+│       └── val_data.txt
+├── model						# The directory to save models
+├── BiLSTM.py					# BiLSTM model
+├── preprocess_data.py			# Preprocess the datasets
+├── train.py					# The entry of this project
+└── utils.py					# Some useful functions
+
+```
+
+
+
+## 运行环境
+
+```shell
+Python 3.7
+PyTorch 1.7.1
+NumPy 1.19.2
+```
+
+请使用 pip install 命令安装以上运行环境（版本不一定非要一样，正常安装最新版一般不会有问题），建议使用 anaconda 或 miniconda 等虚拟环境（自行百度），做好环境管理
+
+执行 `python train.py` 即可按照默认设置训练模型
+
+执行 `python train.py --test` 即可在测试集上进行测试
+
+
+
+## 运行结果
+
+训练10个 epoch，准确率：88.58，召回率：83.45，F1：85.94（实际测试结果略有出入为正常现象）
+
+
+
+## 学习资料
+
+机器学习基本概念：周志华《机器学习》前几章
+
+Word embedding：https://www.zhihu.com/question/32275069
+
+RNN：https://zhuanlan.zhihu.com/p/123211148，https://zhuanlan.zhihu.com/p/28054589，https://zhuanlan.zhihu.com/p/30844905，https://www.bilibili.com/video/BV1JE411g7XF?p=20
+
+LSTM：http://colah.github.io/posts/2015-08-Understanding-LSTMs/
+
+交叉熵：https://zhuanlan.zhihu.com/p/149186719，https://pytorch.org/docs/stable/generated/torch.nn.CrossEntropyLoss.html
+
+梯度下降：https://www.bilibili.com/video/BV1JE411g7XF?p=5
+
+PyTorch document：https://pytorch.org/docs/stable/index.html
+
+PyTorch 教程：https://pytorch.org/tutorials/index.html
+
+课程：李宏毅深度学习，https://www.bilibili.com/video/BV1JE411g7XF

ChineseNER/data/renMinRiBao/tags.txt

@@ -0,0 +1,9 @@
+O
+B-LOC
+B-PER
+I-ORG
+B-ORG
+I-DATE
+I-PER
+I-LOC
+B-DATE