# NLP-Lab

**Repository Path**: wangsc18/nlp-lab

## Basic Information

- **Project Name**: NLP-Lab
- **Description**: USTC NLP course project repo
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: master
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2023-04-06
- **Last Updated**: 2023-04-17

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

参考：[NLP入门（十）使用LSTM进行文本情感分析](https://www.cnblogs.com/jclian91/p/10886031.html)

运行环境：```windows10, python3.8```

数据集分布特征：

![句子长度及出现频数统计图](句子长度及出现频数统计图.png)

![句子长度累积分布函数图](句子长度累积分布函数图.png)

模型结构：

![模型结构](model_lstm.png)

```train.py```运行结果：

```commandline
Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 embedding (Embedding)       (None, 180, 20)           82100     
                                                                 
 lstm (LSTM)                 (None, 100)               48400     
                                                                 
 dropout (Dropout)           (None, 100)               0         
                                                                 
 dense (Dense)               (None, 2)                 202       
                                                                 
=================================================================
Total params: 130,702
Trainable params: 130,702
Non-trainable params: 0
_________________________________________________________________
Epoch 1/6
282/282 [==============================] - 14s 45ms/step - loss: 0.5937 - accuracy: 0.6848
Epoch 2/6
282/282 [==============================] - 12s 44ms/step - loss: 0.4813 - accuracy: 0.7766
Epoch 3/6
282/282 [==============================] - 12s 44ms/step - loss: 0.4468 - accuracy: 0.8043
Epoch 4/6
282/282 [==============================] - 13s 45ms/step - loss: 0.4229 - accuracy: 0.8180
Epoch 5/6
282/282 [==============================] - 12s 44ms/step - loss: 0.3943 - accuracy: 0.8336
Epoch 6/6
282/282 [==============================] - 12s 44ms/step - loss: 0.3724 - accuracy: 0.8366

...

模型在测试集上的准确率为: 0.71.
```