- [Script and Sample Code](#script-and-sample-code)
- [Script Parameters](#script-parameters)
- [Training Process](#training-process)
- [Evaluation Process](#evaluation-process)
- [Model Description](#model-description)
- [Performance](#performance)
- [ModelZoo Homepage](#modelzoo-homepage)
# [TextCNN Description](#contents)
TextCNN is an algorithm that uses convolutional neural networks to classify text. It was proposed by Yoon Kim in the article "Convolutional Neural Networks for Sentence Classification" in 2014. It is widely used in various tasks of text classification (such as sentiment analysis). It has become the standard benchmark for the new text classification framework. Each module of TextCNN can complete text classification tasks independently, and it is convenient for distributed configuration and parallel execution. TextCNN is very suitable for the semantic analysis of short texts such as Weibo/News/E-commerce reviews and video bullet screens.
[Paper](https://arxiv.org/abs/1408.5882): Kim Y. Convolutional neural networks for sentence classification[J]. arXiv preprint arXiv:1408.5882, 2014.
# [Model Architecture](#contents)
The basic network structure design of TextCNN can refer to the paper "Convolutional Neural Networks for Sentence Classification". The specific implementation takes reading a sentence "I like this movie very much!" as an example. First, the word segmentation algorithm is used to divide the words into 7 words, and then the words in each part are expanded into a five-dimensional vector through the embedding method. Then use different convolution kernels ([3,4,5]*5) to perform convolution operations on them to obtain feature maps. The default number of convolution kernels is 2. Then use the maxpool operation to pool all the feature maps, and finally merge the pooling result into a one-dimensional feature vector through the connection operation. At last, it can be divided into 2 categories with softmax, and the positive/negative emotions are obtained.
# [Dataset](#contents)
Note that you can run the scripts based on the dataset mentioned in original paper or widely used in relevant domain/network architecture. In the following sections, we will introduce how to run the scripts using the related dataset below.
- Please click [here](<http://www.cs.cornell.edu/people/pabo/movie-review-data/rt-polaritydata.tar.gz>) to download the data, and change the files into utf-8. Then put it into the `data` directory.
- Note：Data will be processed in src/dataset.py
# [Environment Requirements](#contents)
- Hardware（Ascend）
- Prepare hardware environment with Ascend processor. If you want to try Ascend , please send the [application form](https://obs-9be7.obs.cn-east-2.myhuaweicloud.com/file/other/Ascend%20Model%20Zoo%E4%BD%93%E9%AA%8C%E8%B5%84%E6%BA%90%E7%94%B3%E8%AF%B7%E8%A1%A8.docx) to ascend@huawei.com. Once approved, you can get the resources.
For more configuration details, please refer the script `config.py`.
## [Training Process](#contents)
- running on Ascend
```python
python train.py > train.log 2>&1 &
OR
sh scripts/run_train.sh
```
The python command above will run in the background, you can view the results through the file `train.log`.
After training, you'll get some checkpoint files in `ckpt`. The loss value will be achieved as follows:
```python
# grep "loss is " train.log
epoch: 1 step 149, loss is 0.6194226145744324
epoch: 2 step 149, loss is 0.38729554414749146
...
```
The model checkpoint will be saved in the `ckpt` directory.
## [Evaluation Process](#contents)
- evaluation on movie review dataset when running on Ascend
Before running the command below, please check the checkpoint path used for evaluation. Please set the checkpoint path to be the absolute full path, e.g., "username/textcnn/ckpt/train_textcnn.ckpt".
The above python command will run in the background. You can view the results through the file "eval.log". The accuracy of the test dataset will be as follows:
