https://github.com/philipperemy/sentiment-analysis-nlp

Sentiment Analysis applied to different datasets such as IMDB
https://github.com/philipperemy/sentiment-analysis-nlp

Last synced: 21 days ago
JSON representation

Sentiment Analysis applied to different datasets such as IMDB

• Host: GitHub
• URL: https://github.com/philipperemy/sentiment-analysis-nlp
• Owner: philipperemy
• Created: 2016-02-18T08:07:56.000Z (over 8 years ago)
• Default Branch: master
• Last Pushed: 2016-02-25T03:21:32.000Z (over 8 years ago)
• Last Synced: 2024-10-04T13:31:37.848Z (about 1 month ago)
• Language: Python
• Size: 14.6 KB
• Stars: 19
• Watchers: 3
• Forks: 10
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Simple sentiment analysis examples 

Sentiment analysis applied to different datasets such as IMDB. The perceptron implementation uses [Keras](http://keras.io/), a minimalist, highly modular neural networks library, written in Python and capable of running on top of either TensorFlow or Theano.

## Documentation

What is [Sentiment Analysis (Wikipedia)](https://en.wikipedia.org/wiki/Sentiment_analysis) ?  


Learn in 30 seconds with [Keras](http://keras.io/#getting-started-30-seconds-to-keras).

## Getting Started

```

git clone https://github.com/philipperemy/Sentiment-Analysis-NLP.git

cd Sentiment-Analysis-NLP

chmod +x init.sh

./init.sh

python main_perceptron.py

```

The script will download the [IMDB Sentiment Database](http://ai.stanford.edu/~amaas/data/sentiment/) from Stanford University and unzip it.

Also, we consider the list of positive and negative words from Illinois University, available [here](https://www.cs.uic.edu/~liub/FBS/sentiment-analysis.html#lexicon). This list has around 6800 words that we refer as keywords.

## Execution

You should have a console output similar to the one below.

```

Using TensorFlow backend.

=> loaded 2006 positive words

=> loaded 4783 negative words

=> processed 500 aclImdb/train/neg files.

[...]

=> processed 5000 aclImdb/train/neg files.

=> processed 500 aclImdb/train/pos files.

[...]

=> processed 5000 aclImdb/train/pos files.

I tensorflow/core/common_runtime/local_device.cc:40] Local device intra op parallelism threads: 4

I tensorflow/core/common_runtime/direct_session.cc:58] Direct session inter op parallelism threads: 4

Epoch 1/200

6666/6666 [==============================] - 22s - loss: 0.2177    

[...]  

```

## Results (accuracy)

On 10k reviews with 2/3 training and 1/3 validation set (200 epochs). 


This result is very experimental and can be drastically increased.

```

training= 0.803180318032

validation= 0.732913669065

```

## Visualize the graph (tensorflow feature)

Make sure the backend of Keras is tensorflow. You can check it easily by

```

cat ~/.keras/keras.json

{"epsilon": 1e-07, "floatx": "float32", "backend": "tensorflow"}

```

Finally run the tensorboard command with specifying the log dir.

```

tensorboard --logdir=/Users/philipperemy/PycharmProjects/Sentiment-Analysis-NLP/deep_learning/runs/1455792487

```