Sentiment-Analysis-MK1

A sentiment analysis project based on Sentiment140 training data. This First attempt utilises some basic rules for creating quality input data from the raw data which is fed into a LSTM neural network.

Demo Available Here: https://www.ericsciberras.com/portfolio/#LSTM-Sentiment-Analysis

Getting Started

Play With Already Trained Neural Network

1. Go to html folder and and open nn.html

Train Neural Network Yourself

1. Download the sentiment140 data from here. Alternatively you may use your own dataset however the process_raw_data() function may need to be altered so it is in the correct format.

2. Run
   pip3 install -r requirements.txt
   To install all the python3 libraries required for use.

3. Run
   python3 processData /path/to/dataset/'
   To create 3 files:

   • processed_data.csv : which is the data in correct format for converting sentences to arrays
   • processed_data_2.csv : which is data that is ready to be fed into the neural network
   • tokeniser.pickle : holds the dictionary which is useful for converting the words to numbers and vice versa

4. Run
   python3 neuralnet
   To start training. Note: it is recommended to use tensorflow-gpu for this as running on a cpu will be very slow. This will create a .hdf5 file for each epoch e.g. sentiment-ai-04-0.74. The format is sentiment-ai--<val_acc> and a file called my-model.hdf5 which is the last epoch.

To play with the model:

Python

1. Run
   python3 runmodel /path/to/model.hdf5

OR

Web Browser

1. Install tfjs-converter
   pip install tensorflowjs
2. Convert the keras model to TF.js Layers format
   tensorflowjs_converter --input_format keras path/to/my_model.h5 path/to/tfjs_target_dir
3. Run
   dictionary2json ./tokeniser.pickle ./dict.json
   To turn the dictionary pickle to json.
4. Replace all new created files dict.json,group1-shard1of1 and model.json with the current ones in the html folder.
5. Open nn.html.

Possible Improvements For MK2

• Use more or better datasets (create my own ???)
• Research Word2vec and its possible advantages
• Allow neural network to accept arbitrarily sized sentences
• Use more sophisticated techniques for processing raw datasets (i.e: identifying words that don't contribute to sentiment)
• Optimise parameters for LSTM neural network or use different neural network