B.Tech-Project: Sentimental Data Analysis

All the updates and current work-in-progress for my B.Tech Project, along with Namish Narayan under the guidance of Kshitiz Verma Sir.

Current Work:

• Small Data Model for dataset of 10,662 lines of Data
• Large Data Model for dataset of 1.6 Million lines of Data

Small Data Model Overview:

• Creating Organized and Vectorized Data

  • Sentiment data is extracted from 2 txt file, pos.txt and neg.txt that contains respective type of sentences.
  • The program reads the files line by line, and lemmatizes the words for each line.
  • Repeating the process for every line from data, gives lexicon of the data.
  • Lexicon contents are shuffled (in order to maintain randomness of +ve and -ve data)
  • Training Data, Testing Data, Training Labels and Testing Labels are created, by pairing the vectorized Lexicon's count of occurence for each line, alone with a one-hot-array to denote positive or negative sentiment of that specific line.
  • This data is then stored as pickle, and can be accessed externally.

• Data Feed into Neural Network

- This calculates the total accuracy of the model.

How To Build:

• Install pip3 (for Python 3)
• Install Essential libraries like TensorFlow, NumPy, SciPy

     pip3 install tensorflow

• Install NLTK Library for Python

    pip3 install nltk
    #Open Python in Commandline
    python3
    #Download all the required data of NLTK
    nltk.download('all')

• Clone this repository and run

     python3 neural_net.py

Please feel free to leave all suggestions and fixes in issues. Thank You.