• Search by Module
• Search by Words
• Search Projects

• Java
• Python
• JavaScript
• TypeScript
• C++
• Scala
• Blog

Project: ann-writer (GitHub Link)

• ann-writer-master
  • __pycache__
    • main.cpython-34.pyc
  • Datasets
    • MacbookAirBlog(large).txt
    • HarryPotter(large).txt
    • HarryPotter(xxlarge).txt
    • HarryPotter(xlarge).txt
    • HarryPotter(small).txt
    • HarryPotter(medium).txt
  • Modules
    • UnitTesting.py
    • __pycache__
      • NeuralNetwork.cpython-34.pyc
      • NetworkTrainer.cpython-34.pyc
      • UnitTesting.cpython-34.pyc
      • MachineLearning.cpython-34.pyc
      • ConsoleOutput.cpython-34.pyc
      • NaturalLanguage.cpython-34.pyc
    • .DS_Store
    • NaturalLanguage.py
    • ConsoleOutput.py
    • NetworkTrainer.py
    • MachineLearning.py
  • .DS_Store
  • LICENSE.md
  • main.py
  • README.md

Machine learning text generator

Machine learning that utilises sk-learn, numpy and nltk in an attempt to generate text in the style of any given training data. Written in python3.

Installing

---

pip3 install -U scikit-learn
pip3 install -U numpy
pip3 install -U nltk

Usage

---

Run the program:

python3 main.py

Run the unit test function on the sentence structuring:

python3 main.py -utss

Unit test function for vocabulary:

python3 main.py -utv

Specify the training data file:

python3 main.py -td <filepath>

Specify test sentence: (Generates text that follows on from the input) example input = "the boy ran"

python3 main.py -ts "<input sentence here>"

Specify the number of words generated for given test sentence:

python3 main.py -tsc <genCount>

Output generated text to a file:

python3 main.py -of "<fileLocation>"

Example usage scenario:

python3 main.py -ts "today i will" -tsc 10 -td "Datasets/HarryPotter(xxlarge).txt"

Datasets

---

Includes 6 datasets:

HarryPotter(small).txt = 346 training vectors
HarryPotter(medium).txt = 2500 training vectors
HarryPotter(large).txt = 4550 training vectors
HarryPotter(xlarge).txt = 11429 training vectors
HarryPotter(xxlarge).txt = 15829 training vectors

MacbookAirBlog(large).txt = 3576 training vectors

Change the data sets with the '-td' command. The larger the data set, the longer the program will take to fit and produce a result. The ability to load an already fitted network has not been implemented yet, so the program has to run the initial fit every time.
The Harry potter data sets have been taken from the book directly and the macbook dataset was taken from a random blog.
It is extremely easy to add your own data set, just make sure that it is in the form of a text blob (see provided datasets). And then simply use the command line to select your dataset

python3 main.py -td "Datasets/your_set.txt"

Dataset has to contain more words than the training range (default = 3).

Go here to see results!

Here I show multiple text generations with different training data sets and how accurate the program is at impersonating the training data.