Generate text in the style of the text in the given files.

1. Install Git on your computer (I have no idea how to do this. wahhh.)
2. Download the repository
   1. open the command line
   2. Use the command cd path\to\directory to navigate to the directory where you want to work.
   3. Use the command git clone [email protected]:dhgarrette/saul-bot.git to download the repository. If you open the folder you should see that it has downloaded inside a new folder called saul-bot.

In the upper right of the github home page you can choose download. You'll get a zipped folder that you have to unzip. It will be called saul-bot-master, so either rename to saul-bot or plann accordingly.

The repository will come with a training file (alice.txt) that will create an aliceinwonderlandbot. But if you want it based on someone else's work, you'll need to give it something else to train on. It should be a .txt file full of words by the person you want to roboticize. Save the file in saul-bot.

There are examples below, but in general you'll need to do the following things:

1. Choose a file containing training text (i.e. alice.txt)
2. Decide if you want to use a word-based model (i.e. it only generates words - chars=false) or a character-based model (i.e. it generates character sequences that may or may not be words. chars=true)
3. Decide on the order, i.e. how much linguistic context you want the bot to use. The standard is an order of 3 (order=3), meaning it's a 3-gram or trigram model that chooses each word based on the two words that precede it.
4. Decide on length for your story (i.e. 5 lines. lines=5)
5. Choose a filename for your story (i.e. storyaboutsaul.txt).
6. Open the command line and navigate to the saul-bot folder using the cd command.
7. Enter a command to run the program. There are examples below, but the basic format is: programmingLanguage prograName.py trainingFile.txt --chars=false --order=number --lines=number > outputfilename.txt
python2 ngrams.py alice.txt --chars=false --order=3 --lines=5 > storyaboutsaul.txt

Train a 4-gram word-based language model on a file called alice.txt and generate 5 lines of text.

python2 ngrams.py alice.txt --chars=false --order=4 --lines=5

This will produce a sequence of words that are found in the original text. Each word generated will be based on (at most) the previous three words (--order=4).

Train a 7-gram character-based language model on a file called alice.txt and generate 10 lines of text.

python2 ngrams.py alice.txt --chars=true --order=7 --lines=10

This will produce a sequence of characters, not words, which means that the output text may contain words that are not actual words.

usage: ngrams.py [-h] [--lines LINES] [--order ORDER] [--chars CHARS]
                 [--max_length MAX_LENGTH] [--stop_symbols STOP_SYMBOLS]
                 [--backoff_exponent BACKOFF_EXPONENT]
                 FILE [FILE ...]

positional arguments:
  FILE                  The files to use for language model training.

optional arguments:
  -h, --help            show this help message and exit
  --lines LINES         The number of lines of text to generate. Default: 10
  --order ORDER         The order of the n-gram model (the maximum 'n').
                        Default: 4
  --chars CHARS         Use a character-based model instead of a word-based
                        model. Default: false
  --max_length MAX_LENGTH
                        The maximum number of symbols generated for a line.
  --stop_symbols STOP_SYMBOLS
                        Symbols that will stop the generation of a single
                        text. This is optional, but an example might be
                        --stop_symbols=".?\!"
  --backoff_exponent BACKOFF_EXPONENT
                        A higher exponent means that the model will be less
                        likely to randomly use shorter contexts. Default: 3