A Random Paragraph Generator

This Python program randomly generates paragraphs using a Context Free Grammar (CFG). It takes two JSON files as inputs: a rule dictionary containing phrase structure rules and a lexical dictionary containing words grouped by parts of speech. It outputs the file generated_paragraphs.txt, which contains the randomly generated sentences split between one or more paragraphs.

By having the dictionaries separate from the program, the hope is to eventually allow for the generation of paragraphs in a variety of languages. However, this is a long-term goal, as the program is still being developed to handle a wide range of English sentence generation possibilities.

The Rule Dictionary

The rule dictionary contains phrase structure rules and weights for the outputs. These weights are arbitrarily chosen as their purpose is not to mimic the distribution of various constructions in English sentences, but to ensure that sentences do not expand infinitely.

The format of the entries looks like the following:

{"NP":
    {"outputs":
        ["AdjP N", "N PP", "N Conj N", "N"],
    "weights":
        [15, 30, 40, 90]
    }
}

The Lexical Dictionary

The lexical dictionary contains lists of words organized by parts of speech. Currently the set of words is small for testing purposes.

The format of the entries looks like the following:

{"P":
    ["in", "with", "to", "for", "on", "by"]
}

Current Tasks

• Re-write dictionary files from XML into JSON
• Re-write current version of program from C# into Python
• Develop a system for tracking agreement
• Develop a system for handling multiple tenses
• Develop a system for handling verb complements
• Develop a system for question inversion
• Investigate HPSGs to see if those would work better