A simple markup language to write novel with types.


Keywords
TypeNovel, novel, compiler, language, TypeScript, javascript, fsharp, machine-learning, markup-language, natural-language-processing, writing, writing-tool
License
MIT
Install
npm install typenovel@1.2.0

Documentation

TypeNovel

A simple markup language to write novel with types.

Example

This is what it looks like.

@scene({
  season:"winter"
}){
  Finally, $season("Xmas") has come!
}

Output is here.

<scene data-season="winter">
  Finally, <season>Xmas</season> has come!
</scene>

For more information, take a look at Cheatsheet.

Install

npm install -g typenovel

Then /usr/local/bin/tnc would be installed.

[foo@localhost] tnc --version
1.0.4

Use compiler from your project

You can compile from source by Tnc.fromString.

import { Tnc } from 'typenovel';

const result = Tnc.fromString('@scene(){ foo }', {
  format: 'html', // or 'text'
  minify: false
});

console.error(result.errors);
console.log(result.output);

Or you can compile from file by Tnc.fromFile.

import { Tnc } from 'typenovel';

const result = Tnc.fromFile('sample.tn', {
  format: 'html',
  minify: false
});

console.error(result.errors);
console.log(result.output);

Reader Application

Reader application is available for TypeNovel(compiler included).

Take a look at TypeNovelReader.

Motivation

What is the difference between pro novelist and amature novelist?

I'm running a novel posting platform over recent decades, and I always thought about differences between pro and amature.

I think that amature novelist sometimes lack some context description in their story, especially tend to lack the time description.

To confirm this feeling, I measured the scores how many times they wrote time in their novel, and gave them quolity scores.

The formula is simple.

[Total count of time written sentence] * [Entropy of time written sentence] / [Total sentence count]

For example, score of "Mon"('門' in Japanese) written by 'Natsume Soukeki (1867-1916)' is like this.

histgram: [30, 24, 33, 21, 29, 29, 30, 24, 24, 17]
score: 0.221913(total sentence size = 3879, total time sentences = 261, entropy = 3.298082)

In this way, I measured the average scores of many writers in my platform.

Here is scores of amature writers.

user1: average score:0.055084
user2: average score:0.057143
user3: average score:0.020505
user4: average score:0.071971
user5: average score:0.045247
user6: average score:0.042603
user7: average score:0.086316
user8: average score:0.051926
user9: average score:0.089533

And here is scores of pro writers.

pro1: average score:0.120076, Arthur Conan Doyle
pro2: average score:0.210235, Anton Pavlovich Chekhov
pro3: average score:0.124371, James Augustine Aloysius Joyce
pro4: average score:0.101132, Feodor Dostoyevsky
pro5: average score:0.138546, Nomura Kodou(野村胡堂)
pro6: average score:0.153463, Mori Ougai(森鴎外)
pro7: average score:0.148902, Christian Andersen
pro8: average score:0.110636, Natsume Souseki(夏目漱石)
pro9: average score:0.118976, Nakajima Atsushi(中島敦)

Clearly pro scores around 0.1, but amature scores around 0.05, difference is not so small.

So I thought "How can we fill the differences between pro and amature?".

In other words, "Can we fill the diffecence using some kind of technology?".

My answer is "novel with type"

I think we can force all writers to fill enough information in their text, to use typing.

But what is the type of novel?

I think it's pair of constraint and annotation.

This is what it looks like.

// block level tag starts with '@'
@scene({
  season: "winter", // constraint 'season'
  time: "7:00AM" // constraint 'time'
}){
  // annotation tag starts with '$'.
  // annotated constraint 'time(7:00AM)' as 'morning'
  I woke up at the $time("morning"). // I woke up at the morning.

  // annotated constraint 'season(winter)' as 'Xmas'
  Finally $season("Xmas")!
}

In this example, we have two constraints(season as "winter" and time as "7:00AM") for this scene block.

So we have to annotate them by writing annotation tag.

And we annotated them by $time("morning") and $season("Xmas").

Now we annotated all constraints, let's compile this code!

[foo@localhost]$ tnc sample.tn

And you get html output like this.

<scene data-season="winter" data-time="7:00AM">
  I woke up at the <time>morning</time>.
  "Finally <season>Xmas</season>!"
</scene>

This output is NLP(Natural Language Processing) or ML(Machine Leraning) friendly, so both platform and writers win!

Writers merit: quality of novel

Platform merit: quality of data

By the way, if we don't annotated our constraints, what happens?

Offcource, you'll get errors at the compilation time.

/Users/u1/sample.tn(line:1) 'season' is not annotated in this block!
/Users/u1/sample.tn(line:2) 'time' is not annotated in this block!