markov.js

function compute_ngrams(sentences, order) {
	const ngrams = {"": []}
	for(let i=0; i<sentences.length; ++i) {
		const words = sentences[i].split(/\s+/)
		for(let j=0; j<words.length-order; ++j) {
			const Gram = words.slice(j, j+order).join(' ')
			if(j === 0) ngrams[""].push(Gram)
			const gram = Gram.toLowerCase()
			const next = words[j+order]
			if(ngrams[gram] == null) ngrams[gram] = []
			ngrams[gram].push(next)
		}
	}
	return ngrams
}

function generate_sentence(ngrams, rnd) {
	const choose = (a) => a[Math.floor(rnd()*a.length)]
	let sentence = choose(ngrams['']).split(' ')
	const order = sentence.length
	while(true) {
		const last = sentence.slice(-order).join(' ').toLowerCase()
		const following = ngrams[last]
		if(following == null) break
		sentence.push(choose(following))
	}
	return sentence
}

function generate_exercise(ngrams, word_count, rnd) {
	let words = []
	let chars_left = word_count * 5 + 1
	while (chars_left > 0) {
		const sentence = generate_sentence(ngrams, rnd)
		chars_left -= 1 + sentence.join(' ').length
		words.splice(words.length, 0, ...sentence)
	}
	return new TypeJig.Exercise(words, 0, false, 'ordered');
}

let jig
window.addEventListener('load', () => jig = loadExercisePage(args => {
	const ngrams = compute_ngrams(sentences, 3)
	const nwords = args.word_count==null ? 100 : parseInt(args.word_count)
	return {
		generate: (rnd, options) => generate_exercise(ngrams, nwords, rnd),
		options: { name: "Markov-chain generated sentences" }
	}
}))