data co-occurrence in JSON dataset

Question

I'm having trouble extracting JSON information. My JSON file contains 100 chapters of a novel. Each chapter contains a number of characters found in the chapter.

For instance:

{"ONE": ["PERSON A", "PERSON B", "PERSON C", "PERSON D", "PERSON A"],
"TWO": ["PERSON A", "PERSON D", "PERSON F", "PERSON G", "PERSON H"],
"THREE": ["PERSON F", "PERSON D", "PERSON A", "PERSON A", "PERSON A"]
... "ONE HUNDRED": ["PERSON B", "PERSON A"]
}

My goal is to design a method to extract how many times two characters co-occurred in the whole book, and two characters can only co-occur once in a chapter. eg, within 100 chapters, I want to know how many times PERSON A and PERSON B co-occurred.

I have two methods in mind, A. Use JSON PATH and filter out the dataset (where PERSON A and B co-occurred), and calculate the number of chapters they co-occurred. (I don't know what to query either :P ) B. Although I'm not really good with JAVASCRIPT. My idea is to define an integer, and then run for loops in every chapter of the JSON file.

I wonder if you guys could share your knowledge with me on this! Thanks!

May you share those two methods? – evolutionxbox Jul 18 '21 at 19:41 — evolutionxbox, Jul 18 '21 at 19:41

score 0 · Accepted Answer · answered Jul 18 '21 at 19:31

Here's a function where you can specify if you want the count or an array of chapters

Here is the function broken down

const cooccur = (people, rettype) => {
  let result = Object.keys(
  // the final result will be an array of object keys
     Object.fromEntries(Object.entries(chapters)
     // but to iterate your object, we need to first convert it into an array with Object.entries
     // then with that result, convert it back into an object with Object.fromEntries
        .filter(c => people.filter(r => c[1].indexOf(r) > -1).length === people.length)));
         // this double filter will run through each chapter and filter it based on the second filter's result
         // the second filter takes our people array and finds how many total occurences of both people in a given chapter
         // if the total number of occurences equals the number of people we're searching for, it's a match
  return rettype === 'count' ? result.length : result;
}

let chapters = {
  "ONE": ["PERSON A", "PERSON B", "PERSON C", "PERSON D", "PERSON A"],
  "TWO": ["PERSON A", "PERSON D", "PERSON F", "PERSON G", "PERSON H"],
  "THREE": ["PERSON F", "PERSON D", "PERSON A", "PERSON A", "PERSON A"],
  "ONE HUNDRED": ["PERSON B", "PERSON A"]
}

const cooccur = (people, rettype) => {
  let result = Object.keys(Object.fromEntries(Object.entries(chapters).filter(c => people.filter(r => c[1].indexOf(r) > -1).length === people.length)));
  return rettype === 'count' ? result.length : result;
}

console.log('number of occurences:', cooccur(["PERSON A", "PERSON B"], 'count'));
console.log('occurence chapters:', cooccur(["PERSON A", "PERSON B"], 'chapters'));

of course it does! :) if this solved your question consider accepting it as the solution by clicking the checkmark to the left of the answer. Cheers — Kinglish, Jul 18 '21 at 20:08
another cool feature: this will work for any number of people. Add as many as you want to that funciton argument and it will work - example: `cooccur(["PERSON A", "PERSON B", "PERSON C"], 'chapters')` — Kinglish, Jul 18 '21 at 20:18

score 0 · Answer 2 · answered Jul 18 '21 at 21:30

Probably go with @Kinglish's answer, but I wanted to add this for completeness.

Proper JSON Path doesn't have a syntax for this yet, but we're building an official spec, so now's the best time to propose one. We've actually been working on what expression syntaxes to support recently. I've referenced this question in a comment that explains the proposal.

data co-occurrence in JSON dataset

2 Answers2