<aside> 💡 Here’s a detailed explanation of the HCU grader tool that I have designed. Feel free to reach out to me to ask any questions you may have.

</aside>

⚠️ The Problem

The Helpful Content Update has clearly had a negative impact on MakeUseOf. The best way to turn things around is to remove unhelpful content from the site.

However, the website has a large number of pages, specifically over 60,000, even after excluding the massive Answers and other categories.

So, how do we begin the process of auditing the website to identify and remove the "unhelpful" pages?

✅ The Solution

Well, one thing's for sure, to accomplish this at a page level, we need to find an automated solution. That's precisely why I've designed and built an AI content auditor. This auditor crawls every link on the website to assess and grade the helpfulness of each article.

It utilizes GPT 4's API and employs a carefully crafted combination of prompts to analyze articles based on various metrics. It assigns scores for each metric and generates an overall score for the article.

I've named this tool "HCU Grader" for convenience.

⚙️ How Does The HCU Grader Work?

Tool Architecture

The tool is ingeniously designed with four core components to streamline its functionality.

  1. URL Importer: Capable of ingesting an infinite list of URLs, this module eliminates the constraints on the number of articles you can process at once and basically bypasses ChatGPT’s limitations.
  2. Web Scraper: This segment autonomously navigates and extracts content from the provided URLs, pulling in each article for further analysis.
  3. AI-Powered Scoring Engine: An AI-based scoring mechanism that analyzes the article by following predefined logic and grades the article.
  4. CSV Parser: Finally, the tool compiles all the graded articles and their corresponding scores, exporting this data into a well-structured CSV file for easy consumption and further analysis.

This integrated approach ensures a seamless workflow, from URL importation to final data export.

💡The Logic Behind Grading An Article

To grade an article, we need a weighted system that can assign different values to different aspects of the article. Here is an explanation of the logic behind the weighted scoring system used by HCU Grader. It primarily focuses on the following criteria 👉🏻

  1. Written Content Quality (27 Points)
  2. Depth and Substance (25 Points)
  3. Speed in Answering Main Question (8 Points)
  4. Engagement and Structure (28 Points)
  5. Originality and Uniqueness (12 Points)

Each of these metrics is further broken down into a narrower and more specific set of parameters. This ensures that the results are not inconsistent and that the tool checks everything we want it to.

Here is a further breakdown of the scoring system.