Review Reliability Scoring

Overview

This repository implements a methodology to evaluate the reliability and truthfulness of user reviews. The scoring is based on metrics such as helpfulness votes, review recency, and reviewer ranking. The dataset used is the Amazon 500ECE dataset.

Key Features

• Helpfulness Score: Calculated using the ratio of helpful votes to total votes, weighted by their squared value.
• Reliability Score: Assessed using two approaches:
  1. Most Recent Reviewer: Prioritizes the recency of reviews.
  2. Top-Ranking Reviewer: Prioritizes highly ranked reviewers.
• Truthfulness Score: Combines helpfulness and reliability scores.

Methodology

The methodology follows the formulas outlined below:

Helpfulness Score

For a review by user uᵢ on item pⱼ, the helpfulness score is computed as: