Beyond the h-index: Introducing the Fair Impact Score (FIS) for Network-Driven, Sentiment-Aware Bibliometrics

Traditional academic metrics (such as raw citation counts and the h-index) operate on a linear framework that treats all citations as identical units of value, ignoring contextual topography, structural prestige, and semantic intent. This simplicity leaves bibliometric systems vulnerable to systemic exploitation via citation cartels, self-citation loops, and the artificial inflation of flawed or debunked research. To resolve these vulnerabilities, this repository implements the Fair Impact Score (FIS). The FIS shifts academic evaluation from a flat leaderboard to a graph-based ecosystem by deploying a three-layer correction filtering model. Mathematical Framework The core script translates the primary multi-layered citation algorithm into a functional pipeline: Here’s a polished, journal‑ready rewrite of your equation with smoother phrasing and emphasis on its dynamic weighting logic: FIS(R) = ∑p∈P∑c∈Cp[ Wcentrality(c) × (1−S(p,c)) × σ(c) × γ(c) ] The pipeline evaluates input data across four interdependent dimensions, each designed to capture a distinct aspect of citation behavior and institutional influence: Structural Layer (Wcentrality) Quantifies the prestige of the citing source through recursive graph‑based centrality measures, such as PageRank or Eigenfactor, thereby embedding network‑level authority into the weighting process. Behavioral Layer / Cartel Filter (S(p,c)) Detects and penalizes insular citation practices by measuring overlap in authorship and bibliographic references. High similarity values indicate potential collusion or cartel‑like behavior, reducing the effective citation weight. Semantic Layer (σ) Applies Natural Language Processing (NLP) to the textual context of citations, mapping sentiment into explicit multipliers (Positive = 1.0; Neutral = 0.5; Negative = 0.0). This ensures that the qualitative tone of citation usage directly influences its evaluative weight. Socioeconomic Layer (γ) Implements a lower‑bounded Dynamic Damping Factor (max⁡(0.1,… )) to normalize disparities in institutional funding. By amplifying citations from elite institutions validating work produced in resource‑constrained environments, this adjustment mitigates the Matthew Effect and promotes equity in scholarly recognition. Repository Contents fis_calculator.py: The primary Python 3 execution script implementing the algebraic filters and pipeline logic. README.md: Comprehensive setup documentation, software prerequisite lists, dependency configurations, and terminal execution commands. data/researcher.csv: Baseline data table defining the target researcher profile. data/papers.csv: Tabular metadata capturing the target articles being evaluated. data/citing_network.csv: Anonymized graph network mapping multi-layered citation attributes (prestige, author lists, and text contexts) to feed the execution model. Validation Context The provided sample dataset is configured using the citation profile matrix of the notable epidemiological overview: "Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2): an update" (2020), demonstrating how the algorithm dynamically processes high-prestige endorsements, potential co-author clusters, and negative context triggers in real-time. Keywords Bibliometrics, Scientometrics, Graph Theory, Natural Language Processing, Sentiment Analysis, Fair Impact Score, Citation Cartels, Open Science, H-index, DORA