---
title: Liberata's Scientometrics Framework
url: https://www.emergentmind.com/topics/liberata-s-scientometrics
type: topic
---

# Liberata's Scientometrics Framework

Searching arXiv for the cited Liberata and Scholia papers to ground the article in current arXiv metadata.
{"query":"au:Nielsen OR ti:Scholia and scientometrics with Wikidata", "max_results": 5}
{"query":"ti:\"Liberata -- Graph Scientometrics for a Share Based System of Academic Publishing\"", "max_results": 5}
Liberata's scientometrics denotes a share based framework for academic publishing and quality control in which authorship positions are replaced with contribution shares that sum to unity and encode both ordinality and relative contribution distances, while citations are weighted and corrected through graph-based constructions to derive academic capital and scientometrics for impact, risk, collaboration, collusion, value of quality control, and diversification [2605.02128]. In a related but methodologically distinct line, Scholia provides a Wikidata-based infrastructure for scientometric profiling by querying the Wikidata Query Service to assemble publication lists, publication and citation timelines, co-author networks, citation graphs, and LaTeX/BIBTeX bibliographies [1703.04222].

## 1. Share-based credit allocation

The framework begins from a share-based authorship model. For a manuscript \(m\) with contributor set \(C_m=\{c_1,\dots,c_n\}\), each contributor receives a share \(s_{m,c}\in[0,1]\), and the shares satisfy total-share normalization:
\[
\sum_{c\in C_m}s_{m,c}=1.
\]
The shares encode both which author contributed more and by how much more, and they are fungible: two contributors with equal share on the same paper receive equal credit [2605.02128].

Academic-capital allocation on a single paper is defined as
\[
\mathit{AC}_{m,c}=s_{m,c}\,w_m,
\]
where \(w_m\) is the paper’s weighted-citation score. This couples contribution attribution directly to impact attribution. A plausible implication is that credit assignment is no longer mediated by author order alone, because the primitive quantity is the share \(s_{m,c}\) rather than a discrete position in the byline.

The framework is explicitly motivated by the claim that contemporary scientometric indicators remain anchored in paradigms and axioms from when academic research was conducted in small scholarly communities, whereas current academia is organized in large communities with high rates of information incompleteness regarding work impact and individual contributions [2605.02128]. It further argues that traditional metrics force discretization of credit to authors and prior works despite their fundamentally continuous nature. This suggests that Liberata’s scientometrics is intended as a reparameterization of scientometric accounting, with continuity in both authorship and citation flow.

## 2. Quality control as a marketplace for shares

Liberata supports two open markets—peer review and replication—in which authors sell a fraction of their shares to quality controllers in exchange for expected increases in long-run impact [2605.02128]. In the peer-review marketplace, authors submit a pre-review manuscript \(m\) with share vector \(\{s_{m,a}\}_{a\in A_m}\), and they propose post-review shares \(\{s_{m',a}\}\) and reviewer shares \(\{s_{m',p}\}\).

A transaction occurs iff authors expect net benefit:
\[
\mathbb{E}\bigl[\mathit{AC}_{m',A}\bigr]>\mathbb{E}\bigl[\mathit{AC}_{m,A}\bigr]
\quad
\text{with }s_{m',A}<s_{m,A}\text{ and }\mathbb{E}[w_{m'}]>\mathbb{E}[w_m].
\]
A reviewer \(p\) accepts a bid \(s_{m',p}\) if the time \(t_p\) for review is less than \(s_{m',p}\) times the time to produce an original paper of similar expected impact:
\[
t_p < t_a\,s_{m',p}.
\]

The replication marketplace is defined analogously. Replicators \(r\) trade shares \(s_{m'',r}\) for expectations of higher impact \(w_{m''}\), and they compare their own time cost through
\[
t_r<t_a\,s_{m'',r}.
\]
Within the formalism, peer review and replication are not external certification mechanisms; they are integrated into the same share ledger as authorship. This suggests that quality control is treated as a tradable contribution category rather than a separate administrative layer.

## 3. Weighted citations and correction factors

Liberata distinguishes unweighted citation count from weighted citation. The unweighted count for a manuscript \(m\) is
\[
u_m = \sum_{m'\succ m} \mathbf{1}\{m'\text{ cites }m\}.
\]
The default weighted citation score normalizes each incoming citation by the citing paper’s reference-list length:
\[
w_m = \sum_{m'\succ m} \frac{1}{|\mathrm{ref}(m')|}.
\]
Under this rule, each paper can print at most one unit of citation [2605.02128].

Two correction mechanisms are then introduced. First, publication-rate normalization corrects field-size effects by dividing by the per-author publication rate \(\rho(d)\) of the paper’s field \(d\):
\[
w'_m=\frac{w_m}{\rho\bigl(d(m)\bigr)}.
\]
Second, an author-similarity discount suppresses self-citations. If \(\phi_{m',m}=\bar v_{m'}\cdot\bar v_m\) is the cosine-similarity of the two papers’ author-share vectors, then each citation is discounted by \((1-\phi_{m',m})\):
\[
w'_{m',m}=(1-\phi_{m',m})\,\frac1{|\mathrm{ref}(m')|}.
\]

The abstract states that citations are weighted to guard against frivolous referencing and credit inflation, and that modular correction factors allow multiple measures of impact [2605.02128]. In that sense, the weighting rule is not merely a normalization device; it is part of the framework’s quality-control logic. A plausible implication is that the scientometric unit being propagated is not a raw citation event but a bounded and optionally corrected transfer of academic capital.

## 4. Graph-theoretic structure and academic capital

The system is formalized through two fundamental graphs, Shares and References, and from them constructs academic capital [2605.02128].

| Graph | Structure | Function |
|---|---|---|
| Shares graph \(G_S\) | Bipartite, undirected, node set \(V_S=C\sqcup M\) | Encodes contribution shares |
| References graph \(G_W\) | Directed DAG on manuscript nodes \(M\) | Encodes weighted citation flow |
| Capital graph \(G_C\) | Composition of \(G_S\) and \(G_W\) | Records \(\mathit{AC}_{m,c}\) |

The Shares graph \(G_S\) has node set \(V_S=C\sqcup M\), where \(C=A\cup P\cup R\) consists of authors, reviewers, and replicators, and \(M\) is the set of manuscripts. Its edge set is
\[
E_S=\{(c,m)\mid s_{m,c}>0\},
\]
with weight function \(w_S(c,m)=s_{m,c}\). By construction,
\[
\sum_{c\in C}w_S(c,m)=1
\]
for each manuscript \(m\). In expanded matrix form,
\[
\mathbf G_S =
\begin{bmatrix}
0      & S_A    & S_P    & S_R\\
S_A^T  & 0      & 0      & 0\\
S_P^T  & 0      & 0      & 0\\
S_R^T  & 0      & 0      & 0
\end{bmatrix}.
\]

The References graph \(G_W\) is a directed DAG on manuscript nodes. An edge \((m',m)\) exists if \(m'\) cites \(m\), weighted by \(w_{m,m'}\in[0,1]\), typically \(w_{m,m'}=1/|\mathrm{ref}(m')|\). Its adjacency matrix \(\mathbf G_W\in\mathbb R^{|M|\times|M|}\) satisfies \(\mathbf G_W(i,j)=w_{i,j}\) if \(j\to i\).

The capital graph \(G_C\) is obtained by composition of \(G_S\) and \(G_W\). If
\[
\mathbf w:=\mathbf G_W\mathbf1\in\mathbb R^{|M|}
\]
is the vector of paper capitals, then in condensed form
\[
\mathbf G_C' = \bigl[\mathrm{diag}(\mathbf w)\bigr]\;\mathbf G_S'
\]
with \(\mathbf G_S'=[S_A\;S_P\;S_R]\). Full \(G_C\) reconstructs like \(G_S\). This construction makes the relationship between citation flow and contributor-level allocation explicit: manuscript-level weighted impact is first computed on \(G_W\), then allocated across contributors through \(G_S\).

## 5. Scientometric measures derived from portfolios

The framework defines scientometric measures on portfolios \(\Pi\), where a portfolio is a set of shares on papers \(M_\Pi\) by contributors \(C_\Pi\) [2605.02128]. Total academic capital is
\[
\mathit{AC}_\Pi = \sum_{m\in M_\Pi}\sum_{c\in C_\Pi}s_{m,c}\,w_m.
\]
Return over \(\Delta t\) is
\[
R_\Pi=\frac{\Delta\,\mathit{AC}_\Pi}{\Delta t},\qquad
\mu_\Pi=\mathbb E[R_\Pi],\qquad
\sigma_\Pi=\sqrt{\mathbb E\bigl[(R_\Pi-\mu_\Pi)^2\bigr]}.
\]
Risk asymmetry is measured by skewness,
\[
\gamma_\Pi=\frac{\mathbb E[(R_\Pi-\mu_\Pi)^3]}{\sigma_\Pi^3},
\]
the Sharpe-style ratio is
\[
\mathrm{Sharpe}_\Pi=\mu_\Pi/\sigma_\Pi,
\]
and academic-returns-to-capital is
\[
\mathrm{ARC}_\Pi=\mu_\Pi/\mathit{AC}_\Pi.
\]

Allocation concentration is expressed through three statistics. With weights \(\omega_m=\mathit{AC}_m/\mathit{AC}_\Pi\), the Herfindahl–Hirschman index is
\[
\mathrm{HHI}_\Pi=\sum_{m\in\Pi}\omega_m^2,
\]
the Gini coefficient is
\[
\mathrm{Gini}_\Pi=\tfrac1{2|\Pi|}\sum_{k,l}|\omega_k-\omega_l|,
\]
and normalized entropy is
\[
H_\Pi=-\bigl(\sum\omega_m\ln\omega_m\bigr)/\ln|\Pi|.
\]
Diversification is measured by
\[
\mathrm{DR}_\Pi
=\frac{\sum_{m\in\Pi}\omega_m\,\sigma_m}{\sigma_\Pi}
\ge 1.
\]

Collaboration and collusion indicators are derived from the two-step shares graph \(G_S^2\). The block
\[
\bigl(G_S^2\bigr)_{AA}=S_A^TS_A
\]
measures co-authorship intensity, while excessively large off-diagonal entries in the author–reviewer subblock \(S_A S_P\) may flag collusion. The framework also defines a fair market price for peer review in field \(d\), \(\overline s_{p,d}=\mathbb E[s_{p,d}]\), and a review-risk premium for author group \(\hat A\),
\[
\psi_{p,d}(\hat A)=\hat s_{p,d}-\overline s_{p,d},
\]
with analogous definitions for replication. These constructions broaden scientometrics beyond impact measurement alone by placing risk, concentration, diversification, and quality-control pricing inside the same accounting system.

## 6. Aggregation, field structure, and Wikidata-based operationalization

The framework extends naturally to aggregates such as institutions, regions, time periods, and research fields [2605.02128]. A portfolio \(\Pi_{M',C'}\) collects shares \((m,c)\) with \(m\in M'\) and \(c\in C'\). Choosing \(M'\) as all papers from an institution and \(C'\) as its members yields the institution’s total academic capital, risk, concentration, and related measures. For a region \(\theta\), total capital is
\[
\mathit{AC}_\theta=\sum_{i\in\theta}\mathit{AC}_i,
\]
and per-capita or per-GDP measures and their Gini indices can be formed in the same way.

Field-level metrics use a 4D taxonomy \(D_1\supset D_2\supset D_3\supset D_4\) described as domain \(\to\) department \(\to\) discipline \(\to\) direction. Field-normalized returns are formed by dividing by \(\mu_d\), and field-tag allocation weights
\[
\omega_{d_k}=\mathit{AC}_{d_k}/\mathit{AC}_\Pi
\]
allow HHI and Gini at each level. A journal is defined simply as the collection of papers carrying a given set of \(d_4\) tags, and its impact becomes average paper capital:
\[
\mathbb E(\mathit{AC}_J)=\mathit{AC}_J/|J|.
\]

A distinct operational route to scientometric profiling is provided by Scholia. Scholia is implemented in Python with the Flask framework and exposes aspects such as author, work, and organization by embedding live SPARQL queries against the Wikidata Query Service; the front-end uses HTML iframes for each panel, each iframe points to a WDQS HTTP endpoint, and the Python package builds and injects SPARQL strings, sends them to `https://query.wikidata.org/sparql`, and handles the returned JSON to produce visualizations or downloadables such as BibTeX [1703.04222]. For a researcher such as Liberata, the documented queries include total publications per year, co-author network data, citations per year to all works, overall citation-graph edges, and citation-count retrieval for client-side computation of the h-index, g-index, and i10-index.

The sample interpretations are explicit. Publications per year are visualized as a bar chart for productivity trend; the co-author network uses node size proportional to total collaborations and edge width proportional to number of co-authored papers; citations per year appear as a line chart for impact over time, with cumulative citations as a second line; and the h-index is presented as a single-number summary of both productivity and impact, while the g-index gives more weight to highly-cited papers and is often greater than \(h\), and the i10-index counts moderately-cited works with at least 10 citations [1703.04222]. Scholia also supports BibTeX generation through `python -m scholia.tex write-bib-from-aux example.aux`, allowing `\cite{Q12345678}` or `\cite{10.1000/xyzdoi}` in LaTeX and resolution to the Wikidata item matching that DOI.

Taken together, these two lines of work occupy different levels of abstraction. Liberata’s scientometrics specifies a graph-based accounting and marketplace framework for credit, impact, and quality control, whereas Scholia specifies a Wikidata-centric infrastructure for querying, formatting, and visualizing scholarly metadata. This suggests a practical distinction between a formal scientometric model and an implementation substrate for scientometric profiles.

Source: https://www.emergentmind.com/topics/liberata-s-scientometrics