Citation

PEREIRA, V.; BASILIO, M.P.; SANTOS, C.H.T. (2025). PyBibX: A Python Library for Bibliometric and Scientometric Analysis Powered with Artificial Intelligence Tools. Data Technologies and Applications. Vol. 59, Iss. 2, pp. 302-337. doi: https://doi.org/10.1108/DTA-08-2023-0461

pybibx Web App

New to Python or prefer a graphical interface? The pybibx Web App lets you run your analysis in clicks, not lines of code.

import pybibx

# Start the web service using:
pybibx.web_app()

# Terminate the web service using:
pybibx.web_stop()

• Preview -- pybibx Web App -- in Google Colab

This Google Colab Demo is intended for quick demos only. For the best experience, run the Web UI locally or open it directly in a full browser.

Introduction

A Bibliometric and Scientometric Python library that uses raw files generated by Scopus (.bib files or .csv files), WoS (Web of Science) (.bib files), PubMed (.txt files), and OpenAlex. For OpenAlex, the library supports JSON files obtained through the OpenAlex API and CSV (standard) files exported from the OpenAlex website. Also powered by Advanced AI technologies for analyzing bibliometric, scientometric, and textual data.

To export the correct file formats from Scopus, Web of Science, PubMed, and OpenAlex, follow these steps:

a) Scopus: Search, select articles, click Export, choose BibTeX or CSV, select all fields, and click Export again. When using the CSV format, the exported files include the References for the articles.

b) WoS: Search, select articles, click Export, choose Save to Other File Formats, select BibTeX, choose all fields, and click Send.

c) PubMed: Search, select articles, click Save, choose PubMed format, and click Save to download a .txt file. The exported files do not contain the References for the articles.

d) OpenAlex: API Option -> Retrieve records through the OpenAlex REST API and save the response as a .json file. OpenAlex's official programmatic interface is its API, designed for structured, machine-readable access (OpenAlex Developers). Website Option -> Search and filter records in the OpenAlex website, click Export, and choose CSV (standard). In particular, from the Website option, the exported files do not contain the References for the articles.

e) Compare: Users can compare records exported from Scopus, WoS, PubMed, and OpenAlex using pybibx.compare_sources(). This function identifies shared and unique documents across databases, evaluates source overlap, reports metadata completeness, indicates which documents were matched, and estimates the contribution of each database to the final corpus.

General Capabilities:

• a) Works with Scopus (.bib files or .csv files), WoS (.bib files), PubMed (.txt files), and OpenAlex (API JSON files and website CSV (standard) files).
• b) Identification and Removal of duplicates
• c) Entity Normalization for Authors, Institutions, Countries, Journals, Keywords, and References, with editable mapping files so users can safely standardize variants such as abbreviated author names, institution spellings, keyword synonyms, and reference variants before analysis.
• d) Identification of documents per type
• e) Generates a Health Report to evaluate the quality of the .bib/.csv file
• f) Generates an EDA (Exploratory Data Analysis) Report: Publications Timespan, Total Number of Countries, Total Number of Institutions, Total Number of Sources, Total Number of References, Total Number of Languages (and also the number of docs for each language), Total Number of Documents, Average Documents per Author, Average Documents per Institution, Average Documents per Source, Average Documents per Year, Total Number of Authors, Total Number of Authors Keywords, Total Number of Authors Keywords Plus, Total Single-Authored Documents, Total Multi-Authored Documents, Average Collaboration Index, Max H-Index, Total Number of Citations, Average Citations per Author, Average Citations per Institution, Average Citations per Document, Average Citations per Source
• g) Creates an ID (Identification) for each Document, Authors, Sources, Institutions, Countries, Authors Keywords, Keywords Plus and References. The IDs can be used in graphs/plots to obtain a cleaner visualization
• h) Generates Profiling Functions that are a detailed summary (profile) for any selected entity: Authors, Sources, Institutions, Countries, Authors Keywords, Keywords Plus and References. For each entity, the function aggregates key statistics and metadata, including the list of associated publications, total and average citations, time span of activity, and more.
• i) Creates an Authors' Metrics Table with H-Index, E-Index, G-Index, M-Index and J-Index
• j) Creates a WordCloud from the Abstracts, Titles, Authors Keywords or Keywords Plus
• k) Creates a N-Gram Bar Plot (interactive plot) from the Abstracts, Titles, Authors Keywords or Keywords Plus
• l) Creates a Projection (interactive plot) of the documents based on the Abstracts, Titles, Authors Keywords or Keywords Plus
• m) Creates a Term Growth Plot (interactive plot) showing the temporal evolution of terms from Abstracts, Titles, Sources, Author Keywords, or Keywords Plus.
• n) Creates a Sankey Diagram (interactive plot) with any combination of the following keys: Authors, Countries, Institutions, Journals, Auhors Keywords, Keywords Plus, and/or Languages
• o) Creates a XY Bar Chart (interactive plot) that displays the count of a specified Y-key for each category defined by a specified X-key. The plot visualizes the distribution of entities such as Authors, Countries, Institutions, Journals, Authors Keywords, Keywords Plus, and/or Languages. For example, it can show how many Authors publications are associated with each Country
• p) Creates a XY Heatmap (interactive plot) that displays the count and papers' ID of a specified Y-key for each category defined by a specified X-key. The plot visualizes the distribution of entities such as Authors, Countries, Institutions, Journals, Authors Keywords, Keywords Plus, and/or Languages. For example, it can show how many Authors publications are associated with each Authors Keywords and the papers associated to them.
• q) Creates a TreeMap (interactive plot) from the Authors, Countries, Institutions, Journals, Auhors Keywords, or Keywords Plus
• r) Creates an Authors Productivity Plot (interactive plot) It informs for each year the documents (IDs) published for each author
• s) Creates a Countries Productivity Plot (interactive plot) It informs for each year the documents (IDs) published for each country (each author's country)
• t) Creates a Institutions Productivity Plot (interactive plot) It informs for each year the documents (IDs) published for each institution (each author's institution)
• u) Creates a Sources Productivity Plot (interactive plot) It informs for each year the documents (IDs) published for each source (journal)
• v) Creates a Bar Plot (interactive plot) for the following statistics: Documents per Year, Citations per Year, Past Citations per Year, Lotka's Law, Sources per Documents, Sources per Citations, Authors per Documents, Authors per Citations, Authors per H-Index, Bradford's Law (Core Sources 1, 2 or 3), Institutions per Documents, Institutions per Citations, Countries per Documents, Countries per Citations, Language per Documents, Keywords Plus per Documents and Authors Keywords per Documents
• w) Creates a Top Reference Plot (interactive plot) to visualize the top cited References
• x) Creates a Citation Trajectory Plot (interactive plot) that shows the yearly citation counts for each selected Reference
• y) Creates a Citation Matrix that shows for each Reference, which citing articles (with their publication years) mentioned that Reference, the Unique Reference ID, and the Reference's publication year
• z) Creates a RPYS-Reference Publication Year Spectroscopy (interactive plot) to visualize citation patterns over the years. Revealing the peaks in reference publication years (trough Gaussian Filters) that might indicate influential works or shifts in research trends

Network Capabilities:

• a) Creates a Top Reference Set Matrix, which, for a given group size n, identifies and returns the most frequently co-cited Reference groups of size n
• b) Creates a Reference Co-Citation Network (interactive plot) that visually displays the top n References that are most frequently cited together with a target Reference
• c) Collaboration Plot between Authors, Countries, Institutions, Authors Keywords or Keywords Plus
• d) Computes Hubs & Authorities scores for papers in a citation network, and can also find the top-ranked nodes per decade.
• e) Identifies Sleeping Beauties papers that were uncited for a long time but later received sudden attention.
• f) Identifies the Princes of the Sleeping Beauties.
• g) Citation Analysis (interactive plot) between Documents (Blue Nodes) and Citations (Red Nodes). Documents and Citations can be highlighted for better visualization
• h) Collaboration Analysis (interactive plot) between Authors, Countries, Institutions or Adjacency Analysis (interactive plot) between Authors Keywords or Keywords Plus. Collaboration and Adjacency can be highlighted for better visualization
• i) Similarity Analysis (interactive plot) can be performed using coupling or cocitation methods
• j) World Map Collaboration Analysis (interactive plot) between Countries in a Map
• k) Main Path Analysis identifies the main citation backbone of the field using search path methods such as SPC, SPLC, or SPNP
• l) Creates a Temporal Scholarly Graph Explorer (interactive HTML plot) to explore the bibliometric dataset as a heterogeneous temporal graph. It supports timeline, force, and ego views; allows centering the analysis on Papers, References, Authors, Journals, Institutions, Countries, Authors Keywords, or Keywords Plus; and can export a standalone HTML file for interactive exploration.

Scientometric Capabilities:

• a) Performs Portfolio Analysis to classify entities, such as Authors, Journals, Institutions, Countries, Authors Keywords or Keywords Plus, according to productivity and impact indicators.
• b) Performs Specialization Analysis to measure how concentrated or diversified the scientific production is across entities, themes, or conceptual fields.
• c) Computes Collaboration Impact indicators to evaluate how collaborative patterns relate to scientific visibility, citation performance, and productivity.
• d) Performs Burst Detection to identify Papers, Authors, Journals, Authors Keywords, Keywords Plus or other entities with sudden increases in scientific attention over time.
• e) Performs Knowledge Diffusion Analysis to trace how concepts move between entities, using heatmaps, chord diagrams, Sankey diagrams, or automatic visualization selection.
• f) Computes Reference Diversity indicators for Papers, such as breadth, depth, temporal diversity, and concentration of cited References.
• g) Computes the Disruption Index to identify whether papers are more disruptive or more developmental within the citation network.
• h) Computes Collaboration Brokerage indicators to identify Authors, Institutions, or Countries that act as structural bridges in the collaboration network, connecting otherwise separated scientific communities through co-authorship, inter-institutional, or international collaboration patterns.

Artificial Intelligence Capabilities:

• a) Topic Modelling using BERTopic to cluster documents by topic
• b) Visualize topics distribution
• c) Visualize topics by the most representative words
• d) Visualize documents projection and clusterization by topic
• e) Visualize topics heatmap
• f) Visualize topics over time
• g) Find the most representative documents from each topic
• h) Find the most representative topics according to a word
• i) Find how each word in its abstract semantically aligns with all topics in the model
• j) Creates W2V Embeddings from Abstracts
• k) Find Documents based in words
• m) Calculates the cosine similarity between two words
• n) Make operations between W2V Embeddings
• o) Visualize W2V Embeddings operations
• p) Creates Sentence Embeddings from Abstracts, Titles, Authors Keywords or Keywords Plus
• q) Abstractive Text Summarization using PEGASUS on a set of selected documents or all documents
• r) Abstractive Text Summarization using chatGPT on a set of selected documents or all documents. Requires the user to have an API key (https://platform.openai.com/account/api-keys)
• s) Abstractive Text Summarization using Gemini on a set of selected documents or all documents. Requires the user to have an API key (https://ai.google.dev/gemini-api/)
• t) Extractive Text Summarization using BERT on a set of selected documents or all documents
• u) Ask chatGPT to analyze the following results: EDA Report, WordCloud, N-Grams, Evolution Plot, Sankey Diagram, Authors Productivity Plot, Countries Productivity Plot, Institutions Productivity Plot, Sources Productivity Plot, Bar Plots, Citation Analysis, Collaboration Analysis, Similarity Analysis, and World Map Collaboration Analysis (consult Example 08). Requires the user to have an API key (https://platform.openai.com/account/api-keys)
• v) Ask Gemini to analyze the following results: EDA Report, WordCloud, N-Grams, Evolution Plot, Sankey Diagram, Authors Productivity Plot, Countries Productivity Plot, Institutions Productivity Plot, Sources Productivity Plot, Bar Plots, Citation Analysis, Collaboration Analysis, Similarity Analysis, and World Map Collaboration Analysis (consult Example 09). Requires the user to have an API key (https://ai.google.dev/gemini-api/)

Correction and Manipulation Capabilities:

• a) Filter the .bib, .csv or .txt file by Year, Sources, Bradford Law Cores, Countries, Languages and/or Abstracts (Documents with Abstracts)
• b) Merge Authors, Institutions, Countries, Languages, Sources and/or References that have multiple entries
• c) Merge different or the same database files one at a time. The preference for information preservation is given to the old database, so the order of merging matters (consult Examples 04 and 05)

Usage

1. Install

pip install pybibx

2. Try it in Colab:

• Example 00: Web App -- No Code -- ( Colab Demo )
• Example 01: Scopus ( Colab Demo )
• Example 02: WOS ( Colab Demo )
• Example 03: PubMed ( Colab Demo )
• Example 04: OpenAlex ( Colab Demo )
• Example 05: Scopus + WOS ( Colab Demo )
• Example 06: WOS + Scopus ( Colab Demo )
• Example 07: Scopus + WOS + Pubmed ( Colab Demo )
• Example 08: Compare Databases ( Colab Demo )
• Example 09: Temporal Scholarly Graph ( Colab Demo )
• Example 10: Your Own ( Colab Demo )
• Example 11: Ask chatGPT Analysis ( Colab Demo )
• Example 12: Ask Gemini Analysis ( Colab Demo )

Acknowledgement

This section indicates the libraries that inspired pybibx

• BERT (https://smrzr.io/):

  a) Github: https://github.com/dmmiller612/bert-extractive-summarizer

  b) Paper: DEREK, M. (2019). Leveraging BERT for Extractive Text Summarization on Lectures. arXiv. doi: https://doi.org/10.48550/arXiv.1906.04165

• SciBERT (https://huggingface.co/allenai/scibert_scivocab_uncased):

  a) Github: https://github.com/allenai/scibert

  b) Paper: BELTAGY, I.;, LO, K.; COHAN, A. (2019). SCIBERT: A Pretrained Language Model for Scientific Text. arXiv. doi: https://doi.org/10.48550/arXiv.1903.10676

• BERTopic (https://maartengr.github.io/BERTopic/index.html):

  a) Github: https://github.com/MaartenGr/BERTopic

  b) Paper: GROOTENDORST, M. (2022). BERTopic: Neural Topic Modeling with a Class-based TF-IDF Procedure. arXiv. doi: https://doi.org/10.48550/arXiv.2203.05794

• Bibliometrix (https://www.bibliometrix.org/home/):

  a) Github: https://github.com/massimoaria/bibliometrix

  b) Paper: ARIA, M.; CUCCURULLO, C. (2017). Bibliometrix: An R-tool for Comprehensive Science Mapping Analysis. Journal of Informetrics, 11(4), 959-975. doi: https://doi.org/10.1016/j.joi.2017.08.007

• Gemini (https://gemini.google.com/app):

  a) Github: https://github.com/google-gemini

  b) Paper: Gemini Team Google (2024). Gemini: A Family of Highly Capable Multimodal Models. arXiv. doi: https://arxiv.org/abs/2312.11805

• Gensim (https://radimrehurek.com/gensim/):

  a) Github: https://github.com/piskvorky/gensim

  b) Paper: REHUREK, R.; SOJKA, P. (2010). Software Framework for Topic Modelling with Large Corpora. LREC 2010. doi: https://doi.org/10.13140/2.1.2393.1847

• chatGPT (https://chat.openai.com/chat):

  a) Github: https://github.com/openai

  b) Paper: OPENAI. (2023). GPT-4 Technical Report. arXiv. doi: https://doi.org/10.48550/arXiv.2303.08774

• Metaknowledge (http://www.networkslab.org/metaknowledge):

  a) Github: https://github.com/UWNETLAB/metaknowledge

  b) Paper: McILROY-YOUNG, R.; McLEVEY, J.; ANDERSON, J. (2015). Metaknowledge: Open Source Software for Social Networks, Bibliometrics, and Sociology of Knowledge Research.

• SentenceTransformers (https://www.sbert.net/):

  a) Github: https://github.com/UKPLab/sentence-transformers

  b) Paper: REIMERS, N.; GUREVYCH, I. (2019). Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. arXiv. doi: https://arxiv.org/abs/1908.10084

• PEGASUS (https://ai.googleblog.com/2020/06/pegasus-state-of-art-model-for.html?m=1):

  a) Github: https://github.com/huggingface/transformers

  b) Paper: ZHANG, J.; ZHAO, Y.; SALEH, M.; LIU, P.J. (2019). PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization. arXiv. doi: https://doi.org/10.48550/arXiv.1912.08777

And to all the people who helped to improve or correct the code. Thank you very much!

• Fabio Ribeiro von Glehn (29.DECEMBER.2022) - UFG - Federal University of Goias (Brazil)