CoChat

WISDOM: An AI-powered framework for emerging research detection using weak signal analysis and advanced topic modelling

Open access
Ebadi, Ashkan, Auger, Alain, Gauthier, Yvan · 2026 · Journal of Informetrics

The DOI matches its Crossref record; 61 of the 61 references it cites were found in a registry or an index; none of the 65 sources its claims cite is checked for support yet.

Abstract

The landscape of science and technology is characterized by its dynamic and evolving nature, constantly reshaped by new discoveries, innovations, and paradigm shifts. Moreover, science is undergoing a remarkable shift towards increasing interdisciplinary collaboration, where the convergence of diverse fields fosters innovative solutions to complex problems. Detecting emerging scientific topics is paramount as it enables industries, policymakers, and innovators to adapt their strategies, investments, and regulations proactively. As the common approach for detecting emerging technologies, despite being useful, bibliometric analyses may suffer from oversimplification and/or misinterpretation of complex interdisciplinary trends. In addition, relying solely on domain experts to pinpoint emerging technologies from science and technology trends might restrict the ability to systematically analyze extensive information and introduce subjective judgments into the interpretations. To overcome these drawbacks, in this work, we present an automated artificial intelligence-enabled framework, called WISDOM, for detecting emerging research themes using advanced topic modeling and weak signal analysis. The proposed approach can assist strategic planners and domain experts in more effectively recognizing and tracking trends related to emerging topics by swiftly processing and analyzing vast volumes of data, uncovering hidden cross-disciplinary patterns, and offering unbiased insights, thereby enhancing the efficiency and objectivity of the detection process. As the case technology, we assess WISDOM's performance in identifying emerging research as well as their trends, in the field of underwater sensing technologies using scientific papers published between 2004 and 2021.

Overview

CoChat Summary

TL;DR

WISDOM is an AI framework combining BERTopic topic modeling, automated T5-based topic labeling, and weak-signal analysis to detect emerging research themes. In an underwater sensing case study using 9,046 Scopus publications from 2004–2021, it extracted 73 topics and classified 18 as weak signals and 5 as strong signals across three six-year periods.

Key findings

  • BERTopic extracted 73 distinct topics from 9,046 underwater sensing publications; Topic65 (“submarine image enhancement and color correction”) had the highest document count, while Topics34, 55, and 63 had the lowest.
  • Across P1 [2004–2009], P2 [2010–2015], and P3 [2016–2021], 18 topics appeared as weak signals in at least one period; only Topic21 (“underwater imaging”) was a weak signal in two consecutive periods (P1 and P2). None of the weak-signal topics transitioned into strong signals in subsequent periods.
  • Five topics—Topics1, 5, 52, 55, and 60—emerged as strong signals in at least one period, all in P1 except Topic5 in P2. Topics1, 55, and 60 showed upward trends; Topic60 was projected to grow, while Topic5 and 52 declined.
  • Eighteen topics had consistent signal patterns across all periods; five (Topics19, 37, 61, 70, and 71) were latent signals from P1 to P3, and thirteen were NSWK signals throughout. Topic25 appeared only in P2; Topics39 and 67 appeared only in P3.

Methods

  • Data: Scopus publications on underwater sensing from 2004–2021, collected on September 14, 2022; the dataset contained 9,046 publications. Publications lacking a title or abstract were excluded, and only English-language papers were considered. Title and abstract were merged into a `pubtext` feature; preprocessing included lowercasing, stop-word removal, special-character and punctuation handling, tokenization, and lemmatization.
  • Topic modeling: BERTopic with SBERT embeddings, UMAP dimensionality reduction, HDBSCAN clustering, and c-TF-IDF; n-gram range was (1,2); the model was initially set to 100 topics and extracted 73. Automated labeling selected random papers per topic, applied T5 summarization and a T5-based headline generator, repeated the process fifteen times per topic, then reviewed the labels.
  • Weak-signal analysis: topic proportions from BERTopic were used to build Topic Emergence Maps (TEMs) for three six-year intervals; quadrants classified strong, weak, latent, and not-strong-but-well-known (NSWK) signals. Temporal evolution curves included trend lines and 95% confidence intervals, and time-specific keywords were reported for recent peaks.

Limitations

  • Stated: the approach was tested only on scientific publications; not all emerging technologies are immediately documented in publications, and confidential or proprietary innovations cannot be captured. Definitions of emerging technology vary, not all papers are of equal quality, and only English papers were considered; future work could use other sources, focus on high-impact research, include other languages, and monitor weak signals as they evolve.
  • Stated: the method relies on a domain-expert-crafted search query (available upon request) and on title/abstract metadata because full texts were not accessible.
  • Evident: the evaluation is a single-domain case study (underwater sensing) with no external ground-truth validation reported; automated labels were reviewed, but the paper does not quantify labeling accuracy or compare WISDOM against alternative detection methods.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1016/j.joi.2025.101759

Cites

61 of 61 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1 The high-level conceptual flow of the analytical pipeline.
Figure 1 The high-level conceptual flow of the analytical pipeline.CC-BY · From the PDF
(a) Number of papers of a given token length
(a) Number of papers of a given token lengthCC-BY · From the PDF
Image 3
CC-BY · From the PDF
Figure 4 Similarity matrix, top-20 most frequent topics. The numbers in parentheses within the legend indicate the document counts corresponding to each topic. The values in the cells denote the similarity score.
Figure 4 Similarity matrix, top-20 most frequent topics. The numbers in parentheses within the legend indicate the document counts corresponding to each topic. The values in the cells denote the similarity score.CC-BY · From the PDF
Figure 5 Topics' signals in P 1 , P 2 , and P 3 . Black cells indicate no signal.
Figure 5 Topics' signals in P 1 , P 2 , and P 3 . Black cells indicate no signal.CC-BY · From the PDF
Figure 6 The temporal evolution of research topics identified as weak signals across various periods (solid blue line). The corresponding periods are indicated in parentheses alongside the topics. The gray dashed line represents the trend line, with the shaded area highlighting the 95% confidence interval.
Figure 6 The temporal evolution of research topics identified as weak signals across various periods (solid blue line). The corresponding periods are indicated in parentheses alongside the topics. The gray dashed line represents the trend line, with the shaded area highlighting the 95% confidence interval.CC-BY · From the PDF
Figure 7 The temporal evolution of research topics identified as strong signals across various periods (solid blue line). The corresponding periods are indicated in parentheses alongside the topics. The gray dashed line represents the trend line, with the shaded area highlighting the 95% confidence interval.
Figure 7 The temporal evolution of research topics identified as strong signals across various periods (solid blue line). The corresponding periods are indicated in parentheses alongside the topics. The gray dashed line represents the trend line, with the shaded area highlighting the 95% confidence interval.CC-BY · From the PDF

Verification

Claims the paper makes

0 Valid · 7 Not valid · 50 Not checked · 78 Own results and interpretations

36 more claims

Create a free account to see all 135 claims and every source verdict.

See all claims — free account
  • It may state two separate claims as one. 39
  • Probably not a substantial claim. 76

Create a free account to see these claims.

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 61 references

Research works printed but not linked: 1

  1. Huebert, 2022, The arctic as emerging geopolitical flashpoint, Canada-United States Law Journal, 46, 32

Metadata

Authors
Ebadi, A., Auger, A., & Gauthier, Y.
Affiliations
  • Government of Canada (CA)
  • Defence Research and Development Canada (CA)
  • National Research Council Canada (CA)
Corresponding author
Ebadi, Ashkan
Venue
Journal of Informetrics
Publisher
Elsevier BV
Published
2026-03
Volume
20
Issue
1
Pages
101759
ISSN
1751-1577 (ISSN-L)
OpenAlex
W4403785816
Cited by
17
References
59
Citations by year
2026: 11
FWCI
13.85
Type
journal article
Language
en
Open access
Locations
  • Journal of Informetrics (publishedVersion)
  • arXiv (Cornell University) (submittedVersion, cc-by)
  • arXiv (Cornell University) (cc-by)
Topics
  • Social Sciences › Library and Information Sciences › scientometrics and bibliometrics research
  • Social Sciences › General Social Sciences › Computational and Text Analysis Methods
  • Health Sciences › Medicine › Public Health, Environmental and Occupational Health › Diverse Approaches in Healthcare and Education Studies
Keywords
topic modeling, artificial intelligence, interdisciplinary research trends

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-11-20, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-11-20 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: WISDOM: An AI-powered framework for emerging research detection using weak signal analysis and advanced topic modelling. https://app.cochat.ai/papers/10.1016/j.joi.2025.101759
The paper
Ebadi, A., Auger, A., & Gauthier, Y. (2026). WISDOM: An AI-powered framework for emerging research detection using weak signal analysis and advanced topic modelling. Journal of Informetrics. https://doi.org/10.1016/j.joi.2025.101759

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.