Navigating the AI Timeline: From 1995 to Today

Vincenza Carchiolo

and Michele Malgeri

DIEEI - Universit

a degli Studi di Catania, Italy

Keywords:

Artiﬁcial Intelligence, Review, Data Analysis, NLP.

Abstract:

In recent years, the exponential growth of Artiﬁcial Intelligence (AI) has transcended disciplinary boundaries,

expanding into diverse ﬁelds beyond computer science. This study analyzes AI’s distribution across disciplines

using a large dataset of scientiﬁc publications. Contrary to expectations, substantial AI research extends into

medicine, engineering, social sciences, and humanities. This interdisciplinary presence heralds new possibil-

ities for collaborative innovation to tackle contemporary challenges. The analysis identiﬁes emerging trends,

contributing to a deeper understanding of AI’s evolving role in society.

1 INTRODUCTION

Artiﬁcial intelligence (AI) has experienced an expo-

nential rise in recent years, both in terms of popularity

and societal impact. This is evident from the increase

in scientiﬁc output in the ﬁeld, with a growing number

of publications, conferences, and initiatives dedicated

to AI.

However, within this rapidly evolving landscape,

an intriguing inquiry arises: how is AI research dis-

tributed across different disciplinary ﬁelds? Is it pri-

marily concentrated within the computer science do-

main, or has it diffused into other application areas?

This analysis aims to examine scientiﬁc output in

the ﬁeld of AI with the goal of determining its distri-

bution across different disciplinary sectors. Through

the analysis of a large dataset of scientiﬁc publica-

tions, we will show that, contrary to what one might

assume, a substantial share of AI research does not

take place within the computer science ﬁeld. Rather,

it is found in a variety of application areas, includ-

ing medicine, engineering, social sciences, and even

humanities.

This ﬁnding has important implications for our

understanding of AI and its potential impact on so-

ciety. It demonstrates that AI is no longer conﬁned to

a narrow technical domain but is permeating a wide

range of disciplines and application areas. This opens

up new and exciting possibilities for interdisciplinary

collaboration and for the development of innovative

https://orcid.org/0000-0002-1671-840X

https://orcid.org/0000-0002-9279-3129

AI-based solutions that can address the most pressing

challenges of our time.

Furthermore, this analysis will allow us to iden-

tify emerging trends in AI research and to better un-

derstand the future directions of this rapidly evolv-

ing ﬁeld. Ultimately, this research will contribute to a

more comprehensive understanding of AI and its role

in society.

While a comprehensive analysis of the entire lit-

erature encompassing AI is an insurmountable task,

we have chosen to utilize the Scopus database as the

foundation for our investigation. Scopus, managed

by Elsevier, stands as one of the world’s most com-

prehensive repositories of bibliographic and abstract

data within the scientiﬁc realm (Elsevier B.V., b) (El-

sevier B.V., a). Its extensive disciplinary coverage,

encompassing a broad spectrum of scientiﬁc ﬁelds,

including natural sciences, social sciences, medical

sciences, and engineering, aligns perfectly with our

objective of understanding AI’s impact across diverse

sectors. Furthermore, Scopus provides an array of an-

alytical tools that facilitate effective exploration of the

database in alignment with our research goals.

Scopus is therefore generalist in that it covers a

wide range of scientiﬁc disciplines, but there are some

more specialized or niche research areas that may not

be fully represented in its database. Scopus allows

us to query its database through various APIs. Sco-

pus contains approximately 82 million articles, and

searches can be conducted across different ﬁelds of

research.

In the literature, there are many examples of the

use of the Scopus database. For instance, in (H

ancean

Carchiolo, V. and Malgeri, M.

Navigating the AI Timeline: From 1995 to Today.

DOI: 10.5220/0012856700003756

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 13th International Conference on Data Science, Technology and Applications (DATA 2024), pages 577-584

ISBN: 978-989-758-707-8; ISSN: 2184-285X

577

et al., 2021) (De Stefano et al., 2011) (Carchiolo et al.,

2022a) some examples of generating co-authorship

networks are presented with the idea of represent-

ing the collaborations of the authors. In other cases

(Carchiolo et al., 2022b), (Bordons et al., 2015) (Car-

chiolo et al., 2023), the data extracted from Scopus

have been used to analyze the importance of certain

researchers or their performance in terms of speciﬁc

indices.

Our analysis, grounded in a robust dataset of sci-

entiﬁc publications, maps the distribution of AI re-

search across these diverse disciplines. The data

speaks volumes: AI is not just a tool but a collabo-

rator, opening doors to new realms of knowledge and

understanding.

Section 2 introduces how the dataset was con-

structed, while Section 3 presents the temporal anal-

ysis and thematic distribution of the publications, and

some results are discussed in detail. Section 4 pro-

vides an overview of some of the most cited articles.

We ﬁnally consider further works and concluding re-

marks in Section 5.

2 DATASET

The study of scientiﬁc output in the ﬁeld of Arti-

ﬁcial Intelligence (AI) holds immense signiﬁcance,

given its prominence as a contemporary research fo-

cus that extends beyond the conﬁnes of computer sci-

ence. Researchers across diverse disciplines, even

those seemingly distant from computer science, are

increasingly exploring the potential of AI as a vi-

able solution for their respective ﬁelds of study. For

this study, to capture the subset of articles focusing

on Artiﬁcial Intelligence, a query against the Sco-

pus database was performed using the keyword ﬁeld.

Instead of using only the keyword ”Artiﬁcial Intel-

ligence,” 18 different keywords were chosen, the ta-

ble 1 lists the 18 keywords used and the number of

articles selected for each. With this keyword selec-

tion, a total of 2, 156, 387 articles were selected, of

which 2, 081, 397, about 96.5% were written in En-

glish. This percentage remains relatively constant

over the years; for example, in 2023, it is approxi-

mately 97%.

The analysis of these documents, totaling

2,156,387, reveals that a very small fraction (a few

thousand) are incorrectly categorized and cover a

topic do not related to artiﬁcial intelligence that, typ-

ically, pertain to keywords such as ”Pattern Recogni-

tion,” ”Reinforcement Learning,” ”Optimization Al-

gorithms,” and ”Data Analysis”. However, given the

small percentage, they cannot bias our analysis. Some

Table 1: Keywords in Artiﬁcial Intelligence.

Keyword Document number

Artiﬁcial Intelligence 460 755

Machine Learning 455 294

Deep Learning 348 589

Data Analysis 272 258

Pattern Recognition 227 417

Convolutional Neural Networks 182 998

Computer Vision 180 619

Artiﬁcial Neural Networks 154 046

Natural Language Processing 108 584

Optimization Algorithms 80 415

Reinforcement Learning 77 935

Expert Systems 56 364

Supervised Learning 45 794

Recurrent Neural Networks 43 456

Machine Learning Algorithms 17 022

Unsupervised Learning 14 376

Artiﬁcial Intelligence Applications 950

Cognitive Robotics 796

of the documents are quite old, dating back further

than expected (see Figure 1.(a)) It should not be sur-

prising to ﬁnd articles from the early 1950s, as the

term ”AI” was coined by John McCarthy in 1955 dur-

ing a conference at Dartmouth College. McCarthy

and other scholars laid the groundwork for a new

research ﬁeld aimed at developing machines capa-

ble of learning, reasoning, and problem-solving au-

tonomously. However, traces of AI-related concepts

can be found even before 1955 in the writings of Alan

Turing, Marvin Minsky, who founded the MIT Arti-

ﬁcial Intelligence Laboratory in 1951, one of the pio-

neering research centers in AI, and Arthur Samuel,

who in 1959 developed the ”Gameplay” program,

considered one of the earliest examples of artiﬁcial

intelligence applied to a game, speciﬁcally checkers.

Nevertheless, in recent decades, due to technologi-

cal advancements, the availability of large amounts

of data, and improvements in algorithms, AI has ex-

perienced a resurgence and has begun to inﬂuence an

increasing number of societal sectors.

The initial analyses presented in this paper aimed

to delineate the sectors to which the publications

could be attributed. Scopus organizes its database by

assigning a ”Subject Area” to each publication based

on the publication venue. As depicted in Figure 1.(b),

it is evident that slightly over 50% of the publications

are categorized under Computer Science, a proportion

that has varied between 45% and 60% over the years.

This observation underscores the dominant presence

of Computer Science within the dataset.

Furthermore, we conducted an analysis of the

”Subject Areas” in which Scopus classiﬁes docu-

ments, the table 2 lists all the ”Subject Areas” present

in Scopus, along with the number of publications se-

lected by us attributed to each area. In our case, the

documents are divided into Computer Science and

Engineering for more than 80% of the cases. More-

over, it can be appreciated that the most relevant ap-

plication ﬁelds in table 2

DATA 2024 - 13th International Conference on Data Science, Technology and Applications

578

(a) Growth in publications over time.

(b) CS publications relative to all publications.

Figure 1: Total and Computer Science document analysis.

Table 2: Subject Area and Document Number.

Ref. Subject Area Document Number

CS Computer Science 1 213 347

ENG Engineering 720 705

MAT Mathematics 413 249

MED Medicine 335 620

PA Physics and Astronomy 194 224

BGMB Biochemistry Genetics and Molecular Biology 145 330

MS Materials Science 123 975

DS Decision Sciences 114 903

SS Social Sciences 110 184

NS Neuroscience 87 705

EN Energy 82 665

EPS Earth and Planetary Sciences 80 569

ES Environmental Science 76 425

CHM Chemistry 61 793

ABS Agricultural and Biological Sciences 49 042

AH Arts and Humanities 45 293

CHE Chemical Engineering 43 689

PSY Psychology 41 260

BMA Business Management and Accounting 40 372

HP Health Professions 32 916

MUL Multidisciplinary 31 149

PTP Pharmacology Toxicology and Pharmaceutics 30 093

IM Immunology and Microbiology 26 421

NUR Nursing 14 553

EEF Economics Econometrics and Finance 12 689

VET Veterinary 3 607

DNT Dentistry 3 110

Finally, we investigated the afﬁliations of the doc-

uments, which originate from various countries; how-

ever, the predominant afﬁliations are located in China

and the USA. Among the top 60 afﬁliations, signiﬁ-

cant presence is observed from France, Canada, Sin-

gapore, the UK, Bangladesh, Japan, Germany, India,

Switzerland, Australia, Hong Kong, Italy, Belgium,

and Brazil.

3 AI PUBLICATIONS TRENDS

The choice to focus the analysis of publications start-

ing from 1995 in the present study is motivated by

several crucial considerations to ensure the validity

and relevance of our conclusions within the current

research context. Firstly, it is important to note that

the data obtained from Scopus for the years 1950-

1995 are signiﬁcantly limited in our ﬁeld of investi-

gation. This limited availability of information can be

attributed to various factors, including technological

and methodological constraints of the time, as well

as the reduced use of bibliometric and citation sys-

tems. Secondly, we consider the dynamic evolution

of our research ﬁeld over the years. There has been a

growing consensus regarding the increasing relevance

of recent publications in reﬂecting the current state of

knowledge and developments in the ﬁeld. Therefore,

recent (in our case, around 30 years) publications pro-

vide a more precise and updated picture of the current

scientiﬁc landscape. Additionally, it is worth men-

tioning that signiﬁcant trends and advancements in

our ﬁeld have occurred primarily during the last few

decades, making publications after 1995 particularly

relevant for our analysis. Finally, our decision is also

motivated by the need to maintain consistency with

contemporary research practices and to provide re-

sults that are fully aligned with the current scientiﬁc

context, thus offering signiﬁcant contributions to aca-

demic literature.

3.1 Number of Publications per Year

As a ﬁrst analysis, a cumulative analysis of publica-

tions over the years was chosen, which naturally re-

veals a consistent growth in the number of scientiﬁc

articles in the ﬁeld of artiﬁcial intelligence. Examin-

ing the temporal growth of publications, a clear pos-

itive trend emerges over the years, indicative of an

increasingly intense and continuous research activity.

This approach also allows for the evaluation of long-

term trends, revealing a steady increase in the number

of publications year after year. Moreover, such anal-

ysis enables the identiﬁcation of signiﬁcant turning

points in the growth rate, suggesting moments when

research activity has undergone important changes or

developments. These turning points can be indicative

of innovations in the ﬁeld, changes in research trends,

or signiﬁcant events that have inﬂuenced academic in-

terest. Figure 2 shows the cumulative trend over the

years from 1995 to 2023. It can be readily observed

that around 2010, there is a knee in the curve, indi-

cating a sudden growth in the number of publications.

This year is widely recognized as a pivotal year in the

Navigating the AI Timeline: From 1995 to Today

579

ﬁeld of artiﬁcial intelligence (AI) for a number of cru-

cial reasons, such as, there were signiﬁcant advance-

ments in deep learning.

Figure 2: Cumulative trend of publications from 1995 to

2023.

3.2 Subject Areas by Year

In this section, we present the evolution of Subject

Area from 1995 to 2023. As previously indicated,

Scopus assigns a Subject Area to each publication

based on its publication venue and the topic it cov-

ers within that source. Based on this analysis, it can

be observed (see Figure 3) that the majority of publi-

cations fall within the ﬁelds of Computer Science or

Engineering, with Computer Science being the pre-

dominant subject area.

As illustrated in Figure 4, Computer Science and

Engineering emerge as the two primary Subject Ar-

eas, exhibiting a parallel growth pattern. This pat-

tern suggests a close relationship between these two

ﬁelds, with advancements in Computer Science often

preceding similar developments in Engineering.

Figure 4: Trend of Computer Science & Engineering Sub-

ject Area.

To delve into the trends beyond the two dominant

Subject Areas of Computer Science and Engineer-

ing, Figure 5 presents an overview of the remaining

Subject Areas. This visualization aims to highlight

the diverse patterns and trajectories observed across

a broader range of disciplines. As evident, the Sub-

ject Areas beyond Computer Science and Engineer-

ing exhibit a more nuanced growth pattern. While

some Subject Areas, display a more gradual or even

plateauing growth trajectory. These observations un-

derscore the multifaceted nature of AI research and its

impact across a diverse spectrum of academic ﬁelds.

To further elucidate the distinct growth patterns

observed, we present the trend of subject areas group-

ing them in smaller sets in order to enhance the

speciﬁc behaviours that are hidden in global ﬁg-

ure, therefore ﬁgure 6 presents the trends for the

Subject Areas immediately following the two dom-

inant ﬁelds (Mathematics Physics and Astronomy,

Medicine, Biochemistry, Genetics, and Molecular Bi-

ology) and ﬁgure 7 delves into the trends of the next

four prominent Subject Areas, Materials Science, De-

cision Sciences, Social Sciences, and Neuroscience.

As depicted in Figure 6, Mathematics displays

a growth pattern akin to the top two Subject Ar-

eas. Conversely, Physics and Astronomy, as well

as Biochemistry, Genetics, and Molecular Biology,

exhibit a delayed onset of growth. Furthermore,

Medicine initiates its upward trajectory with a slope

comparable to Computer Science around 2013, while

Physics and Astronomy and Biochemistry, Genetics,

and Molecular Biology follow suit around 2015. In-

triguingly, both Medicine and Biochemistry, Genet-

ics, and Molecular Biology encounter a deceleration

in their growth rates from 2021 onward.

Figure 7 delves into the trends of the next four

prominent Subject Areas: Materials Science, Deci-

sion Sciences, Social Sciences, and Neuroscience.

This visualization sheds light on the evolving dynam-

ics of these emerging ﬁelds within the AI landscape.

Interestingly, Decision Sciences stands out as an ex-

ceptional case, exhibiting a surge in growth starting

from 2018. This upward trajectory suggests a grow-

ing recognition of the potential of AI in decision-

making processes across various domains. In con-

trast, Neuroscience, which initially followed a simi-

lar growth pattern to the other Subject Areas, experi-

enced a stagnation in its growth from 2018 onwards.

This observation may warrant further investigation to

understand the underlying factors contributing to this

trend.

3.3 Geographical Landscape of AI

Research

In the dynamic landscape of Artiﬁcial Intelligence

(AI), analyzing the number of publications by country

per year between 1995 and 2023 serves as a valuable

lens through which to observe global research trends

in this revolutionary ﬁeld. Understanding the geo-

graphical distribution of scientiﬁc output in AI goes

beyond mere mapping. It’s a deep exploration that

enables us to assess the impact of AI in different con-

DATA 2024 - 13th International Conference on Data Science, Technology and Applications

580

Figure 3: Trend of Subject Area from 1995 to 2023.

Figure 5: Trend of Subject Areas excluding dominant ones

from 1995-2023.

Figure 6: Trend of Mathematics Physics and Astronomy,

Medicine, Biochemistry, Genetics, and Molecular Biology.

Figure 7: Trend of Materials Science, Decision Sciences,

Social Sciences, and Neuroscience Subject.

texts, uncovering how this technology is shaping and

inﬂuencing various societies and cultures worldwide.

Through this analysis, centers of excellence in AI re-

search become clear, acting as strategic benchmarks

for international collaboration and attracting top tal-

ent in the ﬁeld. The data gathered serves as a compass

to guide national and international policies and strate-

gic decisions, allowing for effective resource alloca-

tion and the development of targeted strategies to pro-

mote AI adoption in every context. Furthermore, an-

alyzing the number of publications by country proves

to be a powerful tool for monitoring the progress of AI

research over time. It offers a tangible assessment of

this technology’s impact on society and allows for the

identiﬁcation of new trends and emerging research ar-

eas that will shape the future of AI. In doing so, for the

sake of simplicity and efﬁciency, we narrowed down

the analysis to the top 24 countries. This approach

streamlines the data collection and analysis process,

optimizing computational resources and expediting

the study’s completion. While this decision excludes

other countries, it does not diminish the signiﬁcance

of their AI research endeavors. Complementary stud-

ies could be conducted to delve into the research dy-

namics of nations with lower levels of AI scientiﬁc

output. Ultimately, the choice to focus on the 24 top

AI-publishing countries represents a strategic balance

between comprehensiveness and feasibility, yielding

a focused and in-depth perspective on research trends

and patterns within the global epicenters of this trans-

formative ﬁeld.

Figure 8 illustrates the percentage distribution of

publications across the years under consideration, re-

vealing that China and the USA collectively account

for nearly 50% of the publications. Speciﬁcally,

the USA’s contribution peaks at 35% in 1995, while

China reaches the same percentage in 2023.

Figure 8: Percentage of Publications by Country.

To better appreciate the trend for each of the coun-

tries, a cumulative trend analysis was conducted, dis-

playing subsets of selected states for similarities in

subsequent ﬁgures. Figure 9a shows the trend for the

Navigating the AI Timeline: From 1995 to Today

581

(a) China and USA (b) India and UK

Figure 9: Cumulative growth of publications.

USA and China, which are the states with the highest

number of publications. Figure 9b shows the trends

for India and the UK, while ﬁgure 9c shows some of

the European states. Finally, ﬁgure 9d displays the

trends for the other states.

3.4 Keywords Analysis

The analysis of keywords in research articles on Arti-

ﬁcial Intelligence (AI) from 1995 to 2023 represents a

wealth of information for understanding global trends

in this constantly evolving ﬁeld. By examining the

most frequent keywords in articles, we can identify

the most popular themes and research areas within the

ﬁeld of AI during a speciﬁc period. This allows us

to track the evolution of research interests over time

and identify emerging trends. Keywords, that are usu-

ally inserted by authors, can also provide insights into

the methodological approach used by researchers. For

example, the presence of statistical terms may indi-

cate the use of data-driven models, while keywords

related to machine learning may suggest the use of

machine learning algorithms. By examining the most

frequent keywords in strategic years, we can iden-

tify dominant themes, prevalent methodological ap-

proaches, and the impact of AI across various spheres

of society. This approach offers several advantages,

including the ability to highlight long-term trends, re-

duce complexity, and optimize resources.

To do this, the years 2010, 2015, 2020, 2021,

2022, and 2023 were selected, all following the rapid

growth of AI on the international scene. For each

year, the top 160 most recurring keywords were ex-

tracted. Before conducting any analysis, the list of ex-

tracted keywords was processed using some Natural

Language Processing techniques (NLP) (Cambria and

White, 2014) to eliminate keywords that lexically ex-

press the same concept. To perform the ﬁltering from

this list of keywords after transforming all strings to

lowercase, the following steps were carried out:

1. Synonym Identiﬁcation: WordNet (Pedersen

et al., 2004), a lexical database of English words,

was used to identify synonyms for each keyword.

WordNet provides relationships between words,

including synonyms, antonyms, hyponyms (more

speciﬁc words), and hypernyms (more general

words).

2. Acronym Expansion: Using an acronym dictio-

nary, the acronyms and abbreviations found in the

list of keywords were expanded.

3. Similarity Calculation: For each keyword, co-

sine similarity with its potential synonyms or

acronym expansions was calculated (Wang and

Dong, 2020). A removal threshold of 0.8 was cho-

sen.

These operations allowed for the calculation of the

overall occurrence count of each keyword in a given

year, shown in Table 3, thus enabling the determina-

tion of the top 10 most recurring keywords for each

of the years.

To comprehend the trends of interest in speciﬁc ar-

eas of artiﬁcial intelligence, we elected to investigate

the top 15 keywords appearing across approximately

DATA 2024 - 13th International Conference on Data Science, Technology and Applications

582

Table 3: First 10 keyword for year, where AI, ML, CNN, NLP stands for Artiﬁcial Intelligence, Machine Learning, Convolu-

tional neural networks, Natural Language Processing respectively.

2015 2010 2020 2021 2022 2023

AI AI ML Human ML ML

Genetic Algorithm Human Decision Tree ML Human Deep Learning

Algorithm Algorithm CNN CNN CNN Human

Article Pattern Recognition Learning Systems Deep Learning Deep Learning CNN

Pattern Recognition Article Genetics Learning systems Article Learning systems

Computer Science Learning Systems Article AI Female/Male Article

NLP Female/Male AI Data Analysis Magnetic resonance

imaging

Feature Extraction Computer Vision Feature Extraction Neural Networks AI Neural Networks

Artiﬁcial Neural Network ML Human Controlled study Neural Network Female/Male

Controllers Multilayer Neural

Networks

Classiﬁcation Female/Male Classiﬁcation Features Extraction

2 million publications on the subject, selected inde-

pendently of the year (this procedure was conducted

after ﬁltering). Table 4 presents the top keywords and

their respective ranks in each year, while Figure 10

depicts the ranking trend of the top 15 keywords. To

facilitate the reader, the ﬁgure 10 is divided into two

parts, in each of which the change of keywords rank-

ing over the years can be observed. It can be noted

that the keywords ”Controlled Study” and ”Human,”

absent in the early years, have achieved high rankings

(it is worth noting that a low position value indicates a

high ranking), as well as ”Neural Networks” and ”Re-

inforcement Learning”. On the contrary, some key-

words as ”Artiﬁcial intelligence” and ”Artiﬁcial Neu-

ral Network” have lost positions in the ranking.

Table 4: Main keyword and their ranking (ﬁrst 160) in each

year.

2010

2015

2020

2021

2022

2023

Adult 21 27 14 14 21 22

Algorithm 3 3 18 15 16 18

Article 4 5 6 5 5 6

Artiﬁcial Intelligence 1 1 7 7 8 7

Artiﬁcial Neural Network 9 15 32 29 27 23

Classiﬁcation 64 11 10 12 10 12

Computer Vision 17 8 25 16 19 20

Controlled Study 111 19 19 10 13 16

Convolutional Neural Network - 47 3 3 3 4

Data Analysis 41 14 24 8 12 19

Deep Learning - 79 51 4 4 2

Female/male 18 7 11 11 6 9

Forecasting - 30 121 20 17 15

Human 122 2 9 1 2 3

Learning Systems 16 6 4 6 7 5

Machine Learning 23 9 1 2 1 1

Natural Language Processing

Systems

8 - 21 23 18 13

Neural Networks 32 29 23 9 9 8

Pattern Recognition 5 4 - - - -

Reinforcement Learning6 66 120 69 20 15 11

4 HIGHLY CITED PAPER

In this brief section, we aim to demonstrate that nu-

merous articles of great interest have been published

during the period under consideration.

In particular, we would like to highlight some of

Figure 10: Trend of the top 15 keywords.

the most cited articles by listing them in chronological

order.

Among the 200 most cited articles in our dataset,

the most cited article from 1995 (the ﬁrst year of our

analysis) is (Cortes and Vapnik, 1995). The main top-

ics addressed in this article are the use of neural net-

works, particularly in the ﬁeld of pattern recognition,

which is one of the most common ﬁelds of application

and on which researchers have focused their efforts

in the ﬁeld of AI. The most recent of the 200 most

cited is (Jumper et al., 2021). The article presents a

protein structure prediction approach based on deep

learning and this shows how techniques have evolved

and how the ﬁelds of application of greatest interest

have changed

Finally, the two most cited articles with over

100,000 citations are (He et al., 2016) and (Livak and

Navigating the AI Timeline: From 1995 to Today

583

Schmittgen, 2001). In (He et al., 2016) aims to ad-

dress the challenges of training very deep neural net-

works and it propose a new framework called resid-

ual learning which makes it easier to train deeper net-

works compared to previous methods. (Livak and

Schmittgen, 2001) falls within the ﬁeld of molec-

ular biology, with a speciﬁc focus on genetics and

deals with the analysis of data obtained from real-time

quantitative PCR (qPCR) experiments. These two ar-

ticles further demonstrate that the trend in ﬁelds of

application is increasingly shifting from computer vi-

sion to applications in the ﬁeld of genetics.

5 CONCLUSION

The authors emphasize the value of examining re-

views from previous years to gain a comprehensive

understanding of the AI landscape. This longitudi-

nal approach can reveal trends, patterns, and semi-

nal contributions that might be missed by focusing

solely on recent publications. The analysis has high-

lighted some peculiarities of publications in the AI

ﬁeld. Nevertheless, this analysis is still in its early

stages. This could involve identifying emerging key-

words, tracking changes in keyword usage over time,

and exploring the relationships between different key-

words.

ACKNOWLEDGEMENTS

The work is partially supported by UDMA project,

CUP: G69J18001040007.

REFERENCES

Bordons, M., Aparicio, J., Gonz

alez-Albo, B., and D

ıaz-

Faes, A. A. (2015). The relationship between the re-

search performance of scientists and their position in

co-authorship networks in three ﬁelds. Journal of In-

formetrics, 9(1):135–144.

Cambria, E. and White, B. (2014). Jumping nlp curves: A

review of natural language processing research. IEEE

Computational intelligence magazine, 9(2):48–57.

Carchiolo, V., Grassia, M., Malgeri, M., and Mangioni, G.

(2022a). Co-authorship networks analysis to discover

collaboration patterns among italian researchers. Fu-

ture Internet, 14(6).

Carchiolo, V., Grassia, M., Malgeri, M., and Mangioni, G.

(2022b). Network topology to predict bibliometrics

indices: A case study. In Pardede, E., Delir Haghighi,

P., Khalil, I., and Kotsis, G., editors, Information Inte-

gration and Web Intelligence, pages 166–180, Cham.

Springer Nature Switzerland.

Carchiolo, V., Grassia, M., Malgeri, M., and Mangioni, G.

(2023). Correlation between researchers’ centrality

and h-index: A case study. In Braubach, L., Jan-

der, K., and B

adic

a, C., editors, Intelligent Distributed

Computing XV, pages 133–143, Cham. Springer Inter-

national Publishing.

Cortes, C. and Vapnik, V. (1995). Support-vector networks.

Machine Learning, 20(3):273–297.

De Stefano, D., Giordano, G., and Vitale, M. (2011). Issues

in the analysis of co-authorship networks. Quality &

Quantity, 45:1091–1107.

Elsevier B.V. Elsevier Developer - Academic Research.

https://dev.elsevier.com/api\ service\ agreement.

html. Accessed: July, 2021.

Elsevier B.V. Elsevier Developer - API Service Agree-

ment. https://dev.elsevier.com/academic\ research\

scopus.html. Accessed: Juy, 2021.

He, K., Zhang, X., Ren, S., and Sun, J. (2016). Deep resid-

ual learning for image recognition. In 2016 IEEE Con-

ference on Computer Vision and Pattern Recognition

(CVPR), pages 770–778.

ancean, M., Perc, M., and Lerner, J. (2021). The coau-

thorship networks of the most productive european re-

searchers. Scientometrics, 126:201–224.

Jumper, J., Evans, R., and Pritzel, A. e. a. (2021). Highly

accurate protein structure prediction with alphafold.

Nature, 596(7873):583–589.

Livak, K. J. and Schmittgen, T. D. (2001). Analysis of

relative gene expression data using real-time quantita-

tive pcr and the 2(-delta delta c(t)) method. Methods,

25(4):402–408.

Pedersen, T., Patwardhan, S., Michelizzi, J., et al. (2004).

Wordnet:: Similarity-measuring the relatedness of

concepts. In AAAI, volume 4, pages 25–29.

Wang, J. and Dong, Y. (2020). Measurement of text simi-

larity: a survey. Information, 11(9):421.

DATA 2024 - 13th International Conference on Data Science, Technology and Applications

584