Quantitative Text Analysis

Kristoffer L. Nielbo; F.B. Karsdorp; Melvin Wevers; Alie Lassche; Rebekah Baglini; Mike Kestemont; Nina Tahmasebi

doi:10.1038/s43586-024-00302-w

Quantitative Text Analysis

Kristoffer L. Nielbo, F.B. Karsdorp, Melvin Wevers, Alie Lassche, Rebekah Baglini, Mike Kestemont, Nina Tahmasebi

Research output: Contribution to journal/periodical › Article › Scientific › peer-review

2 Citations (Scopus)

Abstract

Text analysis has undergone substantial evolution since its inception, moving from manual qualitative assessments to sophisticated quantitative and computational methods. Beginning in the late twentieth century, a surge in the utilization of computational techniques reshaped the landscape of text analysis, catalysed by advances in computational power and database technologies. Researchers in various fields, from history to medicine, are now using quantitative methodologies, particularly machine learning, to extract insights from massive textual data sets. This transformation can be described in three discernible methodological stages: feature-based models, representation learning models and generative models. Although sequential, these stages are complementary, each addressing analytical challenges in the text analysis. The progression from feature-based models that require manual feature engineering to contemporary generative models, such as GPT-4 and Llama2, signifies a change in the workflow, scale and computational infrastructure of the quantitative text analysis. This Primer presents a detailed introduction of some of these developments, offering insights into the methods, principles and applications pertinent to researchers embarking on the quantitative text analysis, especially within the field of machine learning.

Original language	English
Pages (from-to)	1-16
Journal	Nature Reviews Methods Primers
Volume	4
Issue number	25
DOIs	https://doi.org/10.1038/s43586-024-00302-w
Publication status	Published - 2024

Access to Document

10.1038/s43586-024-00302-w

Cite this

@article{5cf0b019b1474e97a9f58c4237f50652,

title = "Quantitative Text Analysis",

abstract = "Text analysis has undergone substantial evolution since its inception, moving from manual qualitative assessments to sophisticated quantitative and computational methods. Beginning in the late twentieth century, a surge in the utilization of computational techniques reshaped the landscape of text analysis, catalysed by advances in computational power and database technologies. Researchers in various fields, from history to medicine, are now using quantitative methodologies, particularly machine learning, to extract insights from massive textual data sets. This transformation can be described in three discernible methodological stages: feature-based models, representation learning models and generative models. Although sequential, these stages are complementary, each addressing analytical challenges in the text analysis. The progression from feature-based models that require manual feature engineering to contemporary generative models, such as GPT-4 and Llama2, signifies a change in the workflow, scale and computational infrastructure of the quantitative text analysis. This Primer presents a detailed introduction of some of these developments, offering insights into the methods, principles and applications pertinent to researchers embarking on the quantitative text analysis, especially within the field of machine learning.",

author = "Nielbo, {Kristoffer L.} and F.B. Karsdorp and Melvin Wevers and Alie Lassche and Rebekah Baglini and Mike Kestemont and Nina Tahmasebi",

year = "2024",

doi = "10.1038/s43586-024-00302-w",

language = "English",

volume = "4",

pages = "1--16",

number = "25",

}

TY - JOUR

T1 - Quantitative Text Analysis

AU - Nielbo, Kristoffer L.

AU - Karsdorp, F.B.

AU - Wevers, Melvin

AU - Lassche, Alie

AU - Baglini, Rebekah

AU - Kestemont, Mike

AU - Tahmasebi, Nina

PY - 2024

Y1 - 2024

N2 - Text analysis has undergone substantial evolution since its inception, moving from manual qualitative assessments to sophisticated quantitative and computational methods. Beginning in the late twentieth century, a surge in the utilization of computational techniques reshaped the landscape of text analysis, catalysed by advances in computational power and database technologies. Researchers in various fields, from history to medicine, are now using quantitative methodologies, particularly machine learning, to extract insights from massive textual data sets. This transformation can be described in three discernible methodological stages: feature-based models, representation learning models and generative models. Although sequential, these stages are complementary, each addressing analytical challenges in the text analysis. The progression from feature-based models that require manual feature engineering to contemporary generative models, such as GPT-4 and Llama2, signifies a change in the workflow, scale and computational infrastructure of the quantitative text analysis. This Primer presents a detailed introduction of some of these developments, offering insights into the methods, principles and applications pertinent to researchers embarking on the quantitative text analysis, especially within the field of machine learning.

AB - Text analysis has undergone substantial evolution since its inception, moving from manual qualitative assessments to sophisticated quantitative and computational methods. Beginning in the late twentieth century, a surge in the utilization of computational techniques reshaped the landscape of text analysis, catalysed by advances in computational power and database technologies. Researchers in various fields, from history to medicine, are now using quantitative methodologies, particularly machine learning, to extract insights from massive textual data sets. This transformation can be described in three discernible methodological stages: feature-based models, representation learning models and generative models. Although sequential, these stages are complementary, each addressing analytical challenges in the text analysis. The progression from feature-based models that require manual feature engineering to contemporary generative models, such as GPT-4 and Llama2, signifies a change in the workflow, scale and computational infrastructure of the quantitative text analysis. This Primer presents a detailed introduction of some of these developments, offering insights into the methods, principles and applications pertinent to researchers embarking on the quantitative text analysis, especially within the field of machine learning.

U2 - 10.1038/s43586-024-00302-w

DO - 10.1038/s43586-024-00302-w

M3 - Article

SN - 2662-8449

VL - 4

SP - 1

EP - 16

JO - Nature Reviews Methods Primers

JF - Nature Reviews Methods Primers

IS - 25

ER -

Quantitative Text Analysis

Abstract

Access to Document

Fingerprint

Cite this