Textual Data Science for Logistics and Supply Chain Management

Horst Treiblmaier,Patrick Mair

doi:10.3390/logistics5030056

Abstract

Background: Researchers in logistics and supply chain management apply a multitude of methods. So far, however, the potential of textual data science has not been fully exploited to automatically analyze large chunks of textual data and to extract relevant insights. Methods: In this paper, we use data from 19 qualitative interviews with supply chain experts and illustrate how the following methods can be applied: (1) word clouds, (2) sentiment analysis, (3) topic models, (4) correspondence analysis, and (5) multidimensional scaling. Results: Word clouds show the most frequent words in a body of text. Sentiment analysis can be used to calculate polarity scores based on the sentiments that the respondents had in their interviews. Topic models cluster the texts based on dominating topics. Correspondence analysis shows the associations between the words being used and the respective managers. Multidimensional scaling allows researchers to visualize the similarities between the interviews and yields clusters of managers, which can also be used to highlight differences between companies. Conclusions: Textual data science can be applied to mine qualitative data and to extract novel knowledge. This can yield interesting insights that can supplement existing research approaches in logistics and supply chain research.

Highlights

To date, automated analyses of large chunks of textual data, usually labeled as text analysis or text mining, have been used only infrequently in logistics and supply chain management
Several notable exceptions, which illustrate the wide range of possible uses, include their application for literature reviews [1,2,3], conference abstracts and summaries [4], demand planning [5], supply chain risk management [6], identification of key challenges and prospects for logistics services [7], and the creation of insights regarding drivers of change [8]
We discuss the results from applying five different textual science methods: (1) word clouds, (2) sentiment analysis, (3) topic modeling, (4) correspondence analysis, and (5) multidimensional scaling [23]. All of these techniques were applied on the same corpora, and the statistical software R was used throughout the study

Summary

Introduction

To date, automated analyses of large chunks of textual data, usually labeled as text analysis or text mining, have been used only infrequently in logistics and supply chain management. Alfaro et al [10] illustrate how to detect opinion trends via the analysis of weblogs Building on the latter ideas, we illustrate how to gain knowledge from textual data by using an illustrative case from supply chain forecasting. Methods: In this paper, we use data from 19 qualitative interviews with supply chain experts and illustrate how the following methods can be applied: (1) word clouds, (2) sentiment analysis, (3) topic models, (4) correspondence analysis, and (5) multidimensional scaling. Conclusions: Textual data science can be applied to mine qualitative data and to extract novel knowledge. This can yield interesting insights that can supplement existing research approaches in logistics and supply chain research

Objectives

Methods

Results