Text Mining for Data Collection in Production Economics Research
The research of Production Economics has a focus on various allocation problems, e.g. on the optimal use of productive resources in manufacturing and service industries or other production organisations. The contribution of Production Economics studies is often reflected by theory development and applications for emerging problems in industry. Therefore, in the past decades the research is expected to be strongly supported and encouraged by changing environment in industry and society. The research methodologies in Production Economics include mathematical modelling, optimization, decision analysis, simulation, empirical studies and conceptual modelling, among others.
There are abundant research outputs from Production Economics, and from its closely relevant fields such as Management Science and Engineering, and Supply Chain and Operations Management. Even though the research fields have been well developed in the past decades, there are also criticisms indicating that many models are seldom used by managers in practice. Among several reasons, this is due to the difficulties of obtaining data in the past to estimate empirically the model parameters (Little, 2004). In order to bridge the link between theory and practice, research has stressed the importance of data (Bass, 2004), and recent decision models have a trend shifting from rational assumptions to behaviour description (Kundu, et al., 2015). This is then encouraged by the rapid availability of data in recent years.
This presentation therefore aims to provide an overview how to obtain data through text mining for the purpose of supporting the research in Production Economics. After introducing the main processes, tasks and tools of text mining, we present three undergoing studies. In the first example (Gao and Tang, 2017), text mining is used for literature review in supply chain risk management. The intellectual structure of the field is analysed by CiteSpace. Important terms, keywords, cooperation relation, cited references are structed and investigated. Furthermore, in order to extract literature contents, we propose a framework by using text mining approaches to identify the topics in the field. Topic words extracted by Latent Dirichlet allocation technology reveal some interesting variation of topics in the field. The study not only updates the review in supply chain risk management, but also illustrates the possibility to objectively review literature with the support of text mining technology, i.e. our newly developed framework. This framework can also be easily applied to other research fields for providing a general overview of the existing body of literature.
In the second example (Gao et al. 2018), by text mining of online reviews, we extract comparative relations and construct comparison relation networks for competitor identification in a restaurant industry. The first comparative network is established to analyse market structure. The second comparative network identifies top competitors using a competitive index and dissimilarity index. The third network indicates strengths and weaknesses of a restaurant through aspects-comparative relation mining. The above networks are constructed by using the data from mining the online reviews. Finally, the market environment is analysed. Experiments are conducted to validate and verify the proposed method. The research illustrates the effectiveness of competitiveness analysis technology, which can facilitate identifying top competitors and evaluating the market environment and consequently help restaurants develop an effective service improvement strategy. The realisation of this study is again due to the advantages of text mining.
In the third example (Ying, et al., 2018), we focus on the adoption of supply chain finance (SCF) service for small-and-medium enterprise (SME) financing. For the first time in SCF studies, we integrate natural language processing (NLP) method with content analysis – computer aided text analysis - to gather information and data from thousands evaluation and approval reports. In the case financial institute, we identify five lending techniques from the reports: comprehensive capability assessment, asset monitoring, financial analysis, cash flow controlling and supply chain collaboration. Further investigation indicates different emphases of these techniques, based on which we suggest a variety of SCF strategies and consequently discriminate their applicability in mitigating financing difficulty of SME borrowers. By documenting relevant themes and key lending techniques and their roles in SCF program, the study is the beginning of wisdom about supply chain finance practice and the findings should inform third-party financial service providers on how to improve their evaluation and approval strategy in different SCF scenarios. In this study, contents analysis is conducted by text mining.
In the end of this presentation, we illustrate the opportunity and possible difficulties of applying text mining for obtaining data and then for further research in Production Economics. Some potential study topics and future research directions are also discussed.
Prof. Ou Tang is a Professor of Production Economics at Linköping University, Sweden. Prof. Tang is an editor of International Journal of Production Economics. His publication has a focus on Supply Chain and Operations Management. More specifically, the research interest includes inventory modelling, planning and control systems, closed loop supply chain management, supply chain risk management, service operations, supply chain sustainability, text mining and China related operations management issues.
Professor TANG Ou
4:00 - 5:00 pm