In a time when the employment of natural language processing techniques in domains such as biomedicine, national security, finance, and law is flourishing, this study takes a deep look at its application in policy documents. Besides providing an overview of the current state of the literature that treats these concepts, the authors implement a set of natural language processing techniques on internal bank policies. The implementation of these techniques, together with the results that derive from the experiments and expert evaluation, introduce a meta-algorithmic modelling framework for processing internal business policies. This framework relies on three natural language processing techniques, namely information extraction, automatic summarization, and automatic keyword extraction. For the reference extraction and keyword extraction tasks, the authors calculated precision, recall, and F-scores. For the former, the researchers obtained 0.99, 0.84, and 0.89; for the latter, this research obtained 0.79, 0.87, and 0.83, respectively. Finally, the summary extraction approach was positively evaluated using a qualitative assessment.
Top1. Introduction
Data are the pollution of the information age, since they are created and are here to stay (Spence, 2010). This increase in the flow of data that organizations create and collect, necessitates the need to leverage these resources and extract information and subsequent knowledge. The large stream of data unveiled two data formats, namely structured and unstructured data, with each of them requesting different treatment methodologies to derive knowledge. Although many argue that this process is less-time consuming when the data have a consistent representation and a predefined structure, only 20% of the data that organizations have, are actually found in this manner. The rest of the data are found in an unstructured format (Grimes, 2008). These data have no consistency in appearance and are usually text-heavy, making it a challenge to extract patterns and relationships from and among them. This has called for the introduction of Text Mining (TM) as a discipline that “analyses text to extract information that is useful for particular purpose” (Witten, 2004). We consider TM to be a multidisciplinary field which utlizes data mining techniques, information extraction, information retrieval, machine learning, natural language processing (NLP), statistical data analysis and graph theory, among others (Miner et al., 2012). The wide range of techniques that TM fosters, together with its applicability in domains such as biomedicine, national security, finance, social studies, law and so on, show the prominence of such data analyzing techniques (Bholat, Hansen, Santos, & Schonhardt-Bailey, 2015; Fan, Wallace, Rich, & Zhang, 2006; Friedman, Johnson, Forman, & Starren, 1995; Haug, Ranum, & Frederick, 1990; Menger, Scheepers, & Spruit, 2018; Zhao, 2013).
Nevertheless, not all disciplines have been able to taste the same riches. Policies are industry-wide documents that represent written guidelines of acceptable actions to which organizations must adhere. Financial institutions, especially banks, can be thought of industries that have a high number of policies in place. These organizations continuously introduce policies, in order to be fully complaint with regulations that governing bodies impose. Nevertheless, even though such documents are found across industries, they still lack in standardization (A. I. Anton & Earp, 2003) and their domain-specific language often makes them incomprehensible (A. Anton & Earp, 2004). Considering the importance of these documents for the business, but at the same time their inconsistent and exhausting representation, a perception was first created in (Spruit & Ferati, 2019) that TM and its techniques can be used to bring order and understanding into them. This is what motivated the compilation of the following three related research questions (RQ):
- 1.
To what extent has TM been applied on policy documents?
- 2.
Which TM techniques or frameworks have been applied on policy documents?
- 3.
Which TM techniques can be used to obtain information that would enable an easier navigation through the policies?
Nevertheless, in our attempt to validate this perception, the scientific body of literature showed a landscape different from what was anticipated. Thus, next to providing an overview of the current state of literature that treats the concept of using TM on internal bank policies, this paper also introduces a novel TM framework for processing internal business policies. From a design science research (DSR) perspective, we present a meta-algorithmic model (MAM) as the main DSR artifact to structure and support analyses of internal business policies, which was validated through a case study at one of the biggest banks in The Netherlands (Spruit & Jagesar, 2016). The rest of the paper is structured as follows. In Section 2 we present the research approach, which is followed by the related literature in Section 3. Section 4 presents the case study methodology, whereas the results of the case study and implications are presented in Section 5. Conclusions are formulated in Section 6, whereas discussions, limitations and future work are presented in Section 7.