Using Knowledge Organization Systems to Automatically Detect Forward-looking Sentiment in Company Reports to Infer Social Phenomena

We investigate whether existing knowledge organization systems (KOS) for strong and hesitant forward-looking sentiment could be improved to detect social phenomena. Five judges identified examples of strong/hesitant forward-looking sentiment that were used to compare the KOS developed in the study to existing models. The “composite” KOS was subsequently applied to annual company reports to generate word frequency and biologically inspired diversity ratios. Critical realism was used as a philosophy to interpret word patterns. Results indicate the composite KOS improved on existing models identified in the literature for strong forward-looking sentiment. In one company, a statistically significant association was found between increasing diversity of assertive forward-looking sentiment and subsequent declining relative business performance. This supported the Pollyanna effect: the social phenomena of over-positive business language in that company. Sharp increases in mentions of the “future” and “learnings” were discovered in another company which may be explained by an industrial disaster and subsequent crisis management rhetoric, supporting discourse of renewal theory. This study shows that improvements can be made to existing KOS used to detect forward-looking sentiment in reports. Adopting critical realism as a philosophy when analysing “big data” may lead to improved theory generation and the potential for differentiating insights.