Trading on Sentiment in Financial Text: An Ontology Rashmi Jha PhD Student College of Business Administration University of Illinois at Chicago 601 S Morgan Street (MC 294) Chicago, IL 60605 USA E-mail: [email protected] Arkalgud Ramaprasad, PhD Professor Emeritus, University of Illinois at Chicago Visiting Professor, University of Miami 421 Jenkins Building Coral Gables, FL 33124, USA E-mail: [email protected], [email protected] Keywords: Sentiment, Financial Text, Trading, Ontology 1. Introduction The study of sentiments in financial text has received considerable attention from information systems and finance researchers and practitioners over the past decade (for e.g., Antweiler and Frank, 2004; Loughran and McDonald, 2010; Tetlock’s 2007). They have examined the statistical relationships between financial market indicators and sentiments (coarsely categorized as positive, neutral, and negative) extracted from the text. However, the decision to purchase stock takes place in a highly competitive and ambiguous environment. It involves various agents (traders), different sources of textual information and conflicting views, opinions, and sentiments about the particular stock, market, or the economy as a whole. The generally used statistical textual analysis does not capture the semantic complexity of sentiments in the different types of texts fully. We propose a refined taxonomy of sentiment based on the extant literature on the topic. We then combine it with simple taxonomies of financial text and market behavior to propose ontology for articulating the logic of trading on sentiment in financial text. We will discuss how the ontology can be used to study the phenomenon systemically and systematically, and the potential insights from such study. The following describes the ontology of the Trading on Sentiment in Financial Text. 2. Ontology of Trading on Sentiment in Financial Text There is no standard definition of ontology of a complex, ill-structured problem. We will define ontology, as a logically constructed n-dimensional natural language model of the problem statement (Ramaprasad & Papagiri, 2009; Ramaprasad et.al, 201). The dimensions are derived from the problem statement. Each dimension is independent of the other and is taxonomy of discrete categories. Each taxonomy may be flat or hierarchical. Further, the order of categories in a particular dimension at a particular level of the taxonomy may be nominal (no particular order) or ordinal (based on some parameter). In the following we will first present the derivation of ontology for analyzing the problem statement “Trading on Sentiment in Financial Text”. We note that we present an ontology framework not the ontology. The dimensions of the ontology are derived logically from an interdisciplinary synthesis of the literature. This allows us to define the dimensions as well as the problem domain precisely and accurately. The ontology shown in Figure 1 has three dimensions, namely: (a) Sentiment, (b) Financial Text, and (c) Market Behavior. Each dimension is defined by a one- or two-level taxonomy. The taxonomies of the dimensions are logically constructed based on psychology and finance Financial Text Structured Annual reports Quarterly reports 10K reports … Semi-structured [financial text on] Sentiment Positive Love Joy/Happiness Fearlessness Neutral Surprise Negative Anger Sadness Fearfulness [sentiment in] [Influence of] (academic and practitioner) literatures. Market Behavior Stock price Stock volume Stock volatility News outlets Analyst recommendations … Unstructured Discussion forum Blogs … Figure 1. Ontology of Influence of Sentiment in Financial Text In the ontology we use a refined taxonomy for sentiment. Sentiment is usually classified into three broad categories “Positive”, “Neutral” and “Negative” in Finance literature(e.g. Das and Chen,2007;Tetlock,2007). These categories are too generic to explain in-depth the semantic mechanisms of influence of sentiment in financial text on trading behavior. For example, previous researchers have found evidence of differential influence of positive vs. negative sentiment but they have not examined the differential influence of the distinct sentiment of the same valence e.g., negative as suggested by the model proposed in Lerner and Keltner (2000) study. Our taxonomy of sentiment explains the dimension at a finer level of granularity. The second level includes the subcategories from (Ekman, Friesen and Ellsworth’s, 1972) emotion categories that are widely used in computer sciences research (e.g. Liu, 2010). Similarly, we have categorizes Financial Text as Structured, Semi-structured, and Unstructured. We expect the type and frequency of sentiments expressed in these types of text to be different and hence have differing impact on Market Behavior. 3 Application of the Ontology The ontology constitutes a closed description of the problem statement. It provides a detailed yet comprehensive description of the problem domain. Three illustrative components of research problem from 27 (3x3x3) level-1 components which can be derived from the ontology are: 1. Influence of Positive Sentiment in Structured Financial Text on Stock price; 2. Influence of Neutral Sentiment in Unstructured Financial Text on Stock volume; and 3. Influence of Negative Sentiment in Semi-structured Financial Text on Stock volatility. Each component may be instantiated in many ways empirically, in a few ways, or not instantiated at all. Thus, we can empirically specify the conjectures about the problem. First order conjectures would be to find out the ‘main’ effects of positive, neutral and negative sentiments (and their subcategories). Second order conjectures would be to find out the ‘interaction’ effects of positive, neutral and negative sentiments (and their subcategories). 4. Conclusion The ontology provides a systematic basis inquiry into trading on sentiment in financial text. The interaction mechanism among the dimension and between the dimensions can be studied and explained by the semiotic, psychological, or social sciences. They can be studied using statistical and simulation models. References Antweiler, W. and Frank, M. Z. (2004). “Is all that talk just noise? The information content of internet stock message boards.” Journal of Finance, 3, 1259–1294 Das, Sanjiv R. and Chen, Mike, Y. 2007: Yahoo! for Amazon: Sentiment Extraction from Small Talk on the Web 1386. Management Science 53(9), pp. 1375–1388, ©2007 INFORMS Ekman, P., Friesen, W. V. and Ellsworth, P. (1972), Emotion in the Human Face, Oxford University Press. Lerner, J.S.and Keltner, D. (2000). Beyond valence: Toward a model of emotion specific influences on judgment and choice. Cognition and Emotion, 14,473–493. Liu, B. (2010). “Sentiment Analysis and Subjectivity,” Handbook of Natural Language Processing, Second Edition. Loughran, T. and McDonald. (2010). When is a liability not a liability? Textual analysis, dictionaries and 10.ks. Journal of Finance 66, 35–65. Ramaprasad, A. and Papagari, S. S. (2009). “Ontological Design.” Proceedings of DESRIST 2009, Malvern, PA Ramaprasad, A., Prakash, A. N. and Rammurthy, N. (2011). Ramaprasad, A., Prakash, A. N. and Rammurthy, N. (2011). “Construction Project Management System (CPMS): An Ontological Framework.” Proceedings of Research & Academic Conference on Project Management. Pune, India: PMI. Tetlock, Paul C. (2007). “Giving content to investor sentiment: The role of media in the stock market.” Journal of Finance, 62, 1139-1168.
© Copyright 2026 Paperzz