Granular Causality Applications: Using Part-of Relations for Discovering Causality

Causal markers, syntactic structures and connectives have been the sole identifying features for automatically extracting causal relations in natural language discourse. However, various connectives such as “and”, prepositions such as “as”, and other syntactic structures are highly ambiguous in nature, as they have multiple meanings besides causality. As a result, one cannot solely rely on lexico-syntactic markers for detection of causal phenomenon in discourse. This paper introduces the Theory of Granular Causality and describes a new approach to identify causality in natural language. Causality is often granular in nature (Mulkar-Mehta, 2011; Mazlack, 2004), and this property of causality is used to discover and infer the presence of causal relations in text. This is compared with causal relations identified using just causal markers. A precision of 0.91 and a recall of 0.79 is achieved using granularity for causal relation detection, as compared to a precision of 0.79 and a recall of 0.44 using text-based causal words for causality detection. Next, the author presents the findings for discovering causal relations between two sentences in an article. The system achieves a precision of 0.60 for discovering causality between two sentences using granular causality markers as features. The results are encouraging, and show that the granular causality is an important phenomenon in natural language DOI: 10.4018/jcini.2012070105 International Journal of Cognitive Informatics and Natural Intelligence, 6(3), 88-108, July-September 2012 89 Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. tions using Elementary Discourse Units (EDUs). Neilson (1996) describes the way in which the semantic and pragmatic functions of causal markers can be accounted for in terms of linguistic and rhetorical theories of argumentation. All these works consider causality as a sequential set of events at the `same’ level of descriptive specificity. However, causality is often described using a granular structure, where the coarse grained event is described as happening because of a fine grained event. For instance, in a building collapses because the roof caved in, the roof is in integral part of the building, and is a sub-event of the entire building collapsing. This paper focuses on granular causality, and how such granular causality structures can be used to identify causal relations in text. We use the phenomenon of granularity on a regular basis in our everyday life. For planning and scheduling of important tasks, we often divide or split our tasks into smaller pieces, until each task is easily manageable. For instance, the day-to-day activity of shopping for groceries involves some finer grained events such as driving to the grocery store, carrying a list, picking out required items, and paying the cashier. Each of these events in turn involves some finer level events; e.g., driving to the grocery store involves sub-events like opening the car door, starting the engine, planning the route, and driving to the destination. The sequence of fine grained events make up a coarser grained event. When the fine grained events are completed successfully, the coarse grained event is completed successfully. In this sense, granularity decomposition is script or plan decomposition. Historically, to discover causality in discourse, certain words are used as causal markers. Certain markers are domain-specific; others (the most frequently occurring causal markers) are generic; but there is a long tail of high precision but low frequency causal marker words. For instance, the words ‘because’ and ‘cause’ are considered causal markers, but ‘behind’ is not generally considered a causal marker, although it often refers in causality in football newspaper articles, as in The Miami Dolphins went ahead 21-6 at halftime behind three touchdown throws by Dan Marino. Word-based causality detection techniques almost always miss the less frequently occurring but high precision causal markers in discourse. This paper describes the application and implementation of the Theory of Granular Causality for discovering causality and causal markers in various domains. Three experiments are performed: (i) Apply the Theory of Granular Causality for discovering causality within a sentence, (ii) Use the Theory of Granular Causality to extract low frequency causal markers from sentences containing causality, and (iii) Use the Theory of Granular Causality to discover causality between two sentences.