Cognitive intentionality extraction from discourse with pragmatic-tree construction and analysis

In the research trend of moving towards a fine-grained analysis of subjective and cognitive information, another interdisciplinary and promising research direction in text analysis, intentionality extraction, has recently attracted research interest. Intentionality denotes the process through which humans conceive future situations, plan actions, predict the sensory consequences of the action, and update the prediction with self-changing means. Intentionality extraction is fundamental to the human ability to understand both the general laws that govern events and the particular principle of how and why a specific event actually occurred. As intentionality is a cognitive concept defined as the directedness of the mind towards a content or object, no previous research effort clearly defines and extracts intentionality in discourse. This paper begins by analysing discourse and cognitive intentionality and constructs the CIES-PT system for intentionality extraction based on the P-Tree model (a working model to analyse discourse qua sensible behaviour). CIES-PT applies an ant colony system to cluster similar discourse P-nodes to ensure high cohesion and hierarchically aggregates discourse P-nodes within one cluster to guarantee high coherence. In the final step, CIES-PT identifies intention connections among sentences at discourse levels based on the thematization principle. CIES-PT is examined with elaborately designed experimental tasks using the Reuters-21578 database with three classic metrics (Precision, Recall and F-Measure) and average computing time. The experimental results have demonstrated CIES-PT correctness and effectiveness.