Date of Award


Degree Type


Degree Name

Master of Computer and Information Science (MCIS)


Computer Science

First Advisor

Dr. Jorge Fandinno


This research project goal is to achieve explainable automatic text understanding and reasoning for simple English narratives focused on the use of action verbs. text2alm is a system developed at UNO in the scope of a master thesis by Craig Olson. It takes a narrative in English and processes it into a logic program under answer set semantics. This program is then processed with the state-of-the-art answer set solver clingo, which finds an answer set for the program that represents entities and events occurring within the narrative as well as encodes state of affairs at various points of the narrative timeline. In other words, text2alm can be seen as an information retrieval system that converts unstructured information present explicitly and implicitly in English narratives into structured form supported by the predefined vocabulary of relations within text2alm. tExplain, a system designed at UNO within MS project by Adrian Dorsey, extends text2alm by replacing answer set solver clingo with system xclingo. xclingo is a novel tool in answer set solving that is capable of providing human readable explanations to inferences performed by clingo. In comparison to text2alm, tExplain produces explanations for extracted information from narratives. We improve the tExplain system by allowing automation of creating narratives, queries, logic programs, and xclingo output. In this work, we target rigorous evaluation of tExplain capabilities by applying it on the Facebook dataset bAbI. This dataset contains simple narratives which text2alm targets. These narratives are annotated with questions, answers, and explanations. The quality of the explanations generated by tExplain are evaluated against bAbI ’s annotations. This evaluation will also lead to extensions of tExplain capabilities and better understanding of methodological use of system xclingo.


Copyright held by author.