Discovering Intentions in Text and Semantic Calculus

The semantic relations between the constituents of a
text represent the core elements of lexical
semantics. Their discovery benefits numerous Natural
Language Processing applications, such as Text
Understanding and Question Answering. In this book,
we are focusing on INTENTIONs. Intentions refer to
future actions that one plans to do. They shed light
on the beliefs and the desires of an agent, which, in
turn, give clues about unstated intentions and
motivations behind an action. This research describes
several highly accurate mechanisms for automatic
detection of (implicit or explicit) INTENTIONs in
domain independent text paragraphs, presents a
detailed analysis of intention descriptors and
proposes a representation of intentions suitable for
recognizing textual entailment settings.

Our analysis of the interactions between INTENTIONs
and other semantic relations, such as PURPOSE and
CAUSE, led to a more general and formal approach to
combinations of semantic information, Semantic
Calculus - a calculus with semantics that captures
unstated information encoded in text. The impact of
the research proposed here is measured within
Question Answering settings.

Autorentext
PhD in Computer Science from The University of Texas at
Dallas. Research Scientist at Lymba Corporation, Richardson,
Texas. Main fields of research include semantic relation
detection, knowledge representation and natural language reasoning.

Klappentext
The semantic relations between the constituents of a text represent the core elements of lexical semantics. Their discovery benefits numerous Natural Language Processing applications, such as Text Understanding and Question Answering. In this book, we are focusing on INTENTIONs. Intentions refer to future actions that one plans to do. They shed light on the beliefs and the desires of an agent, which, in turn, give clues about unstated intentions and motivations behind an action. This research describes several highly accurate mechanisms for automatic detection of (implicit or explicit) INTENTIONs in domain independent text paragraphs, presents a detailed analysis of intention descriptors and proposes a representation of intentions suitable for recognizing textual entailment settings. Our analysis of the interactions between INTENTIONs and other semantic relations, such as PURPOSE and CAUSE, led to a more general and formal approach to combinations of semantic information, Semantic Calculus - a calculus with semantics that captures unstated information encoded in text. The impact of the research proposed here is measured within Question Answering settings.