In this course, we will discuss the design and construction of a computational lexicon. In particular, we present a computational treatment of Generative Lexicon Theory (GL) as given in Pustejovsky (1995), and illustrate how specific problems in the mapping from the lexicon to syntactic form are mediated in a highly distributed lexical semantic framework. We argue that the tasks of the lexicon designer should achieve at least the following goals:

1. Accounting for the interpretation of words in context;
2. Deriving the potentially infinite number of senses for words from finite resources;
3. Explaining the systematic relatedness between word senses in formal and predictable ways;
4. Providing a semantic representation that is applicable cross-categorially, and is not restricted to verb semantics alone.

We present various generative mechanisms which permit richer modes of composition within the sentence, methods that are both well-defined and computationally constrained. Central to these goals is the use of semantic underspecification as a tool for capturing linguistic generalizations.

We outline the type system necessary for capturing lexical semantic information within GL, and introduce a formal distinction between simple and complex types, and their role in accounting for logical polysemy. We discuss the nature of argumenthood and how best to capture verbal polysemy and polyvalency phenomena in a generative lexical framework. We then review the implementational strategies that are available to the grammar designer to account for these phenomena. The details of type coercion operations are discussed, with special attention paid to how coerced forms differ from canonical syntactic forms (csfs).

Other topics covered will include:

• The lexicalization of causative and inchoative semantics in language;
• The syntax and semantics of nominals and nominalizations;
• he formal distinction between event descriptions, propositions, and factives, and their canonical syntactic forms;
• Formal characterization of co-composition operations.

Finally, we discuss methodological issues for how to compare semantic frameworks with respect to: (1) The degree of composition within an expression; and (2) How many explicitly defined senses are necessary to accomplish a unique interpretation of the phrase.