This thesis describes a tool that facilitates rapid modeling of devices from informal documents. The ModelMaker tool facilitates the construction of models by analyzing the source specification document and presenting it based on the modeler's need. ModelMaker analyzes and indexes the source document for the noun phrases and identifiers it contains. When the modeler specifies the name of a pin or device that the modeler is working on, ModelMaker will recover any behavioral or structural information about the particular pin or device. ModelMaker can return this information based on the order of how relevant the information is to the model that the modeler is trying to build. The modeling language that can be used for this tool is VHDL. The initial VHDL model is derived from a block diagram of the source document using a schematic capture tool. This VHDL model can be edited inside ModelMaker to add behavioral code and to insert source document fragments as comments. / Master of Science

Specifications of digital hardware systems are typically written in a natural language. The objective of this research is automatic information extraction from specifications to aid model generation for system level design automation. This is done by automatic extraction of the noun phrases and the verbs from the natural language specification statements. First, the natural language sentences are parsed using a chart parser. Then, a noun phrase and verb extractor scans these charts to obtain the noun phrases with their frequencies of occurrence. The noun phrases are then classified by semantic types. Also the verbs are automatically assigned their respective roots and classified. Finally, each sentence is summarized as a sequence of "chunks": noun phrases, verbs and prepositions. Vectors are generated from these chunks and imported into MS Excel for plotting occurrence graphs of noun phrases and verbs with respect to the sentences in which they occur. Finally, inter-term dependencies between noun phrases, and between noun phrases and verbs were studied. The frequencies of occurrence, the classification of chunks, the occurrence graphs and the inter-term dependencies together give useful information about the subject, the hardware components and the behavior of a system described by a natural language specification document. / Master of Science