Determining the size of the addressable market is a key aspect of market intelligence and requires identifying and delineating projected budget data from potential customers. The market intelligence arena is characterized by a wide range of disparate sources, many of which are unstructured, ranging across competitive, market, financial, and technology sources, and typically necessitating significant manual work to analyze, reconcile, and integrate. The authors present an approach for classification of data from one of these sources, facilitating aggregation and analysis of intelligence information. We describe a concept proof using machine learning that extends a model for automatic mapping of publicly available budget data to segments and subsegments of a market segmentation taxonomy. This approach automates the tagging of market and market segment for each program and cost element by training classification models on the manually labeled historical data. We describe the evaluation and use of multiple natural language processing (NLP) and classification modeling methods. This work's contribution is demonstrating how NLP and machine learning techniques can provide useful data classification and automatic classification even when source data diverges from its specified taxonomic description.