This research developed a comprehensive methodology to quickly monitor key technical intelligence areas, provided a method that cleanses and consolidates information into an understandable, concise picture of topics of interest, thus bridging issues of managing technology and text mining. This research evaluated and altered some existing analysis methods, and developed an overall framework for answering technical intelligence questions. A six-step approach worked through the various stages of the Intelligence and Text Data Mining Processes to address issues that hindered the use of Text Data Mining in the Intelligence Cycle and the actual use of that intelligence in making technology decisions. A questionnaire given to 34 respondents from four different industries identified the information most important to decision-makers as well as clusters of common interests. A bibliometric/text mining tool applied to journal publication databases, profiled technology trends and presented that information in the context of the stated needs from the questionnaire. In addition to identifying the information that is important to decision-makers, this research improved the methods for analyzing information. An algorithm was developed that removed common non-technical terms and delivered at least an 89% precision rate in identifying synonymous terms. Such identifications are important to improving accuracy when mining free text, thus enabling the provision of the more specific information desired by the decision-makers. This level of precision was consistent across five different technology areas and three different databases. The result is the ability to use abstract phrases in analysis, which allows the more detailed nature of abstracts to be captured in clustering, while portraying the broad relationships as well.

Science and Technology (S and T) information presents a rich resource, essential for managing research and development (R and D) programs. Management of R and D has long been a labor-intensive process, relying extensively on the accumulated knowledge of experts within the organization. Furthermore, the rapid pace of S and T growth has increased the complexity of R and D management significantly. Fortunately, the parallel growth of information and of analytical tools offers the promise of advanced decision aids to support R and D management more effectively. Information retrieval, data mining and other information-based technologies are receiving increased attention. In this thesis, a framework based on text mining techniques is proposed to discover useful intelligence implicit in large bodies of electronic text sources. This intelligence is a prime requirement for successful R and D management. This research extends the approach called Technology Opportunities Analysis (developed by the Technology Policy and Assessment Center, Georgia Institute of Technology, in conjunction with Search Technology, Inc.) to create the proposed framework. The commercialized software, called VantagePoint, is mainly used to perform basic analyses. In addition to utilizing functions in VantagePoint, this thesis also implements a novel text association rule mining algorithm for gathering related concepts among text data. Two algorithms based on text association rule mining are also implemented. The first algorithm called tree-structured networks is used to capture important aspects of both parent-child (hierarchical structure) and sibling relations (non-hierarchical structure) among related terms. The second algorithm called concept-grouping is used to construct term thesauri for data preprocessing. Finally, the framework is applied to Thai S and T publication abstracts toward the objective of improving R and D management. The results of the study can help support strategic decision-making on the direction of S and T programs in Thailand.