Since the past several decades, scientific literature, patents and other semi-structured technology indicators have been generating and accumulating at a very rapid rate. Their growth provides a wealth of information regarding technology development in both the public and private domain. However, it has also caused increasingly severe information overload problems whereby researchers, analysts and decision makers are not able to read, summarize and understand massive technical documents and records manually. The concept and tools of technology intelligence aims to handle this issue.
In the current technology intelligence research, one of the big challenges is that, the frameworks and applications of existing technology intelligence conducted semantic content analysis and temporal trend estimation separately, lacking a comprehensive perspective on trend analysis of the detailed content within an area. In addition, existing research of technology intelligence is mainly constructed on the fundamentals of semantic properties of the semi-structured technology indicators; however, single keywords and their ranking alone, are too general or ambiguous to represent complex concepts and their corresponding temporal patterns. Thirdly, systematic post-processing, forecasting and evaluation on both content analysis and trend identification outputs are still in great demand, for diverse and flexible technological decision support and opportunity discovery.
This research aims to handle these three challenges in both theoretical and practical aspects. It first quantitatively defines and presents temporal characteristics and semantic properties of typical semi-structured technology indicators. Then this thesis proposes a framework of topic-based technology intelligence, with three main functionalities, including data-driven trend identification, topic discovery and comprehensive topic evaluation, to synthetically process and analyse technological publication count sequence, textual data and metadata of target technology indicators. To achieve the three functionalities, this research proposes an empirical technology trend analysis method to extract temporal trend turning points and trend segments, which help with producing a more reasonable time-based measure; a topic-based technological forecasting method to first discover and characterize the semantic knowledge underlying in massive textual data of technology indicators, meanwhile estimating the future trends of the discovered topics; a comprehensive topic evaluation method that links metadata and discovered topics, to provide integrated landscape and technological insight in depth. In order to demonstrate the proposed topic-based technology intelligence framework and all the related methods, this research presents case studies with both patents and scientific literature. Experimental results on Australian patents, United States patents and scientific papers from Web of Science database, showed that the proposed framework and methods are well-suited in dealing with semi-structured technology indicators analysis, and can provide valuable topic-based knowledge to facilitate further technological decision making or opportunity discovery with good performance.