The evolution and diffusion of technological knowledge by developing new Artificial Intelligence and text mining methods

点击翻译按钮获取中文摘要

In my research I intend to study the evolution and diffusion of technological knowledge by developing new Artificial Intelligence and text mining methods. One of the main challenges in the study of technological innovation is the measurement of innovative activities and of novel technological outputs. Since new products and manufacturing processes necessarily entail some degree of novelty and greatly vary across sectors and technological domains, the design of indicators that can measure the inputs and outputs of technological innovative processes presents challenges that have currently only been partially addressed. I propose to study the evolution and diffusion of technological knowledge by analysing large volumes of corpora of text. Given that scientific and technological knowledge can be considered embedded in language and vocabularies, the evolutionary patterns of novel ideas can be potentially traced in large volumes of textual data by extending the capabilities of existing Artificial Intelligence and text mining methods for information extraction. Available methods make use of set-theoretic, algebraic and probabilistic mathematical models in order to extract topics from collections of documents and classify unstructured textual data. Despite representing powerful tools for large scale content analysis, these methods are not suited for the identification and measurements of novel scientific and technological ideas. The main limitation is found in the very outputs of these methods, which in their general form are represented by clusters of frequently co-occurring and statistically correlated terms that are hardly interpretable and devoid of meaning. I plan to build on extant literature in cognitive psychology in order to extend these models and realise computational methods for the automated extraction of meaning and interpretable information from large corpora of text. Studies of concepts and categorization have a long tradition in cognitive psychology and offer theories and extensive experimental evidence for how individuals make sense of the large amounts of information they constantly and increasingly must deal with. Knowledge, and technological knowledge in particular, is conveniently categorized in hierarchical classification systems that minimize the cognitive costs of storage, processing and retrieval of information. I intend to show how novelty can be described and measured in terms of the elements of these classification systems and how novel ideas can therefore be traced within relevant textual data. Evolutional patterns of ideas and scientific and technological concepts, as they could be represented over time or across space and knowledge domains, would then offer a rich and interpretable description of technological innovation processes and indicators to measure their inputs and outputs. Building on studies of concepts and categorization, this theoretical framework would thus guide the development of methods for the automated extraction of meaning from collection of documents, which will likely leverage and expand the capabilities of existing methods in Artificial Intelligence and text mining. The data for the development, testing and validation of such methods is largely available and include patents' abstracts and descriptions, technical documentations, and the full texts of publications in trade and academic journals. The methods I intend to develop would be particularly suited for the study of the evolution and diffusion of scientific and technological knowledge and contribute to the field of Artificial Intelligence and text mining.