Abstract: The application of text and data mining technology requires the input of a large number of works as an analysis database. The output works may also be substantially similar to several prior works, and the original crime of copyright infringement on the reproduction and use of works is not necessarily exempt from fair use. Using the input/output dichotomy to analyze text and data mining is an important tool to unravel the nature of the use behavior of their works. Based on this analysis, there is market failure in the acquisition stage of text and data mining input behavior, and it should be conditionally included in the category of fair use to promote the development and progress of technology. In the case that the output works do not constitute “transformational use”, the judgment of its infringement should break through the principle of separate comparison with the prior works, and judge its substantial similarity with the input works by the combined advanced standard of “whole perception method+abstract separation method”.

Key words: text and data mining, input and output, fair use, copyright infringement