The rapid development of edge-cutting technologies, including artificial intelligence, big data, and block chain etc., is radically changing the world. In meeting the need for protection of such new inventions, the number of patent applications for computer software related inventions has increased significantly in Taiwan. On the other hand, however, there is manifested a high proportion of rejection of computer software related inventions for not being in possession of “non-obviousness.”
In determining whether an invention is non-obvious over the prior art, the TIPO would search to locate all relevant prior art references and one prior art reference is chosen as the primary reference. Only if it is found that the located references can be combined in an obvious manner, shall examination be continued to determine whether a prima facie case of obviousness is established. Given that a computer software technology can usually be applied in various technical fields, it would be premature to conclude, only on the basis that the cited references are not relevant, that a skilled artisan would not be motivated to combine the references. The IPO thus revised the old examination guidelines in 2021 in this respect, among others, to rectify the surfaced inadequacies.
According to the revised examination guidelines, for a computer software invention, in the determination of whether an ordinary person skilled in the art can be motivated to combine two or more references, quite a number of parameters should be considered. It is inappropriate of the applicant to contend that an ordinary person skilled in the art would not have motivation to combine the cited references by simply stating that cited references are irrelevant; instead, detailed and sound reasons should be provided.
Moreover, a computer software related invention shall be denied its non-obviousness if it is a simple modification to the technical aspects of a reference, which include: (1) an adaptation of a prior art into a new form; (2) a systematization of the methods of operation carried out by humans; (3) using of a software to achieve a function performed by prior hardware technology; (4) an ordinary knowledge at the time of filing of patent applications for the reproduction of the virtual space of the computer; (5) an application or modification of ordinary knowledge at the time of filing, and (6) characteristics that do not contribute to any technical effect.
As mentioned above, computer software inventions are more prone to rejection for want of “non-obviousness.” For instance, in the case of an application for a neural network, if (1) the content of the invention is only the use of a general method of deep learning to generate some data and apply it to some specific field; and (2) there are already results from the use of conventional algorithms by prior techniques and the results have been applied to some specific field, the examiner may combine this prior technique using conventional algorithms with the prior art related to deep learning, even if the two prior techniques do not fall into the same or analogous technical fields. In addition, the examiner may not cite any of the prior techniques related to deep learning and determine straightforward that the claimed neutral network is a simple modification of conventional algorithms.
In response to the rejection of a computer-related invention, apart from making necessary amendments, to increase the odds securing a patent, the applicant is advised to (1) reiterate the advantageous effects of the invention over the relevant prior art as set out in the specification, (2) clarify that the invention and the cited references have a substantial difference in technology, and thanks to such difference an ordinary person skilled in the art would not contrive of making modification to the references; and (3) the references did not provide any teachings regarding making potential changes to the technology and therefore an ordinary person skilled in the art would not have incentive to make a modification.