The practical application of artificial intelligence advances in various fields by the appearance of deep neural networks. Automated translation is a typical success story. Higher accuracy of automatic translation by deep neural network after 2016 is revolutionary. There are not only general-purpose types but also field-specific types, and many services have been put on the market and are rapidly spreading. This paper explains the unique aspects of automatic translation in the background of higher accuracy, and describes in detail the significant contribution that non-researchers can make to higher accuracy of automatic translation. This paper also describes the present state and prospects of the automation of sequential and simultaneous interpretation, which will be possible by linking automatic translation with speech technology.

As the digitization and standardization have matured the intellectual property world, there are many examples of value creation associated with Digital Transformation (DX). As examples, we introduce Japio-AI translation and automatic classification for social issues, which have been done by the AI Research Center in the Japan Patent Information Organization (Japio). Regarding AI translation, the features and strengths of our machine translation (MT) is shown, focusing on the MT of Chinese patent documents. As for automatic classification, the significance of connecting social issues such as carbon neutrality and the Sustainable Development Goals (SDGs) to patent information and its impact on society. These patent information services are expected to promote DX in the field of intellectual property.

Recently, the application of artificial intelligence (AI) to intellectual property information business has become familiar. Several patent search tools using commercial AI have also appeared. However, the reality is that there are many basic items, limitations, and issues that users should keep in mind when using these commercial AI research tools. In this report, regarding the utilization of AI in patent search, we will look back on the introduction process of concept search in the past and describe the comparison between concept search and AI search from the viewpoint of the user of both the patent search system and machine learning.

PATENTSCOPE is a free patent document search service provided by the World Intellectual Property Organization (WIPO), which allows you to search and browse over 102 million patent documents (as of the end of March 2022). PATENTSCOPE continues to evolve, with the addition of some non-patent literature searches from March 2021, and the addition of a Markush structure search function for chemical compound searches from September 2021. This paper presents an overview of PATENTSCOPE’s continuing evolution, including its features and latest functions.

In case of patent searches by chemical structure search, index databases are indispensable because of the reliability of their chemical substance indexes and the convenience of their search functions. On the other hand, it was demonstrated that some full-text patent databases can search for chemical substances unindexed in the index database, and that it can search chemical substances without a time lag. It can be expected that they will be useful especially in prior art search before filing applications or invalidation material search. Chemical substance extractions of the full-text patent database differ in each system. And it can be said that it is insufficient in any system at present in terms of accuracy. We expect that enrichment of chemical substance dictionaries and advancement of AI technologies will enhance accuracy of the chemical substance extractions in future.

The Japan FARMDOC Association (JFA), a group of persons engaged in patent research at pharmaceutical companies, sets 5-6 research themes every year. In this paper, we introduce the results of two projects, “Markush Structure Study Group” and “Manufacturing Method Study Group”. The Markush Structure Study Group investigated the details of databases available at the time of 2017, and compared the results of several survey cases. At the Manufacturing Method Study Group held in 2021, we studied search method in process patents for small molecules and macrocycles, which have problems in exhaustive search, and obtained useful research know-how. Furthermore, it was shown in various survey cases that PatSnap could complement CAplus, and therefore the research is being continued this year.

Japan EPI Association studied IP related analytical methods under the subject of IP Landscape. The theme was “rice cooker”. Each of 3 teams separately analyzed on rice cooker manufacturers in domestic market. We checked external factors in the market, then we proceeded on SWOT analysis, 5 forces analysis, positioning analysis, patent analysis, text mining, word clouds and so on. Each teams led to each conclusion regarding business strategy based on these analysis. This article indicates some findings on analytical methods obtained through our research activities.