Added a new feature to TM-RoBo's trademark search function that more deeply reflects application and registration status

IP-RoBo Inc. (Headquarters: Minato-ku, Tokyo; President: Masafumi Iwahara; hereinafter IP-RoBo) provides 'TM-RoBo', an artificial intelligence service that supports trademark searches such as the registrability of trademarks and the risk of infringement through use. We will add a new function to TM-RoBo's trademark search feature that reflects the application and registration status related to the searched trademark more deeply than ever before, and will begin offering this service on Wednesday, April 1, 2026. With this feature addition, the accuracy of splitting positions into individual words, indicators related to distinctiveness, and the final comprehensive indicator will be significantly improved, enabling easier, faster, and more accurate searches than before. ■ Background In April 2018, we released 'TM-RoBo', which quantifies various indicators useful for determining the registrability of searched trademarks using AI that has learned the judgments of experts such as the Patent Office and courts. Since then, we have aimed to dramatically improve the efficiency of trademark search operations by sequentially adding new features. Compared to conventional systems, which could only search by inputting a series of katakana pronunciations, the released in July 2020 is a groundbreaking feature that realizes the following: 1. By inputting the text trademark exactly as it appears (including kanji, hiragana, and alphabets), designated goods/services, and similar group codes, the AI assigns pronunciations and calculates linguistic strength statistical indicators (*1). 2. When a combined trademark consisting of multiple words is input, the AI exhaustively extracts all possible combination words that can be observed separately, conducts a detailed analysis for each combination, calculates various indicators such as TMR (*2) and TMC (*3), and outputs the final comprehensive indicator TMS (*4). 3. If necessary, it searches for and displays co-existing registration examples useful for determining the essential parts. *1 An indicator of the relevance between the target word and the target goods/services, relating to distinctiveness. *2 An indicator relating to how original the pronunciation of the target word is in relation to registered trademarks. *3 An indicator relating to the possibility that the target combined word will be observed separately. *4 A comprehensive safety indicator for the target combined word or searched trademark. ■ Challenges Regarding the accuracy of the linguistic strength statistical indicators related to distinctiveness mentioned above, we have improved it by adding numerous trial decision data and our own proprietary data. However, because the required volume of data is enormous, inaccurate results were sometimes output. To address this, we have sequentially carried out additional development to improve accuracy as follows: 1. February 10, 2022: Added linguistic strength statistical indicator data targeting specific fields. 2. October 1, 2024: Added examination stage data and examination standard data. In particular, the addition of examination stage data in step 2 contributed greatly to the improvement of accuracy, as the amount of data was several times larger than before. However, there are still a considerable number of words that do not appear at all during the examination stage, and improving the distinctiveness evaluation of these words remained a challenge. ■ Feature Overview To solve the above challenges, in addition to the conventional approach of collecting data judged to lack distinctiveness, we decided to add an approach that deeply analyzes the registration status of trademarks. Specifically, we have realized the following main functions.