NTT Technocross Corporation will begin offering the “tasokarena Masking Tool Local LLM Edition” from May 15, 2025, which utilizes generative AI to further improve the accuracy of the masking function.
Due to legal provisions such as the revised Personal Information Protection Act, it is now possible to use customer data that has been processed in accordance with the law for research and analysis. Accordingly, the use of text data such as response history in the contact center field is increasing, as well as for training AI models at AI generation service providers.
Also Read: Dataiku Adds AI Agent Tools to Tackle AI Adoption Barriers
In this context, Tasokarena, an anonymous processing information creation software that can both anonymize and mask data, will begin offering a local LLM version of its masking tool, which uses a new local LLM to improve the extraction of masked targets. By using local LLM for text masking, the extraction rate of personal information has been significantly improved by understanding the context, taking into account the characteristics of Japanese, which contains a mixture of kanji, hiragana, and katakana. This enables masking with higher accuracy, reducing the burden of visually checking and correcting masked data. Furthermore, by using local LLM, masking can be performed without connecting to an external network, reducing the risk of confidential information leaks.
SOURCE: NTTTX
