FRONTEO's self-developed AI engine "KIBIT" Unique Features
Four unique features of AI "KIBIT
Vectorization of proprietary technologies 
Proprietary algorithms consisting of many patents
Intuitive visualization of data
Green Micro AI for energy savings
Point 1
Vectoring proprietary technology
One of KIBIT's main features is that it has developed its own technology faithful to the distribution hypothesis*1 for "vectorization," or the replacement of words with numbers, which is indispensable for analyzing natural language (human words) using AI.
Technically speaking, this technique "captures the relationship between words and sentences based on word co-occurrence relationships. The approach to vectorization using a method derived from the distributional hypothesis has been proven to produce better results than Transformer, which is widely used in generative AI and other fields*2.
*2〈 Yamada et al.
Point 2
Proprietary algorithm consisting of many patents
To perform advanced analysis of documents (natural language) and networks, we have developed proprietary algorithms that differ from the Transformer* widely used in generative AI. We have more than 70 patents related to algorithms and data analysis and visualization methods, such as word pattern analysis to capture context, appropriate weighting of words, and optimization of feature values.
Furthermore, since we even provide in-house developed software that incorporates AI, we can implement it while customizing and additionally developing it according to the company's data and issues.
* Transformer: one of the models of deep learning, a method of natural language processing published in 2017.
Point 3
intuitively visualize and map data
The results of data analysis by KIBIT are not only output, but are also visualized to further enhance their value.
By plotting appropriately processed data* on a flat surface and creating a "two-dimensional map" that can be visually recognized by humans, it is possible to see how the information interacts and relates to each other through color, scattering, and other expressions, leading to the discovery of unexpected information that would otherwise be missed in a simple analysis. This leads to "serendipity," the discovery of unexpected information that would otherwise be missed by mere analysis.
Point 4
Green Micro AI for energy savings
Generative AI, for which expectations are rising from all quarters, is actually pointed out to consume a large amount of electricity and the associated CO2 emissions and water consumption perspectives.
KIBIT, on the other hand, is a power-saving and environmentally friendly (=Green) AI with simple algorithms using minimal parameters (=Micro). The strength of KIBIT is its ability to analyze and learn data at extremely low power consumption and high speed, using the CPU level of an ordinary PC rather than a GPU comparable to that of a data center.
