Gopher Protocol Inc., a company specializing in the development of Internet of Things (IoT) and Artificial Intelligence enabled mobile technologies, including a global mesh network technology platform for both mobile and fixed solutions, announced it has completed the Avant! AI Knowledge Retrieval Algorithm (KRA) and has commenced its cognitive system development.
In order to understand context in natural language, the content needs to be analyzed. This is considered an advanced task for AI systems. Avant! AI engine KRA includes robust semantic analysis for large text collections and multimedia objects, which includes several neural network architectures that have been incorporated into Avant! knowledge retrieval tasks. It is the goal to further expand the networks in the future, creating a cognitive system that will be able to be trained. The knowledge retrieval models are based on few methods, among them are feedback relevance, supervised probabilistic techniques and Bayesian inference networks. Each model is targeted to address a different aspect of the semantic analysis according to content type. The combination of wide spectrum of techniques to address knowledge retrieval enables fast and accurate context identification, extraction and understanding.
“One of the most important and difficult tasks in information retrieval is to generate queries that can succinctly identify relevant documents and reject irrelevant documents. In Avant! AI we are addressing this task by context divide-and-conquer approach,” provided Danny Rittman, Gopher’s Chief Technology Officer. “Our KRA system analyzes documents, identifies and indexes text and then assigns different technique to “understand” its meaning and make a decision if it is viable or nonrelevant. In the case where the text is providing essential information, it will be added to the relevant database, otherwise it will be rejected. Typically, it is difficult to accomplish a successful search/analysis at the initial search, therefore we conduct searches iteratively and reformulate queries based on evaluation of the previously retrieved documents. One of the advanced methods that we developed is for automatic generation of improved queries. This method is a derivative of relevance-feedback technique. A query can be improved iteratively by taking an available query terms and adding more from the relevant documents, while subtracting terms from the irrelevant documents. This approach, in addition to probabilistic learning strategies enables our Avant! AI to extract knowledge fast and with the highest accuracy. With its upcoming cognitive capabilities users will be able to train Avant! with knowledge and information,” continued Dr. Rittman.