BotSupply ^Companies

The Chatbot CMS. Where creativity and AI intersect. We combine the latest AI trends with our creative vision to create, shape & build leading AI solutions & bots, that feel and work great. We help leading brands drive their business outcomes creating for them great customer experiences and products using our platform, conversational UX and cutting edge AI. - The Conversational Experience Platform: Using our extensive hands-on experience building bots we're building - in our opinion - the ultimate enterprise-level bot platform. The Conversational Experience Platform enables anyone from a partner to a customer support employee to a marketeer to control all aspects of their bot from building, deploying and measuring it, to setting up advanced personalization, testing and more. Our proprietary NLP (see below) comes built into the platform with add-on modules for Watson, Luis, Dialogflow et al. The cloud-first modular platform is highly customizable and offers APIs for developers to add custom integrations and build individual platform experiences to each bot owner. - Proprietary AI: Our world-class AI team have created proprietary algorithms which can understand Danish, English, Italian, Swedish (beta), and Norwegian (beta). We use the power of representational learning which helps our NLP core engine to understand the context in any given query without performing any kind of translation. Our NLP core engine is a hybrid Architecture of Statistical Modeling and Neural Language Modelling. On top of that to achieve higher accuracy on any provided domain we perform transfer learning with the provided dataset. The approach of transfer learning and domain adaptation has been inspired by Computer Vision applications, and we have empirically proved it to be very effective for many of our use cases. Furthermore, this makes our NLP engine to be domain and language agnostic. We are also researching on automating the continuous learning process, which will help our engine to improve upon feedback. We're researched on developing an algorithm where the model only modifies the old learned information if the new information is related to it. Otherwise, all the previously learned weights are kept intact. We train a memory model to keep a track of all the past relationships(similarity measure) within the data and a contribution measure of each data point to the global loss. This way we can avoid catastrophic forgetting of the previous correct learned weights