We use cutting edge Deep Learning architectures to build in house Language Models in Danish, Italian and English.
The model performs excellent for Information Retrieval (IR) and Information Generation (IG).
The research focus is on making the Neural Network based language models learn in real time by incorporating Reinforcement Learning (RL) based feedback mechanism to it.
A Recurrent Neural Network based encoder-decoder chatbot that learns to generate answers on the fly.
The chatbot learns from the conversation and generates answers based on that itself.
A real time feedback mechanism that can be incorporated to any model and it learns real time using Q learning approach based on the defined rewards and penalties.
We research in Deep Learning based architectures to model high level abstractions in data.
We research in beating the state-of-the-art accuracy in specific use cases by experimenting with Auto Encoders, Convolutional and Recurrent Neural Networks.
These technologies are used in identifying objects in images, visual search, object classification and detection and many more.
Recognize products in real time without the hassle of clicking pictures and getting the results back.
Models are hosted in Mobiles devices and not in clouds.
Solving the most important problem with Deep Learning Architectures: less data.
Optimizing the architecture to learn maximum with fewer datasets.
