Technology

Large Linguistic Models proved their usability in many various domains of human activity. As Artificial Neural Networks (ANNs) they work mainly in the request-response mode. However today we can also construct AI systems using ANNs that can initiate interaction. An example could be systems collecting knowledge. They can ask questions about issues that are poorly or not covered at all. Often AI system’s activity is driven by some goal function. Such function “tells” the system how far it is from the ideal solution or if the activity direction is properly chosen. Goal functions can be rewarding or penalizing ones, to put it simply. In machine learning (ML) domain we call this process “reinforcement learning with reward”. There are of course also more versatile and complex approaches that do not need any human engagement. ML system takes a lot of data and tries to adjust to the underlying patterns. This trait is also specific to the so called self-organizing dynamical systems. There is no need for any theory, assumptions or algorithms formulated by human beings. Such systems can be fully autonomic allowing us, humans, to focus on more important and essential topics.

Using various approaches to ML in AWT we can use AI solutions to optimize the foreign language learning process. Moreover, due to the automation and relatively low costs, it can be made individually. AI assistant can work and train linguistic skills with group of people engaging them mainly into human-human interactions as well as it can help to develop competences using human-machine individual approach. ML systems, that we call AI agents, can involve also in a group activities forming mixed teams including many humans and many machines. No matter what the approach is, the goal stays the same - to engage students in learning any foreign language using their interests.

ANNs can take as an input a lot of different kind of data. One of them can be pure texts or images. Although in so called multi-modal approach they can take the picture with sounds and description. As an example could be video clip of a pop song. We have one concise and interdependent stream of data including images (“pure” video formed by frames changing in time), sounds (e.g. musical sound track) as well as texts (lyrics transformed with use of speech to text methods). AI tries to “understand” internal relations between different data and use them for knowledge and skills development. Then it can be able, e.g, to translate or transcript movies and video clips containing sounds and music.

We all know the good and well anchored teaching practices like listening to songs the students like, writing essays, reading books, watching movies or TV news channels, just to name a few. They all need literally a huge amount of work and can be also boring for students when improperly chosen. Moreover, with one teacher in a class we can simply forget about selecting material to suit individual needs and hobbies. However, with great help of AI and modern communication technologies available worldwide and also decreasing computational costs we can reach a whole new level of foreign languages learning.

AI technology allows us also to engage students of different ages with use of widely understood “gamification”. We all know that the more involvement and fun one has while learning, the more effective the learning process is. It can be not only achieved, but also remarkably reinforced by AI assistance with gamification of the the learning activities. Through team games and plays focused on student interests and hobbies AI assistant can become a creative, tireless, always emphatic and forgiving as well as cheap and massively replicable Game Master and team leader. And when a student expresses the desire to lead the group or to make a new storytelling, AI will help them to achieve results optimal for the entire learning team.