This asset is composed of Natural Language Processing (NLP) components tailored for integration in games. All components are available online, through an Internet connection, and rely on services provided by a computing framework at L2F/INESC-ID. Additionally, some components are also available for offline usage.
Asset integration in a game is achieved with simple networking principles, as each component follows a client/server architecture. For components featuring the offline usage mode, this architecture allows a component’s server to run on the local machine or on a remote server providing the service to several clients. Component usage in online mode only requires integration of clients in the game development framework, since the server part is only accessed/managed remotely (through HTTP requests).
As most of our components are based on Machine Learning (ML), language coverage is limited by availability of training data, thus variable across components, with some supporting multiple languages while others only support Portuguese or English.
The asset is fully functional and ready to use by end developers, for online usages (relying on remote services, which require an Internet connection) and with the Unity game engine (code guidelines not yet implemented).
Some of the asset’s components are still being improved, even when using the online mode. Namely, speech to text is not complete and emotion detection from text is not yet validated.