How does natural language help solve interoperability issues?
The difficulties raised by the lack of interoperability within information systems are of two kinds:
- The ever-increasing variety of systems, devices, applications to be made to work together, each having its own interface (API) with its own formats, commands, parameters…
- The fracture between digital systems and their users: each solution has its own graphical interface, its dashboard, its dedicated application… The number of applications to load on a smartphone becomes excessive and often complicates the work of users.
To make all these elements interoperate, it is necessary to have a common language that is rich, scalable and durable. A first solution would be to define a posteriori this universal language; this is what many standardization bodies, manufacturers’ groups, etc. are working on.
Agora proposes another way: to generalize the use of the natural language, the one that humans have collectively defined and developed. Indeed this one presents many favorable characteristics:
- Rich, it allows to place commands (“open the door”) and to make requests (“what is the CO2 level?”);
- Flexible, it adapts easily to new situations (if you can do it, then you can say it);
- Intuitive, it allows many modes of expression: learning a language consists in mastering the rules of assembling words, not in learning “by heart” all the possible sentences (in almost infinite number);
- Natural, it reduces the gap between technical solutions and their users;
- Durable thanks to its fluidity, it is not about to go out of fashion, and its lifespan will be far superior to all technical standards and conventions;
- Common, it is the property of all and belongs to no one.
What are the characteristics of Agora's natural language processing (NLP)?
To benefit from the advantages of natural language and make digital systems interoperable, a number of technological challenges must be overcome. Agora’s NLP is the result of several years of research, conducted by our teams with the help of international specialists, and has been the subject of an academic publication. Its properties allow it to address the most constrained operational environments:
- High precision to cope with the ambiguities inherent to natural language, thanks to a contextual implementation;
- Deterministic, for high operational safety;
- Ultra-fast (a few milli-seconds);
- Lightweight, it can be integrated into edge functions, gateways and devices with Raspberry Pi-like computing resources
- Flexible, it is easily adaptable to new actions (“intent”), attributes, etc. and allows the easy handling of particular cases (jargons…);
- Multi-language, it detects the user’s language and facilitates interactions in international environments.