First, it seems important for us to explain the stakes of working in a multilingual context. These stakes are both linguistic and computational.
Dealing with multilinguality is fundamental for our research. It is a whole side of our working method. It is a way to structure and improve our knowledge on natural languages. In fact, multilingual treatments bring to make common linguistic features appear and to isolate language specific ones. We can say that multilinguality implies generalization.
Back to monolingual reflections, we notice that a detour by multilingual considerations is also a way of improving the knowledge on one particular language. We claim that these improvements could not have been obtained without this detour.
The computational improvements are mainly conceptual ones. Knowledge engineering take advantages of the separation between common and language specific features.
We are going to illustrate in concrete terms these claims. The framework states in the Natural Language Diagnosis and the problem is to find a good way of tokenizing an input.
| Up | Down | Previous | Next |