Training a model has proven to be too inaccurate with the sizes of corpora snelSLiM will often be worked with (far less than 10 million, often even smaller than 1 million words), including pre-trained, high quality models for every relevant language would not be realistic, as well as yield problems with registers, dialects, colloquial language, neologisms, etc.
Training a model has proven to be too inaccurate with the sizes of corpora snelSLiM will often be worked with (far less than 10 million, often even smaller than 1 million words), including pre-trained, high quality models for every relevant language would not be realistic, as well as yield problems with registers, dialects, colloquial language, neologisms, etc.