Vocabulary size: closed vs. open
A decisive determinant of both a speech BCI's usefulness and the difficulty of the task. A closed (fixed) vocabulary decoder classifies among a small pre-defined set of words or sentences, for example 50 words, which is accurate but expressively limited. An open-vocabulary decoder emits sub-word units (phonemes or characters) and can in principle produce any word, including proper names and out-of-vocabulary items, at the cost of higher error.
Progress in the field is frequently measured as pushing accuracy up while enlarging the vocabulary, from tens of words to well over one hundred thousand.
A 50-word closed system near 100 percent accuracy and a 125,000-word open system at roughly 24 percent WER are not the same technology at different settings, they solve different problems.