Gavagai

Gavagai has built an unsupervised and self-learning system that understands and continuously learns from streaming unstructured language/text data. Our approach solves some of the most important problems in language technology: the variability problem, the out-of-vocabulary problem, and the problem of many languages that is, the possibility of using different words or expressions to communicate the same idea, the problem of terms encountered in input which are not present in a system's dictionary or database of known terms, and the problem that English is only one of many languages, respectively. Gavagai’s scalable approach to these problems is necessary for dealing with real world NLP applications. One such application is for example, automatic analysis and coding of open-ended free-text responses from surveys and similar. Here, texts can be about anything, they can be misspelled, and they might not abide by correct grammar. Furthermore, answers can be in Chinese, Indonesian, Russian, or in any number of languages, and they might contain expressions a traditionally trained model has not seen before.

show less