jloveric/neuralphrasex

Mixed neural network / fuzzy similarity based phrase database matching. Uses universal sentence embeddings from Tensorflow combined with KNN and fuzzy similarity to perform slot filling based on a list of potential phrases. The approach allows for one shot learning and the efficiency will be determined by KNN (or other approximate nearest neighbor). The name "PhraseX" comes from "Regex", as it is a more flexible way of matching phrases than using regex and better suited for natural language.

JavaScript

MIT License

Total donated

Undistributed

Share with your subscribers: