4 ms·
I have found pre-trained models for language detection and Wikipedia word embeddings useful. Everything else I have to train from scratch. I have successfully d
by rpedela 8y ago
I have found pre-trained models for language detection and Wikipedia word embeddings useful. Everything else I have to train from scratch. I have successfully done NER on search keywords where there were many examples of misspelled words or weird punctuation. I used Spacy but I had to train the model from scratch.