the.com/charcnn

reads words like a detective reads ransom notes, one letter at a time.

means a convolutional neural network that learns from characters instead of whole words, building meaning up from letters, prefixes, and typos.

from emerged around 2015 as researchers like yoon kim and xiang zhang asked what happens if you skip word embeddings entirely and let cnns slide filters over raw character sequences, catching morphology and misspellings that word-level models miss.

for instance

kim et al 2016character-aware language model, beat word-level baselines on ptb

zhang and lecun 2015char-level cnn for text classification, no words needed at all

elmo character encoderused charcnn layers before feeding into bilstm, 2018

fasttext subword tricksrelated idea, embeddings built from character n-grams

the.com/
what’s happening now · the.com · generated