Weekly ML Resources Roundup — May 25
Even in this era of end-to-end NLP models, tokenization is performed in an explicit step. There have been several efforts recently to make character-level models work that take in UTF-8/Unicode sequences as input; however they have been often shown to underperform…