Issue 5, Oct 19, 2026
The Bigram Patch
NLP Engineer interview at Cohere, and the interviewer asks:
“Your cheap bag-of-words filter handles most traffic because the LLM costs far more per call. Support shows you it flags ‘not bad at all’ as negative. A teammate proposes adding bigrams. Why is that only a partial fix?”
Don’t say: “Add bigrams. Then ‘not bad’ is one feature and the negation is handled.”
Bigrams do reach an adjacent negator, and they do raise sentiment accuracy which is exactly why "we added bigrams" is not an answer to a negation complaint.
Share this trapShare on LinkedIn
The full answer, with the mechanism and the arithmetic, is for paid subscribers on Substack.