4 ms·
> In what universe, besides the one of a sentiment-analysis-algorithm, would that be considered a "negative" comment? The algorithm likely has issues identifyi
by zamalek 11y ago
> In what universe, besides the one of a sentiment-analysis-algorithm, would that be considered a "negative" comment?
The algorithm likely has issues identifying if criticism is constructive. If you look at the negative/positive score of each word in isolation (e.g. fail, problem, drop, can't, hard) the comment is overwhelmingly negative.
You actually have to parse and understand the comment to prove that it's very positive. They could have ditched the algorithm, but then they would lose a quantifiable positivity score and instead would have had to rely on human intuition - which has it's problems, e.g. spend the whole day looking at negative comments and neutral comments might start looking positive.
The study may have had fundamental issues at the outset.
- thaumasiotes 11y ago> instead would have had to rely on human intuition - which has it's problems, e.g. spend the whole day looking at negative comments and neutral comments might start looking positive. If you assume the people looking to validate their ideas are familiar with the HN community, then this effect makes the analysis more accurate.