Open taalbrecht opened 8 years ago
Four possible implementation ideas: -Cosine distance from custom weighting vector as mentioned in issue post. -LDA for each sentence with one topic seeded with conclusion words -Random forest classification using training set -latent class analysis with training set
Keyword list for vector of taglines that are weighted to identify the sentences that state article conclusions (for example, weighted vector that might place weights of .8 on "summary", .15 on "conclusion", 0.05 on "result")