I am currently testing a proof of concept around DeepDive (relatively similar to the spouse example in terms of structure but with a few additional tweaks). When testing precision I get good results, but when testing recall I came across an issue: I incidentally had the same article twice in my dataset, and I extracted the same mention (obviously) in both articles. I use the standard feature generation process used in the spouse example, but I happen to find 2 different expectation numbers for the same mention.
To be clear, I get in article_1/sentence_x expectation for mention_1 equal to say 0.32, and in article_2/sentence_x expectation for mention_2 equal to 0.64 (with article_1/sentence_x exactly the same as article_2/sentence_x, and mention_1=mention_2).
How can that be? NB I use v0.9.
Many thanks.
I am currently testing a proof of concept around DeepDive (relatively similar to the spouse example in terms of structure but with a few additional tweaks). When testing precision I get good results, but when testing recall I came across an issue: I incidentally had the same article twice in my dataset, and I extracted the same mention (obviously) in both articles. I use the standard feature generation process used in the spouse example, but I happen to find 2 different expectation numbers for the same mention. To be clear, I get in article_1/sentence_x expectation for mention_1 equal to say 0.32, and in article_2/sentence_x expectation for mention_2 equal to 0.64 (with article_1/sentence_x exactly the same as article_2/sentence_x, and mention_1=mention_2). How can that be? NB I use v0.9. Many thanks.