search

fandangOrg / fandango

FAke News discovery and propagation from big Data ANalysis and artificial intelliGence Operations
1 stars 1 forks source link

Relevance score for text #103

Closed jdgryse closed 3 years ago

jdgryse commented 3 years ago

Concerning this article: https://www.reuters.com/article/uk-factcheck-masks-idUSKBN29X2S5

The text relevance score is problematically low. The formula for the text relevance score is the following: IF( word_count > 400, 1, modulus( word_count / 400, 1 ) ). This article has more than 400 words, so the relevance score should be 100%, which is not the case (somewhere around 5% or so). So either the formula is not correctly implemented, or the text on which the score is being computed is not correct.

Screenshot 2021-02-04 at 10 46 37
mmagaldi-eng commented 3 years ago

I've checked into ES: body is > 470 words, so @pstalidis should revise the formula implementation.

pstalidis commented 3 years ago

I am not calculating the relevance score, I am using whatever is provided by the analysers. the text relevance score should be checked by @macagari

macagari commented 3 years ago

I checked that, there was an issue in the code, I'll update today the new image so you can let me know if it is working better

macagari commented 3 years ago

@jdgryse I took a look and I was calling an old module, now it is correct. I'll update it today or next Monday. Thank you!