Multiverse: Multilingual Evidence for Fake News Detection

. 2023 Mar 27;9(4):77. doi: 10.3390/jimaging9040077

Algorithm 3 Multilingual evidence for fake news detection: feature extraction.

Input: news information n, languages to use for comparison $l \in L$ the maximum amount N of news from the web search to compare with.
Output: cross-lingual evidence feature set $(s_{i}, a_{i})$ of similarity with the original news and source credibility rank for each news $w_{i}$ item from a multilingual web search.

1:
function Multiverse(n, L, N)
2:
$c r o s s_l i n g u a l_e v i d e n c e : = []$
3:
for $l \in L$ do
4:
$h e a d l i n e_{l} = Translate (n [h e a d l i n e], l a n g = l)$
5:
$W = Search (h e a d l i n e_{l}, t o p = N)$
6:
for $w \in W$ do
7:
$s o u r c e_r a n k = AlexaRank (w)$
8:
# For similarity score cosine- or NLI-based function can be chosen
9:
$s i m i l a r i t y = cross_lingual_news_similarity(n,w, l)$
10:
$c r o s s_l i n g u a l_e v i d e n c e . append (s i m i l a r i t y, s o u r c e_r a n k)$
11:
end for
12:
end for
13:
return $c r o s s_l i n g u a l_e v i d e n c e$
14:
end function