From: KAGN:knowledge-powered attention and graph convolutional networks for social media rumor detection
Notation | Explanation |
---|---|
\(D\) | The training news samples |
\(P\) | A piece of news composed of a sequence of words |
\(EP\) | The relevant entities of P |
\(CE\) | The entity-related concepts of EP |
\(p\) | The representation of P |
\(EP^{\prime}(q^{\prime})\) | The representation of entities |
\(CE^{\prime}\left( {r^{\prime}} \right)\) | The representation of entity-related concepts |
\(e_{i}\) | An entity in knowledge base |
\(ce_{i}\) | Set of all concepts for an entity |
\(e_{i}^{\prime }\) | Embedded representation of an entity |
\(ce_{i}^{\prime }\) | Embedded representation of a concepts set |
\(c_{j}\) | A concept of \(ce\left( {e_{i} } \right)\) |
\(\tilde{q}\) | Entity representation incorporating conceptual knowledge |
\(\tilde{p}\) | Textual representation of news incorporating entity and concept knowledge |
\(\hat{p}\) | Textual representation of news obtained by gating mechanism |
\(G\) | The post-entity-concept graph |
\(EN\) | The unique entities nodes sets of graph G |
\(en_{i}\) | A uniquely numbered entity node |
\(CN\) | The unique concepts nodes sets of graph G |
\(cn_{i}\) | A uniquely numbered concept node |