A Knowledge Graph Entity Disambiguation Method Based on Entity-Relationship Embedding and Graph Structure Embedding
Algorithm 1
Entity disambiguation algorithm based on entity and graph embedding.
Entity Disambiguation based on Entity and Graph Embedding
Input: A set of facts (h, r, t) with entity ambiguity in knowledge graph G, θ is the walk length of a node, x is the start node in G, Ω is the set of x’s neighbor node, and α is a threshold for choosing neighbor node.
Output: knowledge graph G without entity ambiguity facts
(1)
Initialize facts (h, r, t) in G
(2)
x = SampleRandomWalk(x)
(3)
jump = 0
(4)
while jump <θ do
(5)
if x ∈ Ω then
(6)
appendToKG(x)
(7)
if Rand() > α then
(8)
x = selectNeighborNode(x);
(9)
else
(10)
x = SampleRandomWalk(x)
(11)
jump = jump + 1
(12)
(13)
(14)
(15)
put (x', r, t') into G//x' is the highest similarity entity candidate
