Embeddings of Gene Ontology (GO) using OWL2Vec*

The naming convention of the files are as follows where each symbol is separated by a dash (e.g. 3-L-R-2-50): First Symbol - Document/Embedding Settings 1 - Structure Document, IRI embeddings [D(s), V(iri)] 2 - Lexical Document, IRI embeddings [D(s,l), V(iri)] 3 - Lexical Document, Word embeddings [D(s,l), V(word)] 4 - Lexical Document, IRI & Word embeddings [D(s,l), V(iri, word)] 5 - Combined Document - random, IRI embeddings [D(s,l,rc), V(iri)] 6 - Combined Document - random, word embeddings [D(s,l,rc), V(word)] 7 - Combined Document - random, IRI & Word embeddings [D(s,l,rc), V(iri, word)] 8 - Combined Document - traversal, IRI embeddings [D(s,l,tc), V(iri)] 9 - Combined Document - traversal, word embeddings [D(s,l,tc), V(word)] 10 - Combined Document - traversal, IRI & Word embeddings [D(s,l,tc), V(iri, word)] Second Symbol - Annotation Settings A - All annotations L - Limited annotations (label, hasExactSynonym) Third Symbol - Walker Type R - Random walker WL - Weisfeiler Lehman sub-tree kernel Fourth Symbol - Walk Depth 2 - creates walks similar to (A, subclassof, B) 4 - creates walks similar to (A, subclassof, B, subclassof, C) 6 - creates walks similar to (A, subclassof, B, subclassof, C, subclassof, D) Fifth Symbol - Embedding Size 50 - output vector space will be [44,272, 50] 100 - output vector space will be [44,272, 100] 200 - output vector space will be [44,272, 200] Example 3-L-R-2-50 Will extract lexical information from the ontology and output Word embeddings only [D(s,l), V(word)]. The annotations used are limited (label and hasExactSynonym). It uses a random walker to extract ontology information. The walk depth is 2. The output embedding size is 50 for each class.