Machine translation of entities

1. A method for performing machine translation, comprising:

identifying an unlinked entity pair within a cross-lingual entity graph;

the cross-lingual entity graph comprises a plurality of source entities, a plurality of target entities, and a plurality of links;

at least two of the plurality of links comprise different link types;

the unlinked entity pair comprises a source entity of the plurality of source entities and a target entity of the plurality of target entities;

determining a sub-graph of the cross-lingual entity graph that includes the unlinked entity pair and at least one link of the plurality of links;

inputting the sub-graph into a graph neural network (GNN) that has been trained using different link types;

in response to inputting the sub-graph into the GNN, outputting, by the GNN, a value;

determining that the value satisfies a criterion;

creating and storing a link between the source entity and the target entity in the cross-lingual entity graph;

the link indicates that the target entity is a valid translation of the source entity; and

the cross-lingual entity graph is created by, before the sub-graph is input into the GNN, using taxonomy links to create a merged graph by merging a source language graph that does not comprise taxonomy links with a taxonomy graph and connecting the taxonomy graph with a target language graph.

2. The method of claim 1 , further comprising:

inputting the sub-graph into an encoder of the GNN;

in response to inputting the sub-graph into the encoder, outputting, by the encoder, a source entity embedding;

inputting the source entity embedding and a target entity embedding into a decoder of the GNN;

in response to inputting the source entity embedding and the target entity embedding into the decoder, outputting, by the decoder, the value.

3. The method of claim 2 , further comprising obtaining the different link types from a taxonomy and creating the source entity embedding using at least two type-specific graph convolutional networks (GCNs) that each correspond to a different link type obtained from the taxonomy.

4. The method of claim 1 , further comprising creating the cross-lingual entity graph by extracting the plurality of source entities from data records stored in a source language, creating embedding links between the plurality of source entities and taxonomy entities obtained from a taxonomy, the plurality of links comprise links of different types between the taxonomy entities.

5. The method of claim 1 , further comprising creating the cross-lingual entity graph by extracting the plurality of target entities from data records stored in a target language, and creating embedding links between the plurality of target entities.

6. The method of claim 1 , each different link type indicating a different type of semantic relationship obtained from a taxonomy.

7. The method of claim 1 , the plurality of links comprise within language links (WLLs); the graph neural network trained using both WLLs and ground truth between language links (BLLs).

8. The method of claim 1 , further comprising extracting the source entity from a search query.

9. The method of claim 1 , further comprising extracting the source entity from a data record of a connections network-based system.

10. A machine translation system, comprising:

at least one processor; and

computer memory operably coupled to the at least one processor;

the computer memory comprises instructions that, when executed by the at least one processor, cause the at least one processor to:

configure the computer memory according to a relational graph convolutional network (RGCN),

the RGCN comprises a plurality of graph convolutional sub-networks trained using a cross-lingual entity graph;

the cross-lingual entity graph comprises a first plurality of entities connected by a plurality of within language taxonomy links, a second plurality of entities connected by a plurality of within language embedding links, and a third plurality of entities connected by a plurality of between language links;

the plurality of within language taxonomy links indicate at least two different link types obtained from a taxonomy;

the plurality of within language embedding links indicate semantic similarities between connected entities;

the plurality of between language links each indicate a valid machine translation; and

before the cross-lingual entity graph is used to train the plurality of graph convolutional sub-networks, the cross-lingual entity graph is created by (i) merging a source language semantic graph that does not comprise taxonomy links with the taxonomy and (ii) connecting the taxonomy to a target language graph.

11. The machine translation system of claim 10 , each graph convolutional sub-network trained using a different sub-graph of the cross-lingual entity graph.

12. The machine translation system of claim 11 , each different sub-graph of the cross-lingual entity graph comprising a different particular link type of the at least two different link types obtained from the taxonomy.

13. The machine translation system of claim 10 , the RGCN further comprising an encoder configured to generate entity embeddings using the plurality of within language taxonomy links.

14. The machine translation system of claim 13 , the RGCN further comprising a decoder configured to compute alignment scores that indicate semantic alignment between the entity embeddings and other entity embeddings not computed using the plurality of within language taxonomy links.

15. A method for performing machine translation, comprising:

embedding a source entity in a cross-lingual entity graph stored in computer memory;

identifying a target entity in the cross-lingual entity graph;

the target entity comprises a valid translation of the source entity;

the cross-lingual entity graph comprises a plurality of source entities, a plurality of target entities, a plurality of within language links that each indicate a different link type obtained from a taxonomy, and a plurality of between language links created using entity alignment scores computed by a graph convolutional network (GCN); and

the cross-lingual entity graph is created by (i) merging a source language semantic graph that does not comprise taxonomy links with the taxonomy and (ii) connecting the taxonomy with a target language graph.

16. The method of claim 15 , further comprising extracting the source entity from a search query.

17. The method of claim 15 , further comprising extracting the source entity from a data record of a connections network-based system.

18. The method of claim 15 , further comprising identifying the target entity by determining that a particular type of within language link (WLL) exists between the source entity and a taxonomy entity in the cross-lingual entity graph; and determining that a between language link (BLL) exists between the taxonomy entity and the target entity in the cross-lingual entity graph.

19. The method of claim 15 , further comprising identifying the target entity by iteratively computing alignment scores between the source entity and a plurality of different target entities in the cross-lingual entity graph.

20. The method of claim 15 , further comprising using the plurality of within language links to identify the target entity.