Semantic parsing using encoded structured representation

1. A method of semantic parsing, the method comprising:

receiving an input comprising a plurality of words;

generating a structured representation of the plurality of words, wherein generating the structured representation comprises generating a graph including semantic information and syntactic information of the plurality of words;

encoding the structured representation into a latent embedding space, wherein encoding the structured representation comprises:

generating bi-directional node embeddings for each node of a plurality of nodes of the graph; and

generating a graph embedding from the bi-directional node embeddings; and

decoding the encoded structured representation from the latent embedding space into a logical representation of the plurality of words.

2. The method of claim 1 , wherein generating the graph comprises:

determining a word sequence of the plurality of words; and

determining at least one of a dependency parse tree, constituency parse tree, and an abstract meaning representation graph that is referenced to the word sequence.

3. The method of claim 1 , wherein generating the graph embedding comprises one of:

performing a max-pooling operation on the node embeddings; and

performing an attention-based aggregation on the node embeddings.

4. The method of claim 1 , wherein decoding the encoded structured representation comprises:

determining, using an attention mechanism, alignments between the graph and the logical representation.

5. The method of claim 1 , further comprising:

displaying the input and the logical representation.

6. The method of claim 1 , wherein generating the bi-directional node embeddings comprises:

for each node of the plurality of nodes, categorizing each neighboring node of the node, as one of a forward neighbor and a backward neighbor;

aggregating, over a plurality of iterations, the forward neighbors corresponding to the plurality of nodes into a first vector; and

aggregating, over the plurality of iterations, the backward neighbors corresponding to the plurality of nodes into a second vector.

7. The method of claim 6 , wherein generating the bi-directional node embeddings further comprises:

for each iteration of the plurality of iterations:

concatenating a current forward representation of the node with a first neighborhood vector to provide a first concatenated vector; and

concatenating a current backward representation of the node with a second neighborhood vector to provide a second concatenated vector.

8. The method of claim 7 , wherein generating the bi-directional node embeddings further comprises:

updating the current forward representation of the node by feeding the first concatenated vector into a fully connected layer having a non-linear activation function; and

updating the current backward representation of the node by feeding the second concatenated vector into the fully connected layer.

9. The method of claim 8 ,

wherein updating the current forward representation of the node and updating the current forward representation of the node occur at each iteration of the plurality of iterations prior to a final iteration, and

wherein generating the bi-directional node embeddings further comprises:

at the final iteration, concatenating a final forward representation of the node with a final backward representation of the node.

10. The method of claim 6 ,

wherein, for each iteration of the plurality of iterations, a distinct aggregator is used when aggregating the forward neighbors and when aggregating the backward neighbors.

11. The method of claim 1 , wherein generating the graph embedding comprises:

aggregating the bi-directional node embeddings using one of: an element-wise max-pooling operation, or an attention-based aggregation function.

12. A system comprising:

one or more computer processors configured to perform an operation for semantic parsing, the operation comprising:

receiving an input comprising a plurality of words;

generating a structured representation of the plurality of words, wherein generating the structured representation comprises generating a graph including semantic information and syntactic information of the plurality of words;

encoding the structured representation into a latent embedding space, wherein encoding the structured representation comprises:

generating bi-directional node embeddings for each node of a plurality of nodes of the graph; and

generating a graph embedding from the bi-directional node embeddings; and

decoding the encoded structured representation from the latent embedding space into a logical representation of the plurality of words.

13. The system of claim 12 , wherein generating the graph comprises:

determining a word sequence of the plurality of words; and

determining at least one of a dependency parse tree, constituency parse tree, and an abstract meaning representation graph that is referenced to the word sequence.

14. The system of claim 12 , wherein generating the graph embedding comprises one of:

performing a max-pooling operation on the node embeddings; and

performing an attention-based aggregation on the node embeddings.

15. The system of claim 12 , wherein decoding the encoded structured representation comprises:

determining, using an attention mechanism, alignments between the graph and the logical representation.

16. The system of claim 12 , further comprising:

displaying the input and the logical representation.

17. A computer program product comprising:

a computer-readable storage medium having computer-readable program code embodied therewith, the computer-readable program code executable by one or more computer processors to perform an operation for semantic parsing, the operation comprising:

receiving an input comprising a plurality of words;

generating a structured representation of the plurality of words, wherein generating the structured representation comprises generating a graph including semantic information and syntactic information of the plurality of words;

encoding the structured representation into a latent embedding space, wherein encoding the structured representation comprises:

generating bi-directional node embeddings for each node of a plurality of nodes of the graph; and

generating a graph embedding from the bi-directional node embeddings; and

decoding the encoded structured representation from the latent embedding space into a logical representation of the plurality of words.

18. The computer program product of claim 17 , wherein generating the graph comprises:

determining a word sequence of the plurality of words; and

determining at least one of a dependency parse tree, constituency parse tree, and an abstract meaning representation graph that is referenced to the word sequence.

19. The computer program product of claim 17 , wherein generating the graph embedding comprises one of:

performing a max-pooling operation on the node embeddings; and

performing an attention-based aggregation on the node embeddings.

20. The computer program product of claim 17 , wherein decoding the encoded structured representation comprises:

determining, using an attention mechanism, alignments between the graph and the logical representation.