Generative knowledge search engine for multi-query enabled network knowledge completion

1 . A method comprising:

obtaining at least one search query and inventory data that includes information about a plurality of enterprise assets and configuration of an enterprise network;

generating a contextual schema based on the inventory data, wherein the contextual schema includes a plurality of query sub-graphs indicative of an intention of the at least one search query and are generated based on one or more latent correlations among the plurality of enterprise assets indicated in the inventory data, and wherein the plurality of query sub-graphs include meta-paths representing relations between portions of the at least one search query obtained from the inventory data;

generating a plurality of prediction graphs by machine learning models expanding the plurality of query sub-graphs, wherein the plurality of prediction graphs include additional nodes representing domain knowledge data connected to leaf nodes of the plurality of query sub-graphs and reconstructed meta-paths, and wherein the additional nodes and reconstructed meta-paths are determined based on learned relationships and network domain knowledge data;

generating a solution graph by merging the plurality of prediction graphs; and

providing a response to the at least one search query based on the solution graph, wherein the response is specific to the enterprise network.

2 . The method of claim 1 , wherein the at least one search query includes a multi-query search inquiry, and generating the contextual schema includes:

generating an enterprise asset graph based on the inventory data, wherein the enterprise asset graph includes one or more relationships among the plurality of enterprise assets in the inventory data; and

pruning the multi-query search inquiry based on the enterprise asset graph to generate the plurality of query sub-graphs specific to the enterprise network.

3 . The method of claim 2 , wherein the multi-query search inquiry includes a plurality of search queries that are input as one search by a user and further comprising:

determining a persona of the user,

wherein pruning the multi-query search inquiry is further based on the persona of the user.

4 . The method of claim 1 , wherein the at least one search query includes a multi-query search inquiry, and generating the contextual schema includes:

generating a plurality of search nodes based on a plurality of queries in the multi-query search inquiry;

determining one or more relationships between the plurality of search nodes based on the inventory data; and

generating the plurality of query sub-graphs each indicative of the one or more relationships between the plurality of search nodes in the inventory data.

5 . The method of claim 1 , wherein generating the solution graph includes:

applying a pre-trained artificial intelligence model to the network domain knowledge data to generate the solution graph based on the plurality of query sub-graphs.

6 . The method of claim 1 , wherein generating the plurality of prediction graphs includes:

generating the plurality of prediction graphs by training the plurality of query sub-graphs using a plurality of graph generative pre-trained transformers.

7 . The method of claim 1 , wherein generating the solution graph includes:

merging the plurality of prediction graphs based on one or more common patterns to generate the solution graph.

8 . The method of claim 7 , further comprising:

generating the response that includes an end-to-end solution to each question in a multi-query search inquiry.

9 . The method of claim 1 , wherein providing the response includes:

configuring at least one network asset in the enterprise network based on the solution graph.

10 . The method of claim 1 , wherein providing the response includes:

providing one or more network solution recommendations that identify one or more enterprise assets and suggested network related configurations for the enterprise network.

11 . An apparatus comprising:

a memory;

a network interface configured to enable network communications; and

a processor, wherein the processor is configured to perform a method comprising:

obtaining at least one search query and inventory data that includes information about a plurality of enterprise assets and configuration of an enterprise network;

generating a contextual schema based on the inventory data, wherein the contextual schema includes a plurality of query sub-graphs indicative of an intention of the at least one search query and are generated based on one or more latent correlations among the plurality of enterprise assets indicated in the inventory data, and wherein the plurality of query sub-graphs include meta-paths representing relations between portions of the at least one search query obtained from the inventory data;

generating a plurality of prediction graphs by machine learning models expanding the plurality of query sub-graphs, wherein the plurality of prediction graphs include additional nodes representing domain knowledge data connected to leaf nodes of the plurality of query sub-graphs and reconstructed meta-paths, and wherein the additional nodes and reconstructed meta-paths are determined based on learned relationships and network domain knowledge data;

generating a solution graph by merging the plurality of prediction graphs; and

providing a response to the at least one search query based on the solution graph, wherein the response is specific to the enterprise network.

12 . The apparatus of claim 11 , wherein the at least one search query includes a multi-query search inquiry, and the processor is configured to generate the contextual schema by:

generating an enterprise asset graph based on the inventory data, wherein the enterprise asset graph includes one or more relationships among the plurality of enterprise assets in the inventory data; and

pruning the multi-query search inquiry based on the enterprise asset graph to generate the plurality of query sub-graphs specific to the enterprise network.

13 . The apparatus of claim 12 , wherein the multi-query search inquiry includes a plurality of search queries that are input as one search by a user and the method further comprises:

determining a persona of the user,

wherein the processor is further configured to prune the multi-query search inquiry based on the persona of the user.

14 . The apparatus of claim 11 , wherein the at least one search query includes a multi-query search inquiry, and the processor is configured to generate the contextual schema by:

generating a plurality of search nodes based on a plurality of queries in the multi-query search inquiry;

determining one or more relationships between the plurality of search nodes based on the inventory data; and

generating the plurality of query sub-graphs each indicative of the one or more relationships between the plurality of search nodes in the inventory data.

15 . The apparatus of claim 11 , wherein the processor is configured to generate the solution graph by:

applying a pre-trained artificial intelligence model to the network domain knowledge data to generate the solution graph based on the plurality of query sub-graphs.

16 . The apparatus of claim 11 , wherein the processor is configured to generate the plurality of prediction graphs by:

training the plurality of query sub-graphs using a plurality of graph generative pre-trained transformers.

17 . One or more non-transitory computer readable storage media encoded with software comprising computer executable instructions that, when executed by a processor, cause the processor to perform a method including:

obtaining at least one search query and inventory data that includes information about a plurality of enterprise assets and configuration of an enterprise network;

generating a contextual schema based on the inventory data, wherein the contextual schema includes a plurality of query sub-graphs indicative of an intention of the at least one search query and are generated based on one or more latent correlations among the plurality of enterprise assets indicated in the inventory data, and wherein the plurality of query sub-graphs include meta-paths representing relations between portions of the at least one search query obtained from the inventory data;

generating a plurality of prediction graphs by machine learning models expanding the plurality of query sub-graphs, wherein the plurality of prediction graphs include additional nodes representing domain knowledge data connected to leaf nodes of the plurality of query sub-graphs and reconstructed meta-paths, and wherein the additional nodes and reconstructed meta-paths are determined based on learned relationships and network domain knowledge data;

generating a solution graph by merging the plurality of prediction graphs; and

providing a response to the at least one search query based on the solution graph, wherein the response is specific to the enterprise network.

18 . The one or more non-transitory computer readable storage media according to claim 17 , wherein the at least one search query includes a multi-query search inquiry, and the computer executable instructions cause the processor to generate the contextual schema by:

generating an enterprise asset graph based on the inventory data, wherein the enterprise asset graph includes one or more relationships among the plurality of enterprise assets in the inventory data; and

pruning the multi-query search inquiry based on the enterprise asset graph to generate the plurality of query sub-graphs specific to the enterprise network.

19 . The method of claim 1 , wherein each of the plurality of query sub-graphs is a customized query graph to which at least two meta-paths are applied that represent respective relational hypotheses.

20 . The method of claim 1 , wherein the plurality of query sub-graphs are search nodes that are expanded to include the one or more latent correlations based on the inventory data.