Analytical search results for a complex query
A search platform can receive a complex query, and generate raw data in response to the complex query, based on artificial intelligence searching of a semantic search space, relevant to the context of the complex query. The platform can obtain, generate and/or recommend analytical models capable of providing robust answers to the complex query. The platform can augment the semantic search space with analytical models and search the augmented search space for analytical models relevant to the complex query. The answers, provided by the platform to the complex query, can be the result of performing complex analytical workflows.
1 . A computer-implemented method, performed by one or more processors of a search platform, comprising:
registering, via an interface of the search platform, multiple kernels and a workflow that performs a processing pipeline by executing the registered multiple kernels, wherein one or more kernels include data constraints;
receiving a complex query via the search platform;
semantically searching a context of the complex query, and generating a semantic search space;
searching the semantic search space and retrieving raw data relevant to the complex query;
obtaining the one or more kernels relevant to the context of the complex query;
processing the obtained the one or more kernels with a natural language processing (NLP) pipeline, generating summaries of objectives of the obtained kernels;
determining relevancy context of the obtained the one or more kernels, based at least in part on the summaries;
tagging the obtained the one or more kernels with one more relevancy context;
generating an augmented search space, the augmented search space comprising the semantic search space and the workflow;
embedding the obtained the one or more kernels into the augmented search space, using at least in part the relevancy context tags; and
searching the augmented search space, at least in part, based on machine learning matching of the context of the complex query with one or more relevancy tags;
controlling by a mesh runner the performance of the workflow by executing the obtained the one or more kernels using the raw data or a portion of the raw data as input into one or more of the executed the obtained the one or more kernels; and
generating results comprising an answer to the complex query based on the performance of the workflow.
2 . The computer-implemented method of claim 1 , further comprising:
prior to executing the workflow, transforming, by a processing module, the raw data into a data format compatible with the data constraints.
3 . The computer-implemented method of claim 1 , wherein the workflow is determined to be performed is based on the raw data, and wherein the executed mesh runner matches data constraints of the obtained the one or more kernel to the raw data.
4 . The computer-implemented method of claim 1 , further comprising:
obtaining the workflow for generating the results, the workflow comprising the obtained the one or more kernels, each kernel comprising at least one analytical model;
determining input and output data constraints of each of the obtained the one or more kernels;
determining portions of the raw data compatible with the input data constraints of the obtained the one or more kernels in the workflow;
transforming output of the obtained the one or more kernels, at least in part based on the input data constraint of another kernel;
executing the workflow, comprising executing the obtained the one or more kernels and the models therein; and
generating the results.
5 . The computer-implemented method of claim 1 , wherein the obtained the one or more kernels are connected, via the interface, to form a directed computational graph of the multiple kernels.
6 . A non-transitory computer storage medium that stores executable program instructions that, when executed by one or more computing devices, configure the one or more computing devices to perform operations comprising:
registering, via an interface of a search platform, multiple kernels and a workflow that performs a processing pipeline by executing the registered multiple kernels, wherein one or more kernels include data constraints;
receiving a complex query via the search platform;
semantically searching a context of the complex query, and generating a semantic search space;
searching the semantic search space and retrieving raw data relevant to the complex query;
obtaining the one or more kernels relevant to the context of the complex query;
processing the obtained the one or more kernels with a natural language processing (NLP) pipeline, generating summaries of objectives of the obtained kernels;
determining relevancy context of the obtained the one or more kernels, based at least in part on the summaries;
tagging the obtained the one or more kernels with one more relevancy context;
generating an augmented search space, the augmented search space comprising the semantic search space and the workflow;
embedding the obtained the one or more kernels into the augmented search space, using at least in part the relevancy context tags; and
searching the augmented search space, at least in part, based on machine learning matching of the context of the complex query with one or more relevancy tags;
controlling by a mesh runner the performance of the workflow by executing the obtained the one or more kernels using the raw data or a portion of the raw data as input into one or more of the executed the obtained the one or more kernels; and
generating results comprising an answer to the complex query based on the performance of the workflow.
7 . The non-transitory computer storage of claim 6 , wherein the operations further comprise:
prior to executing the workflow, transforming, by a processing module, the raw data into a data format compatible with the data constraints.
8 . The non-transitory computer storage of claim 6 , wherein the operations further comprise: wherein the workflow is determined to be performed is based on the raw data, and wherein the executed mesh runner matches data constraints of the obtained the one or more kernel to the raw data.
9 . The non-transitory computer storage of claim 6 , wherein the operations further comprise:
obtaining the workflow for generating the results, the workflow comprising the obtained the one or more kernels, each kernel comprising at least one analytical model;
determining input and output data constraints of each of the obtained the one or more kernels;
determining portions of the raw data compatible with the input data constraints of the obtained the one or more kernels in the workflow;
transforming output of the obtained the one or more kernels, at least in part based on the input data constraint of another kernel;
executing the workflow, comprising executing the obtained the one or more kernels and the models therein; and
generating the results.
10 . The non-transitory computer storage of claim 6 , wherein the multiple kernels are connected, via the interface, to form a directed computational graph of the obtained the one or more kernels.
11 . A system comprising a search platform having one or more processors, wherein the one or more processors are configured to perform operations comprising:
registering, via an interface of the search platform, multiple kernels and a workflow that performs a processing pipeline by executing the registered multiple kernels, wherein one or more kernels include data constraints;
receiving a complex query via the search platform;
semantically searching a context of the complex query, and generating a semantic search space;
searching the semantic search space and retrieving raw data relevant to the complex query;
obtaining the one or more kernels relevant to the context of the complex query;
processing the obtained the one or more kernels with a natural language processing (NLP) pipeline, generating summaries of objectives of the obtained kernels;
determining relevancy context of the obtained the one or more kernels, based at least in part on the summaries;
tagging the obtained the one or more kernels with one more relevancy context;
generating an augmented search space, the augmented search space comprising the semantic search space and the workflow;
embedding the obtained the one or more kernels into the augmented search space, using at least in part the relevancy context tags; and
searching the augmented search space, at least in part, based on machine learning matching of the context of the complex query with one or more relevancy tags;
controlling by a mesh runner the performance of the workflow by executing the obtained the one or more kernels using the raw data or a portion of the raw data as input into one or more of the executed the obtained the one or more kernels; and
generating results comprising an answer to the complex query based on the performance of the workflow.
12 . The system of claim 11 , wherein the operations performed by the one or more processors further comprise:
prior to executing the workflow, transforming, by a processing module, the raw data into a data format compatible with the data constraints.
13 . The system of claim 11 , wherein the workflow is determined to be performed is based on the raw data, and wherein the executed mesh runner matches data constraints of the obtained the one or more kernel to the raw data.
14 . The system of claim 11 , wherein the operations performed by the one or more processors further comprise:
obtaining the workflow for generating the results, the workflow comprising the obtained the one or more kernels, each kernel comprising at least one analytical model;
determining input and output data constraints of each of the obtained the one or more kernels;
determining portions of the raw data compatible with the input data constraints of the obtained the one or more kernels in the workflow;
transforming output of the obtained the one or more kernels, at least in part based on the input data constraint of another kernel;
executing the workflow, comprising executing the obtained the one or more kernels and the models therein; and
generating the results.