Semantic concept discovery over event databases

1. A computer-implemented concept discovery method, the method comprising:

preparing a concept index for concepts built over a set of input data comprising structured data having input terms, the preparing being based on a payment and performing a minimal curation by a lightweight mapping of known entities and linking the known entities using a Uniform Resource Indicator (URI) until a cost of a running of the minimal curation meets the payment;

building a vector representation of the concepts in the input data where every value in the input data is transformed into an embedding vector by a virtual document creation process that turns each row in the structured data into a context in a corpus of text wherein values in different columns in structured data describe a different feature;

receiving a set of query terms as an additional input;

mapping the set of query terms to the concepts in the concept index;

calculating:

a co-occurrence score for each of the concepts in the concept index by measuring their frequency of co-occurrence with the input terms' concepts, the calculation relying on an efficient index to measure a level of co-occurrence of concepts in a collection of events in the structured data and uses this as a measure of relevance; and

a similarity score for each of the concepts in the concept index by measuring the similarity of their vector representations according to a vector similarity measure that uses the values in the different columns in the structure data; and

ranking the concepts with respect to their relevance to the input terms by a combination of the co-occurrence score and the similarity score,

wherein the set of input data is prepared by a common ingestion pipeline operating on a cluster in a cloud node, without interaction with an existing database, for both event databases and knowledge sources and the common ingestion pipeline crawls remote sources, cleaning invalid records and applying filters, and then storing a result as the concept index.

2. The computer-implemented method of claim 1 , wherein the input data comprises event databases and knowledge bases.

3. The computer-implemented method of claim 1 , further comprising:

receiving a natural language question as a further input; and

extracting query terms from the question.

4. The computer-implemented method of claim 1 , further comprising:

receiving at least one of an image and a video as a further input; and

extracting query terms using at least one of optical character recognition (OCR), speech recognition, examination of captions, and machine recognition of objects and people.

5. The computer-implemented method of claim 1 , further comprising measuring a relatedness of the concept index to the query terms.

6. The computer-implemented method of claim 1 , embodied in a cloud-computing environment.

7. A computer program product for concept discovery, the computer program product comprising a computer-readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform:

preparing a concept index for concepts built over a set of input data comprising structured data having input terms, the preparing being based on a payment and performing a minimal curation by a lightweight mapping of known entities and linking the known entities using a Uniform Resource Indicator (URI) until a cost of a running of the minimal curation meets the payment;

building a vector representation of the concepts in the input data where every value in the input data is transformed into an embedding vector by a virtual document creation process that turns each row in the structured data into a context in a corpus of text wherein values in different columns in structured data describe a different feature;

receiving a set of query terms as an additional input;

mapping the set of query terms to the concepts in the concept index;

calculating:

a co-occurrence score for each of the concepts in the concept index by measuring their frequency of co-occurrence with the input terms' concepts, the calculation relying on an efficient index to measure a level of co-occurrence of concepts in a collection of events in the structured data and uses this as a measure of relevance; and

a similarity score for each of the concepts in the concept index by measuring the similarity of their vector representations according to a vector similarity measure that uses the values in the different columns in the structure data; and

ranking the concepts with respect to their relevance to the input terms by a combination of the co-occurrence score and the similarity score,

wherein the set of input data is prepared by a common ingestion pipeline operating on a cluster in a cloud node, without interaction with an existing database, for both event databases and knowledge sources and the common ingestion pipeline crawls remote sources, cleaning invalid records and applying filters, and then storing a result as the concept index.

8. The computer program product of claim 7 , wherein the input data comprises event databases and knowledge bases.

9. The computer program product of claim 7 , further comprising:

receiving a natural language question as a further input; and

extracting query terms from the question.

10. The computer program product of claim 7 , further comprising:

receiving at least one of an image and a video as a further input; and

extracting query terms using at least one of optical character recognition (OCR), speech recognition, examination of captions, and machine recognition of objects and people.

11. The computer program product of claim 7 , further comprising measuring a relatedness of the concept index to the query terms.

12. A concept discovery system through augmented reality, the system comprising:

a processor; and

a memory, the memory storing instructions to cause the processor to perform:

preparing a concept index for concepts built over a set of input data comprising structured data having input terms, the preparing being based on a payment and performing a minimal curation by a lightweight mapping of known entities and linking the known entities using a Uniform Resource Indicator (URI) until a cost of a running of the minimal curation meets the payment;

building a vector representation of the concepts in the input data where every value in the input data is transformed into an embedding vector by a virtual document creation process that turns each row in the structured data into a context in a corpus of text wherein values in different columns in structured data describe a different feature;

receiving a set of query terms as an additional input;

mapping the set of query terms to the concepts in the concept index;

calculating:

a co-occurrence score for each of the concepts in the concept index by measuring their frequency of co-occurrence with the input terms' concepts, the calculation relying on an efficient index to measure a level of co-occurrence of concepts in a collection of events in the structured data and uses this as a measure of relevance; and

a similarity score for each of the concepts in the concept index by measuring the similarity of their vector representations according to a vector similarity measure that uses the values in the different columns in the structure data; and

ranking the concepts with respect to their relevance to the input terms by a combination of the co-occurrence score and the similarity score,

wherein the set of input data is prepared by a common ingestion pipeline operating on a cluster in a cloud node, without interaction with an existing database, for both event databases and knowledge sources and the common ingestion pipeline crawls remote sources, cleaning invalid records and applying filters, and then storing a result as the concept index.

13. The system of claim 12 , embodied in a cloud-computing environment.