Semantic concept matching using enhanced weak supervision

1. A method for augmenting data labels of a machine learning task, the method comprising:

applying at least one labeling function for each of a plurality of pairs of semantic concepts and producing a labeling matrix;

computing pairwise similarity scores using similarity metrics for each of the pairs of semantic concepts;

augmenting the labeling matrix using the pairwise similarity scores to increase a density of the labeling matrix; and

inputting the labeling matrix to a label aggregator to apply a single label for each of the pairs of semantic concepts labeled by the at least one labeling function.

2. The method of claim 1 , further comprising:

computing a set of similarity scores between the pairwise similarity scores for each of the pairs of semantic concepts;

computing a single similarity score between each two pairs of semantic concepts, using the set of similarity scores respective to each of the pairs of semantic concepts; and

projecting a label from a labeled one of the semantic pairs to an unlabeled one of the semantic pairs in a case that the single similarity score is above a predetermined threshold.

3. The method of claim 1 , further comprising training a discriminative machine learning model with the labeling matrix.

4. The method of claim 1 , further comprising using a discriminative machine learning model that was trained with the labeling matrix to match semantic concepts.

5. The method of claim 1 , wherein the pairs of semantic concepts comprise heterogeneous data sources of a geographical region and wherein the method further comprises identifying a hazardous area.

6. The method of claim 1 , wherein the pairs of semantic concepts comprise data from independent parking spot providers, and wherein the method further comprises producing a homogenized data layer comprising parking spot availability information.

7. The method of claim 1 , wherein the pairs of semantic concepts comprise data from an existing geographic data source with known metadata, and wherein the method further comprises:

creating mapped data for a digital twin of a smart city;

creating subsequent pairs of semantic concepts comprising new data sources;

applying the at least one labeling function for the subsequent pairs of semantic concepts; and

enriching the mapped data for the digital twin of the smart city using the subsequent pairs of semantic concepts.

8. The method of claim 1 , wherein the pairs of semantic concepts comprise data from heterogeneous data sources of energy consuming systems in an area, and wherein the method further comprises producing a homogenized data layer queryable by an energy management application.

9. The method of claim 1 , wherein:

at least two labeling functions label a single pair of the pairs of semantic concepts;

the label aggregator aggregates all labels applied to the single pair of semantic concepts; and

the single pair of semantic concepts is labeled with the label applied most frequently to the single pair of semantic concepts.

10. The method of claim 1 , wherein the at least one labeling function abstains from applying a label to a first pair of the semantic concepts, and applies a label to a second pair of the semantic concepts, and wherein the method further comprises:

calculating a deterministic or probabilistic value using a distance between the first pair of semantic concepts and the second pair of semantic concepts; and

labeling the first pair of semantic concepts using the deterministic or probabilistic value, based on the deterministic or probabilistic value being greater than a predetermined threshold.

11. The method of claim 1 , wherein the set of similarity scores is a summation of a finite set of the pairwise similarities that is computable to a definite number.

12. The method of claim 1 , wherein the label aggregator creates a training dataset for training a discriminative machine learning model.

13. The method of claim 1 , wherein the step of computing similarity scores using similarity metrics for each of the pairs of semantic concepts further comprises:

creating a first pair of semantic concepts from a first concept and a second concept;

computing a first similarity score using the similarity metrics for the first pair of semantic concepts;

creating a second pair of semantic concepts from a third concept and a fourth concept;

computing a second similarity score using the similarity metrics for the second pair of semantic concepts;

computing a third similarity score using the similarity metrics for the first concept of the first pair of semantic concepts and the third concept of the second pair of semantic concepts; and

computing a fourth similarity score using the similarity metrics for the second concept of the first pair of semantic concepts and the fourth concept of the second pair of semantic concepts.

14. A system for augmenting data labels of a machine learning task comprising one or more hardware processors which, alone or in combination, are configured to provide for execution of the following steps:

applying at least one labeling function for each of a plurality of pairs of semantic concepts and producing a labeling matrix;

computing pairwise similarity scores using similarity metrics for each of the pairs of semantic concepts;

augmenting the labeling matrix using the pairwise similarity scores to increase a density of the labeling matrix; and

inputting the labeling matrix to a label aggregator to apply a single label for each of the pairs of semantic concepts labeled by the at least one labeling function.

15. A tangible, non-transitory computer-readable medium having instructions thereon which, upon being executed by one or more hardware processors, alone or in combination, provide for execution of the following steps:

applying at least one labeling function for each of a plurality of pairs of semantic concepts and producing a labeling matrix;

computing pairwise similarity scores using similarity metrics for each of the pairs of semantic concepts;

augmenting the labeling matrix using the pairwise similarity scores to increase a density of the labeling matrix; and

inputting the labeling matrix to a label aggregator to apply a single label for each of the pairs of semantic concepts labeled by the at least one labeling function.