Refinement and calibration mechanism for improving classification of information assets
Techniques are described for refining the manual classification of assets classified or categorized using the terms of a business glossary. A semantic refinement mechanism is used to refine the manual classification of such assets, as well as subsequently evaluate the refined asset classifications. Further, the refined asset classifications may be used as a training set for a machine learning classifier. That is, should the classification of an asset contributing to a refinement change, the refinement based on that classification may be undone, at least in some cases.
1. A computer-readable storage medium storing executable instructions for configuring a computing appliance, which, when executed, performs an operation for refining asset classifications, the operation comprising:
receiving a plurality of assets, each asset having a classification of a term, wherein each term is selected from a business glossary which provides a hierarchy of controlled vocabulary of terms used within an organization and wherein each asset is characterized using a set of attributes selected from a domain ontology; and
upon determining a first term assigned to a first one of the assets satisfies a set of refinement criteria, refining the classification of the first asset by assigning the first asset a second term from the business glossary, wherein the second term is more precise in the business glossary than the first term and wherein the refinement criteria includes:
determining that the term of a second one of the assets comprises a descendent of the classification of the first asset, and
determining that each attribute of the first asset is at a lower level in the domain ontology than a corresponding attribute in the second asset.
2. The computer-readable storage medium of claim 1 , wherein the operation further comprises:
storing a reference to the first asset and the second asset.
3. The computer-readable storage medium of claim 2 , wherein the operation further comprises:
upon determining the second asset has been reclassified, reevaluating, by operation of the one or more computer processors, the refined classification assigned to the first asset against the refinement criteria.
4. The computer-readable storage medium of claim 3 , wherein the operation further comprises:
undoing the refined classification assigned to the first asset if the first asset no longer satisfies the set of refinement criteria.
5. The computer-readable storage medium of claim 3 , wherein the operation further comprises:
comparing a weight assigned to a user who assigned the classification to the second asset with a weight assigned to a user who reclassified the second asset;
undoing the refined classification of the second asset if the weight of the user who assigned the classification to the second asset exceeds the weight assigned to the user who reclassified the second by a specified threshold; and
retaining the refined classification of the second asset if the weight of the user who assigned the classification to the second asset does not exceed the weight assigned to the user who reclassified the second by the specified threshold.
6. The computer-readable storage medium of claim 1 , wherein the operation further comprises, training a machine learning classifier based on the classifications assigned to the plurality of assets and further based on the refined classification assigned to the first asset.
7. The computer-readable storage medium of claim 1 , wherein the classification of the terms are assigned by users, wherein each user assigning classifications has an assigned weight, and wherein the operation further comprises determining that a difference of the weights assigned to users who assigned the classifications terms of the first asset and the second asset exceeds a threshold.
8. A computing system, comprising:
a processor;
a memory storing one or more executable components configured to perform an operation for refining asset classifications, the operation comprising:
receiving a plurality of assets, each asset having a classification of a term, wherein each term is selected from a business glossary which provides a hierarchy of controlled vocabulary of terms used within an organization and wherein each asset is characterized using a set of attributes selected from a domain ontology, and
upon determining a first term assigned to a first one of the assets satisfies a set of refinement criteria, refining the classification of the first asset by assigning the first asset a second term from the business glossary, wherein the second term is more precise in the business glossary than the first term and wherein the refinement criteria includes:
determining that the term of a second one of the assets comprises a descendent of the classification of the first asset; and
determining that each attribute of the first asset is at a lower level in the domain ontology than a corresponding attribute in the second asset.
9. The computing system of claim 8 , wherein the operation further comprises:
storing a reference to the first asset and the second asset.
10. The computing system of claim 9 , wherein the operation further comprises:
upon determining the second asset has been reclassified, reevaluating, by operation of the one or more computer processors, the refined classification assigned to the first asset against the refinement criteria.
11. The computing system of claim 10 , wherein the operation further comprises: undoing the refined classification assigned to the first asset if the first asset no longer satisfies the set of refinement criteria.
12. The computing system of claim 10 , wherein the operation further comprises:
comparing a weight assigned to a user who assigned the classification to the second asset with a weight assigned to a user who reclassified the second asset;
undoing the refined classification of the second asset if the weight of the user who assigned the classification to the second asset exceeds the weight assigned to the user who reclassified the second by a specified threshold; and
retaining the refined classification of the second asset if the weight of the user who assigned the classification to the second asset does not exceed the weight assigned to the user who reclassified the second by the specified threshold.
13. The computing system of claim 8 , wherein the operation further comprises, training a machine learning classifier based on the classifications assigned to the plurality of assets and further based on the refined classification assigned to the first asset.
14. The computing system of claim 8 , wherein the classification of the terms are assigned by users, wherein each user assigning classifications has an assigned weight, and wherein the operation further comprises determining that a difference of the weights assigned to users who assigned the classifications terms to the first asset and the second asset exceeds a threshold.