Machine-learning based similarity engine
An embodiment may involve storage containing incident logs and mappings between incident logs and vector representations generated by a machine learning (ML) model. The embodiment may further involve one or more processors configured to: receive, from a client device, a request corresponding to an additional incident log; transmit, to the ML model, additional values as appearing in the additional incident log, wherein reception of the additional values causes the ML model to generate an additional vector representation of the additional incident log; obtain confidence measurements respectively representing similarities between the additional vector representation and each of the vector representations corresponding to the incident logs; determine, based on the confidence measurements, a set of one or more incident logs that are semantically relevant to the additional incident log; and transmit, to the client device, representations of the one or more incident logs and their corresponding confidence measurements.
1. A system comprising: persistent storage containing a plurality of incident logs, wherein mappings between the incident logs and vector representations thereof were generated by a machine learning (ML) model based on respective values in a pre-determined set of fields in the incident logs; and one or more processors configured to: receive, from a client device, a classification request corresponding to an additional incident log; responsive to the classification request, transmit, to the ML model, additional values from the pre-determined set of fields as appearing in the additional incident log, wherein reception of the additional values causes the ML model to generate an additional vector representation of the additional incident log; obtain confidence measurements respectively representing similarities between the additional vector representation and each of the vector representations corresponding to the incident logs; determine, based on the confidence measurements, a set of one or more incident logs that are semantically relevant to the additional incident log; and transmit, to the client device, representations of the set of one or more incident logs and their corresponding confidence measurements.
2. The system of claim 1 , wherein the one or more processors are further configured to:
before receiving the classification request: (i) generate a representation of a graphical user interface, wherein the graphical user interface includes an identifier for the additional incident log, a description of the additional incident log, and an actuatable classification control, and (ii) transmit, to the client device, the representation of the graphical user interface, wherein the actuatable classification control, when actuated by way of the client device, causes transmission of the classification request.
3. The system of claim 1 , wherein transmitting the representations of the set of one or more incident logs and their corresponding confidence measurements comprises: generating a representation of a graphical user interface, wherein the graphical user interface includes a list-based view of the set of one or more incident logs and their corresponding confidence measurements, and wherein the list-based view specifies in line items for each of the one or more incident logs: (i) an identifier, (ii) a description, and (iii) a corresponding confidence measurement; and transmitting, to the client device, the representation of the graphical user interface.
4. The system of claim 1 , wherein the one or more processors are further configured to: receive, from the client device, a selected incident log from the set of one or more incident logs; and store, in the persistent storage, an association between the selected incident log and the additional incident log.
5. The system of claim 4 , wherein the association is a parent-child relationship or a sibling relationship.
6. The system of claim 1 , wherein the persistent storage further includes a plurality of incident solutions, wherein each of the incident solutions includes a further set of fields that define a solution to a problem expressed in one of the incident logs, and wherein the one or more processors are further configured to: determine, based on the representations of the set of one or more incident logs, a set of one or more incident solutions associated with the set of one or more incident logs.
7. The system of claim 6 , wherein the one or more processors are further configured to: generate a representation of a graphical user interface, wherein the graphical user interface includes a list-based view of the representations of the set of one or more incident solutions, and wherein the list-based view specifies in line items for each of the set of one or more incident solutions: (i) a identifier, and (ii) a description; and transmit, to the client device, the representation of the graphical user interface.
8. The system of claim 7 , wherein the list-based view also specifies corresponding confidence measurements in the line items for each of the set of one or more incident solutions.
9. The system of claim 6 , wherein the one or more processors are further configured to: receive, from the client device, a selected incident solution from the set of one or more incident solutions; and store, in the persistent storage, an association between the selected incident solution and the additional incident log.
10. The system of claim 1 , wherein the ML model is based on a word vector model or a paragraph vector model, and wherein training the ML model comprises mapping text within the pre-determined set of fields for each of the incident logs respectively into the vector representations, wherein the vector representations are encoded in an n-dimensional semantic space related to content of the text.
11. The system of claim 1 , wherein obtaining confidence measurements comprises:
calculating respective cosine similarity measurements between the additional vector representation and each of the vector representations corresponding to the incident logs; and
calculating the confidence measurements as multiples of the respective cosine similarity measurements that are greater than zero.
12. The system of claim 1 , wherein each of the incident logs in the plurality of incident logs and the additional incident log comprises a description of an incident relating to a managed network.
13. The system of claim 1 , wherein the one or more processors are disposed within a computational instance of a remote network management platform, wherein the computational instance is dedicated to a managed network associated with the client device, wherein the ML model is disposed within a central instance of the remote network management platform, and wherein the central instance stores further ML models used by further computational instances of the remote network management platform.
14. A computer-implemented method comprising: receiving, from a client device, a classification request corresponding to an additional incident log, wherein persistent storage contains a plurality of incident logs, and wherein mappings between the incident logs and vector representations thereof were generated by a machine learning (ML) model based on respective values in a pre-determined set of fields in the incident logs; responsive to the classification request, transmitting, to the ML model, additional values from the pre-determined set of fields as appearing in the additional incident log, wherein reception of the additional values causes the ML model to generate an additional vector representation of the additional incident log; obtaining confidence measurements respectively representing similarities between the additional vector representation and each of the vector representations corresponding to the incident logs; determining, based on the confidence measurements, a set of one or more incident logs that are semantically relevant to the additional incident log; and transmitting, to the client device, representations of the set of one or more incident logs and their corresponding confidence measurements.
15. The computer-implemented method of claim 14 , further comprising: receiving, from the client device, a selected incident log from the set of one or more incident logs; and storing, in the persistent storage, an association between the selected incident log and the additional incident log.
16. The computer-implemented method of claim 14 , wherein the persistent storage further includes a plurality of incident solutions, wherein each of the incident solutions includes a further set of fields that define a solution to a problem expressed in set of one of the incident logs, and wherein the method further comprises: determining, based on the representations of the set of one or more incident logs, a set of one or more incident solutions associated with the set of one or more incident logs.
17. The computer-implemented method of claim 16 , further comprising: receiving, from the client device, a selected incident solution from the set of one or more incident solutions; and storing, in the persistent storage, an association between the selected incident solution and the additional incident log.
18. The computer-implemented method of claim 14 , wherein the ML model is based on a word vector model or a paragraph vector model, and wherein training the ML model comprises mapping text within the pre-determined set of fields for each of the incident logs respectively into the vector representations, wherein the vector representations are encoded in an n-dimensional semantic space related to content of the text.
19. The computer-implemented method of claim 14 , wherein obtaining confidence measurements comprises:
calculating respective cosine similarity measurements between the additional vector representation and each of the vector representations corresponding to the incident logs; and
calculating the confidence measurements as multiples of the respective cosine similarity measurements that are greater than zero.
20. An article of manufacture including a non-transitory computer-readable medium, having stored thereon program instructions that, upon execution by a computing system, cause the computing system to perform operations comprising: receiving, from a client device, a classification request corresponding to an additional incident log, wherein persistent storage contains a plurality of incident logs, and wherein mappings between the incident logs and vector representations thereof were generated by a machine learning (ML) model based on respective values in a pre-determined set of fields in the incident logs; responsive to the classification request, transmitting, to the ML model, additional values from the pre-determined set of fields as appearing in the additional incident log, wherein reception of the additional values causes the ML model to generate an additional vector representation of the additional incident log; obtaining confidence measurements respectively representing similarities between the additional vector representation and each of the vector representations corresponding to the incident logs; determining, based on the confidence measurements, a set of one or more incident logs that are semantically relevant to the additional incident log; and transmitting, to the client device, representations of the set of one or more incident logs and their corresponding confidence measurements.