Natural language outputs for path prescriber model simulation for nodes in a time-series network

1 . A method of generating natural language outputs, the method comprising:

accessing a model of a system, wherein the system is represented by a hierarchy of nodes in a data structure, and nodes in the hierarchy of nodes comprise time series of data;

identifying a time series represented by a node in the data structure that will generate a future anomaly;

accessing a template that is selected from a plurality of templates based on a corresponding to a type of the time series, wherein the template comprises semantic tags;

populating the semantic tags in the template using data from the time series;

sending a phrase from the template to a natural language model;

receiving a plurality of similar phrases having a same semantic meaning as the phrase from the natural language model;

selecting one of the plurality of similar phrases and replacing the phrase in the template; and

causing language from the template to be displayed on a display device.

2 . The method of claim 1 , wherein identifying the time series that will generate the future anomaly comprises:

simulating the model of the system to generate a simulated future time series; and

determining that the simulated future time series includes data points that fall outside of a threshold region.

3 . The method of claim 1 , wherein identifying the time series that will generate the future anomaly comprises:

identifying a trend such that the time series increases or decreases in a single direction in an aggregate.

4 . The method of claim 1 , wherein the template comprises semantic tags that are replaced by data points in the time series.

5 . The method of claim 1 , wherein the time series comprises an entity name, a value type, and a plurality of values.

6 . The method of claim 1 , wherein the natural language model comprises a Transformer model.

7 . The method of claim 1 , further comprising:

selecting the plurality of similar phrases from a plurality of output phrases from the natural language model, wherein the plurality of similar phrases are selected based on a being above a threshold.

8 . The method of claim 1 , wherein selecting the one of the plurality of similar phrases and replacing the phrase in the template comprises:

randomly selecting one of the plurality of similar phrases.

9 . A non-transitory computer-readable medium comprising instructions that, when executed by one or more processors, cause the one or more processors to perform operations comprising:

accessing a model of a system, wherein the system is represented by a hierarchy of nodes in a data structure, and nodes in the hierarchy of nodes comprise time series of data;

identifying a time series represented by a node in the data structure that will generate a future anomaly;

accessing a template that is selected from a plurality of templates based on a corresponding to a type of the time series, wherein the template comprises semantic tags;

populating the semantic tags in the template using data from the time series;

sending a phrase from the template to a natural language model;

receiving a plurality of similar phrases having a same semantic meaning as the phrase from the natural language model;

selecting one of the plurality of similar phrases and replacing the phrase in the template; and

causing language from the template to be displayed on a display device.

10 . The non-transitory computer-readable medium of claim 9 , further comprising:

generating a plurality of phrases from the phrase before sending the phrase to the natural language model.

11 . The non-transitory computer-readable medium of claim 10 , wherein the plurality of phrases are generated from the phrase by substituting words in the phrase with synonym words.

12 . The non-transitory computer-readable medium of claim 11 , wherein the plurality of phrases are converted into word vectors.

13 . The non-transitory computer-readable medium of claim 12 , wherein the word vectors are provided as a seed to the natural language model.

14 . A system comprising:

one or more processors; and

one or more memory devices comprising instructions that, when executed by the one or more processors, cause the one or more processors to perform operations comprising:

accessing a model of a system, wherein the system is represented by a hierarchy of nodes in a data structure, and nodes in the hierarchy of nodes comprise time series of data;

identifying a time series represented by a node in the data structure that will generate a future anomaly;

accessing a template that is selected from a plurality of templates based on a corresponding to a type of the time series, wherein the template comprises semantic tags;

populating the semantic tags in the template using data from the time series;

sending a phrase from the template to a natural language model;

receiving a plurality of similar phrases having a same semantic meaning as the phrase from the natural language model;

selecting one of the plurality of similar phrases and replacing the phrase in the template; and

causing language from the template to be displayed on a display device.

15 . The system of claim 14 , wherein the hierarchy of nodes in the data structure comprises a plurality of non-cyclical, linear parent-child relationships.

16 . The system of claim 14 , wherein the operations further comprise:

simplifying the model by removing relationships between the hierarchy of nodes that affect parent nodes less than a threshold amount;

simulating the model to identify a node comprising a time series of data that risks missing a predefined target value;

generating a pathway of actions comprising changes to driver nodes of the node that cause the time series of data to move within a threshold distance of the predefined target value in the future, wherein the template describes the action pathway.

17 . The system of claim 16 , wherein simplifying the model further comprises removing parameters from the model that affect simulated values less than a threshold amount.

18 . The system of claim 17 , wherein simplifying the model further comprises removing non-driver notes from the hierarchy of nodes.

19 . The system of claim 18 , wherein simplifying the model further comprises assigning partial delay equations to relationships between the hierarchy of nodes.

20 . The system of claim 19 , wherein simplifying the model further comprises:

initializing the partial delay equations using domain-specific values; and

assigning default values to partial delay equations without domain-specific values.