Techniques for real-time generation of temporal comparative and superlative analytics in natural language for real-time dynamic data analytics

• US 9,965,514 B2
• Filed: 12/19/2014
• Issued: 05/08/2018
• Est. Priority Date: 12/19/2014
• Status: Active Grant

• Current Assignee: Software AG USA Incorporated (Software AG)
• Original Assignee: Software AG USA Incorporated (Software AG)
• Inventors: Crupi, John, Thope, Karthic
• Primary Examiner(s): Leroux, Etienne
• Application Number:
  US14/576,413
• Publication Number:
  US 20160179922A1
• Time in Patent Office: 1,236 Days
• CPC Class Codes: G06F 16/243 Natural language query formulation
• CPC Class Codes: G06F 16/2455 Query execution

Abstract

Certain example embodiments relate to techniques for real-time generation of temporal comparative and superlative analytics. For instance, certain example embodiments analyze real-time events to generate building blocks that are assemblable into natural language question templates. These natural language question templates, which may include natural language phrases and templated fields, may be comparative or superlative in nature. User selectable options or parameters may be inserted in place of the templated fields, and complete questions, including user selections and the comparative or superlative, may be translated into a formal query language for execution on the real-time data source. In certain example embodiments, a GUI may enable users to build pre-constructed questions with selectable parameters.

Claims

1. A method of forming a natural language query template, the method comprising:

obtaining a sample of real-time events from a data stream;

identifying, from data in the events in the sample, using processing resources including at least one hardware processor, measures and dimensions associated with that data;

classifying, using the processing resources, the identified measures and dimensions as belonging to one or more distinct measures and one or more distinct dimensions, respectively;

selecting at least one of the one or more distinct measures and/or at least one of the one or more distinct dimensions for inclusion in the natural language query template, the selecting being performed by the processing resources automatically, the natural language query template including natural language expressions and templated fields, at least one of the templated fields enabling user selection of one of a comparative and a superlative; and

building, using the processing resources, the natural language query template by arranging the at least one selected distinct measure and/or the at least one selected distinct dimension in the natural language query template as user-selectable options in at least some of the templated fields;

wherein a natural language query is buildable from the natural language query template and specified user-selectable options, and wherein the natural language query, in turn, is transformable into a formalized query executable on the data stream.

2. The method of claim 1, wherein:

the identifying further identifies timestamps associated with the events in the sample; and

the classifying further identifies one or more timeframes relevant to the events in the sample based on the identified timestamps.

3. The method of claim 2, wherein the one or more timeframes include one or more time windows in or over which the formalized query is to be run, the one or more time windows being included as respective user-selectable options in one or more of the templated fields.

4. The method of claim 1, wherein the selecting identifies measures that have a high rate of change within a predefined time window of interest.

5. The method of claim 1, wherein the selecting identifies those measures whose number of unique values divided by the total number of events in the sample that include that respective measure exceed a predetermined threshold.

6. The method of claim 5, wherein the predetermined threshold varies based on the data type of the values.

7. The method of claim 5, wherein values that are within a predetermined distance of one another are treated as being non-unique.

8. The method of claim 1, wherein the selecting identifies dimensions by testing for conformance with well-known patterns for commonly tested dimensions.

9. The method of claim 1, wherein the natural language query template is built in connection with at least one question part that has an associated mapping, the mapping specifying for its associated question part:

whether the associated question part is one of a comparative or superlative type question part,

what natural language text is to be included in the associated question part,

whether the associated question part is allowed to vary over time, and

an indication as to what function(s) is/are to be performed on a value associated with the associated question part.

10. The method of claim 9, wherein each said mapping further specifies for its associated question part a type of variable to be used with the associated question part, variable types for the question parts being one of time-based and value-based variable types.

11. The method of claim 9, wherein additional functions are user-definable.

12. A method of forming a natural language query template, the method comprising:

obtaining a sample of real-time events from a data stream;

identifying, from the events in the sample, using processing resources including at least one hardware processor, measures and dimensions associated therewith;

classifying, using the processing resources, the identified measures and dimensions as belonging to one or more distinct measures and one or more distinct dimensions, respectively;

selecting, using the processing resources, at least one of the one or more distinct measures and/or at least one of the one or more distinct dimensions for inclusion in the natural language query template, the natural language query template including natural language expressions and templated fields, at least one of the templated fields enabling user selection of one of a comparative and a superlative; and

arranging, using the processing resources, the at least one selected distinct measure and/or the at least one selected distinct dimension in the natural language query template as user-selectable options in at least some of the templated fields;

wherein the natural language query template with specified user-selectable options is transformable into a formalized query executable on the data stream; and

wherein the obtaining of the sample comprises;

(a) obtaining an initial sample having a predetermined sample size, each said real-time event in the initial sample having an associated timestamp;

(b) identifying all unique timestamps in the initial sample;

(c) determining whether the number of unique timestamps in the initial sample size divided by the number of real-time events in the initial sample meets or exceeds a predetermined threshold value;

(d) in response to a determination that the number of unique timestamps in the initial sample size divided by the number of real-time events in the initial sample meets or exceeds the predetermined threshold value, treating the initial sample as the sample; and

(e) in response to a determination that the number of unique timestamps in the initial sample size divided by the number of real-time events in the initial sample does not meet or exceed the predetermined threshold value, increasing the predetermined sample size and repeating (a)-(e), as needed.

13. The method of claim 12, wherein each increase of the predetermined sample size doubles the then-current predetermined sample size.

14. The method of claim 12, wherein each increase of the predetermined sample size increases the then-current predetermined sample size in a predefined manner, unless the then-current predetermined sample size exceeds a second threshold value lower than the predetermined threshold value and the then-current predetermined sample size exceeds a preset threshold amount, in which case the sample is considered complete.

15. The method of claim 12, wherein (b) treats timestamps within a given proximity to one another as being non-unique.

16. An event processing system, comprising:

an event channel configured to receive real-time events from one or more computing systems;

a non-transitory computer readable storage medium; and

processing resources including at least one processor and a memory configured to control the system to at least;

obtain a sample of real-time events from the event channel;

identify, from data in the events in the sample, measures and dimensions associated with that data;

classify the identified measures and dimensions as belonging to one or more distinct measures and one or more distinct dimensions, respectively;

select at least one of the one or more distinct measures and/or at least one of the one or more distinct dimensions for inclusion in a natural language query template that includes natural language expressions and templated fields, at least one of the templated fields enabling user selection of one of a comparative and a superlative;

arrange the at least one selected distinct measure and/or the at least one selected distinct dimension in the natural language query template as user-selectable options in at least some of the templated fields;

responsive to the arranging, store to the non-transitory computer readable storage medium the natural language query template in association with the arranged at least one selected distinct measure and/or the arranged at least one selected distinct dimension; and

enable the natural language query template with specified user-selectable options to be transformed into a formalized query executable on the event channel.

17. The system of claim 16, wherein the processing resources are further configured to control the system to store the formalized query to the non-transitory computer readable storage medium.

18. The system of claim 16, further comprising a complex event processing (CEP) engine configured to executed the formalized query.

19. The system of claim 16, wherein:

the identifying further identifies timestamps associated with the events in the sample; and

the classifying further identifies one or more timeframes relevant to the events in the sample based on the identified timestamps.

20. The system of claim 16, wherein the obtaining of the sample comprises:

(a) obtaining an initial sample having a predetermined sample size, each said real-time event in the initial sample having an associated timestamp;

(b) identifying all sufficiently unique timestamps in the initial sample;

(c) determining whether the number of sufficiently unique timestamps in the initial sample size divided by the number of real-time events in the initial sample meets or exceeds a predetermined threshold value;

(d) in response to a determination that the number of sufficiently unique timestamps in the initial sample size divided by the number of real-time events in the initial sample meets or exceeds the predetermined threshold value, treating the initial sample as the sample; and

(e) in response to a determination that the number of sufficiently unique timestamps in the initial sample size divided by the number of real-time events in the initial sample does not meet or exceed the predetermined threshold value, increasing the predetermined sample size and repeating (a)-(e), as needed.

21. The system of claim 16, wherein the selecting identifies measures that have a high rate of change within a predefined time window of interest.

22. The system of claim 16, wherein the selecting identifies dimensions by testing for conformance with well-known patterns for commonly tested dimensions.

23. The system of claim 16, wherein the natural language query template is built in connection with at least one question part that has an associated mapping, the mapping specifying for its associated question part:

whether the associated question part is one of a comparative or superlative type question part,

what natural language text is to be included in the associated question part,

whether the associated question part is allowed to vary over time, and

an indication as to what function(s) is/are to be performed on a value associated with the associated question part.

24. The system of claim 23, wherein each said mapping further specifies for its associated question part a type of variable to be used with the associated question part, variable types for the question parts being one of time-based and value-based variable types.

25. A non-transitory computer readable storage medium having a program for forming a natural language query template stored thereon, the program comprising instructions that, when executed, cause a computing system including at least one processor and a memory to at least:

obtain a sample of real-time events from a data stream;

identify, from the events in the sample, measures and dimensions associated therewith;

classify the identified measures and dimensions as belonging to one or more distinct measures and one or more distinct dimensions, respectively;

automatically select at least one of the one or more distinct measures and/or at least one of the one or more distinct dimensions for inclusion in the natural language query template, the at least one selected distinct measure and/or the at least one selected distinct dimension being selected dynamically based on the classification and having a propensity for subsequent user selection that is higher than non-selected distinct measure(s) and/or non-selected distinct dimension(s), the natural language query template including natural language expressions and templated fields, at least one of the templated fields enabling user selection of one of a comparative and a superlative; and

arrange the at least one selected distinct measure and/or the at least one selected distinct dimension in the natural language query template as user-selectable options in at least some of the templated fields;

wherein the natural language query template with specified user-selectable options is transformable into a formalized query executable on the data stream.