Scalable systems and methods for discovering and summarizing test result facts
Techniques are described herein for producing machine-generated findings given a set of user experience test results. In some embodiments, the system generates the findings using an artificial intelligence and machine learning engine. The findings may highlight areas that are predicted to provide the most insight into optimizing a product's design. A finding may be generated based on all or a subset of the test result elements, including qualitative and/or quantitative data contained therein. A finding may summarize a subset of the UX test results that are interrelated. A finding may link a summary to one or more references extracted from the set of test results to show support for the machine-generated insights in the underlying raw test data. Machine-generated findings reports may provide near instantaneous guidance for optimizing product designs while removing extraneous information from a vast quantity of raw test result data.
1. A method comprising:
receiving a set of results for a user experience test;
generating a set of clusters, wherein each respective cluster includes at least one respective result from the set of results for the user experience test and represents a particular topic;
for each cluster:
generating a set of segments, wherein each segment includes a subset of the set of results in the cluster and is generated based at least in part on test element type;
instantiating different finding generators for different segments in the set of segments, wherein a machine learning model is used by at least one of the finding generators; and
generating, with the different finding generators, a respective finding summary for each segment in the cluster based at least in part on the test element type associated with the segment, wherein the finding summaries comprise natural language text readable by humans, and wherein the machine learning model extrapolates from learned patterns to distinguish between high-quality and low-quality results for the user experience test; and
adjusting, via a training model, one or more internal parameters of the machine learning model used by at least one of the finding generators to generate improved finding summaries, wherein the training model retrains the machine learning model using new input data that is local in time such that the model is no longer trained on data that is older than a threshold age.
2. The method of claim 1 , wherein the finding generators include a first finding generator for generating finding summaries for custom question responses, a second finding generator for generating finding summaries for diagnostic responses, a third finding generator for generating finding summaries for expectation responses, and a fourth finding generator for generating finding summaries for heatmap responses.
3. The method of claim 1 , further comprising:
applying a second machine learning model to predict a selection score for the set of results; and
selecting a subset of the set of results to retain based at least in part on the selection score.
4. The method of claim 1 , further comprising:
adding the finding summary to a findings report;
wherein a root node in the findings report indicates a sentiment associated with the finding summary;
wherein the findings report associates references associated with test results for a plurality of diagnostics with the finding summary.
5. The method of claim 1 , wherein the finding summary is a parent node to one or more summaries for one or more child findings.
6. The method of claim 1 , wherein the finding summary is associated with an outcome for a user expectation.
7. The method of claim 1 , wherein the set of results are received through a heatmap associated with a particular image or user interface;
the method further comprising:
performing spatial clustering based on coordinates associated with the set of results;
wherein the set of clusters are generated for a particular spatial cluster and represent topics associated with the spatial cluster; and
generating an updated version of the particular image or user interface, wherein the updated version of the particular image or user interface includes the finding summary at a location based on a set of coordinates associated with the particular spatial cluster.
8. The method of claim 1 , wherein generating the set of clusters comprises:
generating, using a trained embedding model, a set of vector embeddings for a set of quotations included in the set of results;
wherein the set of clusters is generated based at least in part on cosine similarities between the set of vector embeddings.
9. The method of claim 1 , further comprising:
training the machine learning model using historical finding reports for historical user experience tests.
10. One or more non-transitory computer-readable media storing instructions which, when executed by one or more hardware processors cause:
receiving a set of results for a user experience test;
generating a set of clusters, wherein each respective cluster includes at least one respective result from the set of results for the user experience test and represents a particular topic;
for each cluster:
generating a set of segments, wherein each segment includes a subset of the set of results in the cluster and is generated based at least in part on test element type;
instantiating different finding generators for different segments in the set of segments, wherein a machine learning model is used by at least one of the finding generators; and
generating, with the different finding generators, a respective finding summary for each segment in the cluster based at least in part on the test element type associated with the segment, wherein the finding summaries comprise natural language text readable by humans, and wherein the machine learning model extrapolates from learned patterns to distinguish between high-quality and low-quality results for the user experience test; and
adjusting, via a training model, one or more internal parameters of the machine learning model used by at least one of the finding generators to generate improved finding summaries, wherein the training model retrains the machine learning model using new input data that is local in time such that the model is no longer trained on data that is older than a threshold age.
11. The media of claim 10 , wherein the instructions further cause:
applying a second machine learning model to predict a selection score for the set of results; and
selecting a subset of the set of results to retain based at least in part on the selection score.
12. The media of claim 10 , wherein the instructions further cause:
adding the finding summary to a findings report;
wherein a root node in the findings report indicates a sentiment associated with the finding summary;
wherein the findings report associates references associated with test results for a plurality of diagnostics with the finding summary.
13. The media of claim 10 , wherein the finding summary is a parent node to one or more summaries for one or more child findings.
14. The media of claim 10 , wherein the finding summary is associated with an outcome for a user expectation.
15. The media of claim 10 , wherein the set of results are received through a heatmap associated with a particular image or user interface;
wherein the instructions further cause:
performing spatial clustering based on coordinates associated with the set of results;
wherein the set of clusters are generated for a particular spatial cluster and represent topics associated with the spatial cluster; and
generating an updated version of the particular image or user interface, wherein the updated version of the particular image or user interface includes the finding summary at a location based on a set of coordinates associated with the particular spatial cluster.
16. The media of claim 10 , wherein generating the set of clusters comprises:
generating, using a trained embedding model, a set of vector embeddings for a set of quotations included in the set of results;
wherein the set of clusters is generated based at least in part on cosine similarities between the set of vector embeddings.
17. The media of claim 10 , wherein the instructions further cause:
training the machine learning model using historical finding reports for historical user experience tests.
18. A system comprising:
one or more hardware processors;
one or more non-transitory computer-readable media storing instructions which, when executed by the one or more hardware processors cause:
receiving a set of results for a user experience test;
generating a set of clusters, wherein each respective cluster includes at least one respective result from the set of results for the user experience test and represents a particular topic;
for each cluster:
generating a set of segments, wherein each segment includes a subset of the set of results in the cluster and is generated based at least in part on test element type;
instantiating different finding generators for different segments in the set of segments, wherein a machine learning model is used by at least one of the finding generators; and
generating, with the different finding generators, a respective finding summary for each segment in the cluster based at least in part on the test element type associated with the segment, wherein the finding summaries comprise natural language text readable by humans, and wherein the machine learning model extrapolates from learned patterns to distinguish between high-quality and low-quality results for the user experience test; and
adjusting, via a training model, one or more internal parameters of the machine learning model used by at least one of the finding generators to generate improved finding summaries, wherein the training model retrains the machine learning model using new input data that is local in time such that the model is no longer trained on data that is older than a threshold age.