Natural-language insights of the results are generated from an api. Here is an example.
[
{"SalientFeatureStat":"As the value of span_density increases, the system performance will increase. System developers can consider improving the performance of samples with smaller span_density value."},
{"SalientFeatureStat":"As the value of span_length increases, the system performance will decrease. System developers can consider improving the performance of samples with larger span_length value."},
{"SalientFeatureStat":"As the value of span_econ increases, the system performance will increase. System developers can consider improving the performance of samples with smaller span_econ value."},
{"SalientFeatureStat":"As the value of span_efre increases, the system performance will increase. System developers can consider improving the performance of samples with smaller span_efre value."},
{"MaxPerformanceGapFeatureStat":"The maximum and minimum performance difference of the system on the span_true_label feature's bucket is the largest of all the features."},
{"FrequentMispredictionStat":"The system frequently mispredicted samples with label `O` as `MISC`, the percentage of errors accounted for 0.14754098360655737 of the total."},
{"FrequentMispredictionStat":"The system frequently mispredicted samples with label `ORG` as `O`, the percentage of errors accounted for 0.12786885245901639 of the total."},
{"FrequentMispredictionStat":"The system frequently mispredicted samples with label `O` as `ORG`, the percentage of errors accounted for 0.1180327868852459 of the total."}
]
Now they can be displayed in the analysis report. The System Insights section is only displayed when the API returns non-empty results.
Add system insights to analysis report
Natural-language insights of the results are generated from an api. Here is an example.
Now they can be displayed in the analysis report. The System Insights section is only displayed when the API returns non-empty results.