Linking try in metrics docs (#691)

* Linking try in metrics docs

* more links

* Updating docs with links to trunk

* modifications based on comments

docs/metrics/accuracy.md

@@ -17,6 +17,10 @@ and should be used with caution.
 - :kolena-manual-16: API Reference: [`accuracy` ↗][kolena.metrics.accuracy]
 </div>
 
+!!!example
+    To see an example of Accuracy metric in use, checkout the
+    [MMLU dataset on app.kolena.com/try](https://app.kolena.io/try/dataset/standards?datasetId=32&models=N4IglgJiBcDsDMAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAFgF9Gg&models=N4IglgJiBcBsCcAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAFgF9Gg&models=N4IglgJiBcDsBMAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAFgF9Gg&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXgF1mQA)
+
 ## Implementation Details
 
 Accuracy is generally used to evaluate classification models. Aside from classification, accuracy is also often

docs/metrics/average-precision.md

@@ -25,6 +25,10 @@ metric used across popular object detection benchmarks such as [PASCAL VOC 2012]
 [COCO](https://cocodataset.org/#detection-2015), and
 [Open Images V7](https://storage.googleapis.com/openimages/web/evaluation.html).
 
+!!!example
+    To see an example use of Average Precision, checkout the
+    [KITTI Vision Benchmark Suite on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=44&models=N4IglgJiBcAcCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&models=N4IglgJiBcAcAsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Implementation Details
 
 The general definition of AP is finding the approximation of the area under the [PR curve](./pr-curve.md). The actual

docs/metrics/averaging-methods.md

@@ -15,6 +15,10 @@ most notably:
 - [**Micro**](#micro-average): global average of per-sample TP, FP, FN scores
 - [**Weighted**](#weighted-average): mean of all per-class scores, weighted by sample sizes for each class
 
+!!!example
+    To see an example of Macro Averaging Method, checkout the
+    [KITTI Vision Benchmark Suite on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=44&models=N4IglgJiBcAcCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&models=N4IglgJiBcAcAsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Example: Multiclass Classification
 
 Let’s consider the following multiclass classification metrics, computed across a total of 10 samples:

docs/metrics/bertscore.md

@@ -10,6 +10,10 @@ embeddings to capture the semantic and contextual information of words and phras
 texts. This approach makes BERTScore more effective at assessing the quality of candidate text because it considers not
 only exact word matches but also the overall meaning, fluency, and order of the output.
 
+!!!example
+    To see and an example of Bert Score, checkout the
+    [CNN-DailyMail on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=39&models=N4IglgJiBcCcA0IDGB7AdgMzAcwK4CcBDAFzHRlEhgEYBfWoA&models=N4IglgJiBcAcA0IDGB7AdgMzAcwK4CcBDAFzHRlEhgEYBfWoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63Jn1ECGPFgF1mQA)
+
 ??? question "Recall: BERT & Textual Embeddings"
     BERT (**B**idirectional **E**ncoder **R**epresentations from **T**ransformers) is a popular language model used to
     generate embeddings from words and phrases. Textual embeddings are learned dense token representations that capture

docs/metrics/confusion-matrix.md

@@ -27,6 +27,10 @@ imbalances.
 - :kolena-manual-16: API Reference: [`ConfusionMatrix` ↗][kolena.workflow.plot.ConfusionMatrix]
 </div>
 
+!!!example
+    To see an example of confusion matrix, checkout the
+    [CIFAR-10 on app.kolena.com/try.](https://app.kolena.io/try/dataset/debugger?datasetId=20&aggregations=N4IgNghgRgpmIC4QEEDGqCuAnCqCeIANCAM4D22qMiIWMJGYALkSBAObt3sRNlaJQHLjB58BSXJhz5WEAG4wc7AJYA7dgFkYTABZkAJjQC2uLGRABfYgAcIOYyUEguFNQYAqWDHoBiKuCMkVwx3AH0mbz0AOkhYeGJ1ADMlGDUqf0CaVEgSEhUklVReFTI1WOg4Vj06En0wILVGMET0sAwDGDCIDFREJIgwEhhrECT%2BUxYkGyUqNSYOamJdFXZdMFXdKdAVtaUASRIAIR0mJURIjBHiEJsAOQhjaiRtSKKAAgBxcwwbKyA&aggregations=N4IgNghgRgpmIC4QAUBOMDGBLAzlg9gHYAEAFALIQar4CUIANCDvgK6oYyIjo6tgAXRiAgBzUelEQB%2BVIlBiJMKTLlIADumx4iwiADcYqMVkKjyMAQAt8AE24BbKjRABfJuojGHOeSAlshLYAKqis1gBiWHD2SAGsQQD6AmHWAHSQsPBMpgBmRjCEnFEx3BiQOHi5WBjSBIQZ0HDC1rw2YLGE-GA5RWCstjCJEKwYiLkQYDgw7iC5sk5CGkachAJiXExWWKJWYDtWS6Dbu0YAkjgAQpYCRogprDNM8eoAchAOXEgWKTXEAOI0VjqNxAA&aggregations=N4IgNghgRgpmIC4QCUYGMJjAAgBQFkI0AnAewEoQAaEAZ1IFdi0ZERiZaGwAXakCAHNBHQRB6liiUEJEwxEqUg4Ys-CADcYxIQEsAdoPwweAC1IATNgFsiZEAF8aABwg7rtaSBGN9FgCrEDGYAYrpwVkg%2BDH4A%2BjxBZgB0kLDwNAYAZtow%2BixhEWxokLS0upm6GDy6pPop0HD8Zhy05mCR%2BtxgGXlgDBYwsRAMaIiZmLQwTiCZkrZ8SM7aLPo8Qqw0prqCpmDbpgugWzvaAJK0AEImPNqICQxTNNHOAHIQ1qxIxgmV2ADiZAYzkcQA&aggregations=N4IgNghgRgpmIC4QDECMACAygYwPYCcZ0AKAWQm31wEoQAaEAZ1wFd9sZERDGWwAXeiAgBzEYRER%2BBRKFHiYk6fi4AzVAH1GeQkIgA3GPlEBLAHYjSMfgAtcAEy4BbClRABfBgAcIxp41kQcVYzewAVfBZbZBM4RyRgllCNfkjbADpIWHgGc1UjGDMOGLiubEhGRhNVE2wpE1wzTOg4IVseOzB4sz4wXKKwFnsYDQgWbERVCDBGGE8QVQIXQSQvIw4zflFOBhsTERswfZsV0D2DowBJRgAha34jRFSWOYZErwA5CCdOJCtU2roADiVBYXg8QA&models=N4IglgJiBcDMA0IDGB7AdgMzAcwK4CcBDAFzHRlEhgFYBfWoA&plots=N4IgHiBcoM4PYFcBOBjAplES0wQGwBcQAaEAazQE9MU8BDGGASwDMmU6Cm4A7AOnoAjNHhIghIzNlyE%2BtBszYcuvAXWGjSAEyYx1eNFqgs6eGGgC%2BpatBDxk6TFs50ADnCY8ipCjZABzJEQeLQB9AiQEAgALNQ0xCVFIEGcCNw8vPkDgsIio2MSxHT1BAyNIEzNLCyA)
+
 ## Implementation Details
 
 The implementation of a confusion matrix depends on whether the workflow concerns one or more classes.

docs/metrics/difficulty-score.md

@@ -14,6 +14,10 @@ indicates that models produce the ideal datapoint-level metrics (e.g. lower infe
 and a higher difficulty score indicates that models consistently face problems or "difficulty"
 (e.g. longer inference time, lower [BLEU scores](./bleu.md), and/or lower [recall](./recall.md)).
 
+!!!example
+    To see an example of the Difficulty Score in action, checkout the
+    [Object Detection (COCO 2014) on app.kolena.com/try.](https://app.kolena.io/try/dataset/studio?datasetId=14&models=N4IglgJiBcBMCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYAWAX3qA&models=N4IglgJiBcBMAsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZ4BfOoA&modelResultNullFilters=N4IglgJiBcBMCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYAWAX2TAGcB9DfAG05i2M%2BYBTekA&modelResultNullFilters=N4IglgJiBcBMAsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZ4BfZMAZwH0N8AbDmLYjpgKZ0gA&sortByType=datapoint&sortBy=_kolena.extracted.difficulty_score&filters=datapoint._kolena.extracted.difficulty_score%3AN4IgdgrgtgRgpgJwPoIIZgOZxALlFASzFwAYA6ATgBoQpUAPXARgF8aB7ABwBcD2wAzrlAI4WekjrcAxgAtcAM1QAbAXBqjxSBctQYhOECBpEAbojWKValiyA%3Anr)
+
 !!! note
     For Kolena to calculate the `datapoint.difficulty_score` you must have:
 

docs/metrics/f1-score.md

@@ -26,6 +26,10 @@ object detection, semantic segmentation, and information retrieval.
 - :kolena-manual-16: API Reference: [`f1_score` ↗][kolena.metrics.f1_score]
 </div>
 
+!!!example
+    To see an example of of the F1 Score, checkout the
+    [Object Detection (COCO 2014) on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=14&models=N4IglgJiBcBMCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYAWAX3qA&models=N4IglgJiBcBMAsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZ4BfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Implementation Details
 
 Using [TP / FP / FN / TN](./tp-fp-fn-tn.md), we can define [precision](./precision.md) and [recall](./recall.md).

docs/metrics/mean-absolute-error.md

@@ -12,6 +12,10 @@ MAE represents the mean of the absolute differences between predicted and actual
 treating each discrepancy equally. A large value is indicative of poor performance, and communicates
 how much on average a prediction will deviate from the actual value in the units of the ground truth.
 
+!!!example
+    To see and example of Mean Absolute Error, checkout the
+    [STS Benchmark on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=12&models=N4IglgJiBcCMDsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&models=N4IglgJiBcCMBsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Implementation Details
 
 MAE is calculated by taking the average of the absolute differences between the predicted values and the actual values.

docs/metrics/pearson-correlation.md

@@ -10,6 +10,10 @@ positive linear correlation, where as one variable increases, the other also inc
 to -1 signifies a strong negative linear correlation, meaning as one variable increases, the other decreases
 proportionally. A value of 0 indicates no linear correlation between the variables.
 
+!!!example
+    To see an example of the Pearson Correlation, checkout the
+    [STS Benchmark on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=12&models=N4IglgJiBcCMDsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&models=N4IglgJiBcCMBsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Implementation Details
 
 The correlation coefficient is calculated by dividing the covariance of $x$ and $y$ by their individual standard

docs/metrics/pr-curve.md

@@ -17,6 +17,11 @@ thresholds, showing a trade-off of how precision and recall values change when a
 - :kolena-manual-16: API Reference: [`CurvePlot` ↗][kolena.workflow.plot.CurvePlot]
 </div>
 
+!!!example
+    To see an example of the PR curve, checkout the
+    [Object Detection (COCO 2014) on app.kolena.com/try](https://app.kolena.io/try/dataset/debugger?datasetId=14&aggregations=N4IgNghgRgpmIC4QAUBOMDGBLAzlg9gHYgA0IO%2BArqhjIiOjpWAC6kgQDmn6nEL%2BVIlBceMPgKFIADumx4i7CADcYqLlkKcAsjBYALfABN6AWywZU%2BEAF8y0iOtM5hIAGaEAYljgmkngDl2N2lvX3pPZHYWUJ8wPxAAFSiyHAxBGDD4%2BjSM9khYMCyEgrhbMjdBU356aTVaQhYuOjJ9LE59MHb9NgRQNo61AEkcACE9FjVEFlRKGDsQHippAIhTOiRdGYsAAgBxK0ppWyA&aggregations=N4IgNghgRgpmIC4QCUYGMJngGhAZwHsBXAJzRkRBJjyLABcRcIBzF6liegkxUV9jE7deSahixMQEAG4wSrAJYA7FgFkY9ABYEAJpQC2itCQIgAvrgAOEBQbx8QAM2UAxRXH1JXAOSlOrd09KVwAFKXpAjzAvEAAVcNw8NB4YIJjKZNSpSFgwdNjcuAtcJx4DLkoreXJlelYKXC1FFi0wFq1GBFBm1vkASTwAIU16eUR6EiIYSxB2YisfCAMKJA1J4wACAHFTIisLIA&aggregations=N4IgNghgRgpmIC4QDECMACAygYwPYCcYQAaEAZ1wFd9silCzKwAXEkCAcw8I4mYMShO3GL375EIAGaoA%2BmTyE2EAG4x8nAJYA7DgFkYzABa4AJpIC2m7PlwgAvqQAOEDRbKDp25JrjmkyABybFJOPn6SyAAKbMxhvmD%2BIAAqMaQKBDDhiZIZSqSQsGDZSYVwDqRSBBZ8kk7qtNrMnESkRpocRmAdRqwIoO2d6gCSZABChszqiMz4lDCOINxUToEQFnQgBrPW6ADitpRODkA&aggregations=N4IgNghgRgpmIC4QFkYQHYAIByMAumAwgPYBOpMAxnujAM50gA0IdxArqZTIiBXezB5mICAHMxFMRDxleAWzToRABwikI8xglABrGAE9eAfV3EwMdBAB0AExikAlgDcYt67TzHKZCtVoMIAC%2BLABmZPIyiOiCYCwAFo5i8WBJ8cI6IInJDgCSdABC%2BHgOiHik7DAhIJIcKtiaPEio5Y6UmADipHXBQA&models=N4IglgJiBcBMCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYAWAX3qA&modelResultNullFilters=N4IglgJiBcBMCsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYAWAX2TAGcB9DfAG05i2M%2BYBTekA)
+    on Kolena's public dataset.
+
 ## Implementation Details
 
 The curve’s points (precisions and recalls) are calculated with a varying threshold, and made into points (precision

docs/metrics/roc-curve.md

@@ -10,6 +10,10 @@ classification models by using the [true positive rate (TPR)](./recall.md) and t
 [false positive rate (FPR)](./fpr.md). The curve is built with the TPR on the y-axis and the FPR on the x-axis computed
 across many thresholds, showing a trade-off of how TPR and FPR values change when a classification threshold changes.
 
+!!!example
+    To see an example of the ROC cureve, chekcout the
+    [CIFAR-10 on app.kolena.com/try.](https://app.kolena.io/try/dataset/debugger?datasetId=20&aggregations=N4IgNghgRgpmIC4QEEDGqCuAnCqCeIANCAM4D22qMiIWMJGYALkSBAObt3sRNlaJQHLjB58BSXJhz5WEAG4wc7AJYA7dgFkYTABZkAJjQC2uLGRABfYgAcIOYyUEguFNQYAqWDHoBiKuCMkVwx3AH0mbz0AOkhYeGJ1ADMlGDUqf0CaVEgSEhUklVReFTI1WOg4Vj06En0wILVGMET0sAwDGDCIDFREJIgwEhhrECT%2BUxYkGyUqNSYOamJdFXZdMFXdKdAVtaUASRIAIR0mJURIjBHiEJsAOQhjaiRtSKKAAgBxcwwbKyA&aggregations=N4IgNghgRgpmIC4QAUBOMDGBLAzlg9gHYAEAFALIQar4CUIANCDvgK6oYyIjo6tgAXRiAgBzUelEQB%2BVIlBiJMKTLlIADumx4iwiADcYqMVkKjyMAQAt8AE24BbKjRABfJuojGHOeSAlshLYAKqis1gBiWHD2SAGsQQD6AmHWAHSQsPBMpgBmRjCEnFEx3BiQOHi5WBjSBIQZ0HDC1rw2YLGE-GA5RWCstjCJEKwYiLkQYDgw7iC5sk5CGkachAJiXExWWKJWYDtWS6Dbu0YAkjgAQpYCRogprDNM8eoAchAOXEgWKTXEAOI0VjqNxAA&aggregations=N4IgNghgRgpmIC4QCUYGMJjAAgBQFkI0AnAewEoQAaEAZ1IFdi0ZERiZaGwAXakCAHNBHQRB6liiUEJEwxEqUg4Ys-CADcYxIQEsAdoPwweAC1IATNgFsiZEAF8aABwg7rtaSBGN9FgCrEDGYAYrpwVkg%2BDH4A%2BjxBZgB0kLDwNAYAZtow%2BixhEWxokLS0upm6GDy6pPop0HD8Zhy05mCR%2BtxgGXlgDBYwsRAMaIiZmLQwTiCZkrZ8SM7aLPo8Qqw0prqCpmDbpgugWzvaAJK0AEImPNqICQxTNNHOAHIQ1qxIxgmV2ADiZAYzkcQA&aggregations=N4IgNghgRgpmIC4QDECMACAygYwPYCcZ0AKAWQm31wEoQAaEAZ1wFd9sZERDGWwAXeiAgBzEYRER%2BBRKFHiYk6fi4AzVAH1GeQkIgA3GPlEBLAHYjSMfgAtcAEy4BbClRABfBgAcIxp41kQcVYzewAVfBZbZBM4RyRgllCNfkjbADpIWHgGc1UjGDMOGLiubEhGRhNVE2wpE1wzTOg4IVseOzB4sz4wXKKwFnsYDQgWbERVCDBGGE8QVQIXQSQvIw4zflFOBhsTERswfZsV0D2DowBJRgAha34jRFSWOYZErwA5CCdOJCtU2roADiVBYXg8QA&models=N4IglgJiBcAsA0IDGB7AdgMzAcwK4CcBDAFzHRlEhgFYBfWoA&models=N4IglgJiBcDMA0IDGB7AdgMzAcwK4CcBDAFzHRlEhgFYBfWoA)
+
 !!! info "Guides: TPR and FPR"
 
     The TPR is also known as sensitivity or recall, and it represents the proportion of true positive inferences

docs/metrics/root-mean-squared-error.md

@@ -14,6 +14,11 @@ interpreting the magnitude of prediction errors and penalizing larger discrepanc
 Like [MSE](./mean-squared-error.md) and [MAE](./mean-absolute-error.md) a large value is indicative of
 poor performance.
 
+!!!example
+    To see and example of the Root Mean Squared Error, checkout the
+    [STS Benchmark on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=12&models=N4IglgJiBcCMDsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&models=N4IglgJiBcCMBsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+    on Kolena's public dataset.
+
 ## Implementation Details
 
 RMSE is calculated by first computing the mean of the squared differences between the predicted values and the actual

docs/metrics/spearman-correlation.md

@@ -11,6 +11,10 @@ strong positive monotonic correlation, where as one variable increases, the othe
 signifies a strong negative monotonic correlation, meaning as one variable increases, the other decreases. A value of
 0 indicates no monotonic correlation between the variables.
 
+!!!example
+    To see an example of the Spearman's Correlation, checkout the
+    [STS Benchmark on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=12&models=N4IglgJiBcCMDsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&models=N4IglgJiBcCMBsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIZYBfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXm3Ad63JswC6zIA)
+
 ## Implementation Details
 
 The Spearman's Rank Correlation is calculated by ranking the data points and then applying

docs/metrics/wer-cer-mer.md

@@ -10,6 +10,10 @@ the similarity between reference and candidate texts, with zero being a perfect
 error rate can be infinitely high, match error rate is always between 0 and 1. Each of these metrics
 have their nuances that reveal different errors within texts.
 
+!!!example
+    To see and example of WER, CER and MER, checkout the
+    [LibriSpeech on app.kolena.com/try.](https://app.kolena.io/try/dataset/standards?datasetId=22&models=N4IglgJiBcBsAcAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&models=N4IglgJiBcBsDsAaEBjA9gOwGZgOYFcAnAQwBcxMZRIYBGAX3qA&models=N4IglgJiBcBssBoQGMD2A7AZmA5gVwCcBDAFzAxlEhgEYBfOoA&metricGroupVisibilities=N4IgbglgzhBGA2BTEAuALgJwK6IDQgFtFMIBjKVAbVEhgWXW0QF9cbo4lVMdX26ujXgF1mQA)
+
 ## Substitutions, Deletions, and Insertions
 
 The building blocks of each metric include substitution, deletion, and insertion errors. These errors reveal different