Skip to content

The Bias Visualizer

Module for visualizing results from the bias evaluation.

Source code in genda_lens/genda_lens.py 
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
class Visualizer:
    """Module for visualizing results from the bias evaluation."""

    def __init__(self):
        import seaborn as sns

        sns.set_style("whitegrid", rc={"lines.linewidth": 1})
        sns.set_context("notebook", font_scale=1.2)

    def visualize_results(self, data, framework, model_name, task=None):
        """Visualize output from any of the genderbias tests that can be run in this package.

        Args:
            data (df): detailed output from any of the tests.
            framework (str): choose between "ner", "dawinobias" or "abc".
            model_name (str): model name
            task (str, optional): choose between "lm" or "coref" if the output is either from the dawinobias or abc framework.
        Returns:
            plot (plot): seaborn plot visualization.

        *EXAMPLE*

           ```python
           from genda_lens import Visualizer

           # initiate visualizer
           viz = Visualizer()

           # visualize ner results
           plot = viz.visualize_results(data = detailed_output_ner, framework = "ner", model_name "my-model-name")

           # visualize abc lm results
           plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="lm")

           # visualize abc coref results
           plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="coref")
           ```
        """
        import matplotlib.pyplot as plt
        import pandas as pd
        import seaborn as sns

        sns.set_style("whitegrid", rc={"lines.linewidth": 10})

        df = pd.DataFrame()

        if framework == "abc":
            df["Stereotypical Occupations"] = [
                "Female occupations",
                "Male occupations",
                "Main Effect",
            ] * 2
            df["Anti-reflexive Pronoun"] = ["Female"] * 3 + ["Male"] * 3
            markers = ["o", "o", "o"]
            x = "Anti-reflexive Pronoun"
            nuance = "Stereotypical Occupations"

            if task == "coref":
                try:
                    mean_fem = list(data.loc["Mean Rate of Detected Clusters"])[1]
                    mean_male = list(data.loc["Mean Rate of Detected Clusters"])[3]
                    fpr_fem_pron_fem_occ = list(data.loc["Rate of Detected Clusters"])[
                        0
                    ]
                    fpr_fem_pron_male_occ = list(data.loc["Rate of Detected Clusters"])[
                        1
                    ]
                    fpr_male_pron_fem_occ = list(data.loc["Rate of Detected Clusters"])[
                        2
                    ]
                    fpr_male_pron_male_occ = list(
                        data.loc["Rate of Detected Clusters"]
                    )[3]
                except:
                    mean_fem = float(data.iloc[2, 2])
                    mean_male = float(data.iloc[2, 4])
                    fpr_fem_pron_fem_occ = float(data.iloc[4, 1])
                    fpr_fem_pron_male_occ = float(data.iloc[4, 2])
                    fpr_male_pron_fem_occ = float(data.iloc[4, 3])
                    fpr_male_pron_male_occ = float(data.iloc[4, 4])

                points = [
                    fpr_fem_pron_fem_occ,
                    fpr_fem_pron_male_occ,
                    mean_fem,
                    fpr_male_pron_fem_occ,
                    fpr_male_pron_male_occ,
                    mean_male,
                ]

                df["False Positive Rates"] = points
                y = "False Positive Rates"
                title = f"ABC Coref Task: {model_name}"

            elif task == "lm":
                try:
                    points = [
                        float(data.iloc[4, 1].split(" ")[0]),
                        float(data.iloc[4, 2].split(" ")[0]),
                        float(data.iloc[2, 1].split(" ")[0]),
                        float(data.iloc[4, 3].split(" ")[0]),
                        float(data.iloc[4, 4].split(" ")[0]),
                        float(data.iloc[2, 3].split(" ")[0]),
                    ]
                except:
                    points = [
                        float(data.iloc[4, 0].split(" ")[0]),
                        float(data.iloc[4, 1].split(" ")[0]),
                        float(data.iloc[2, 0].split(" ")[0]),
                        float(data.iloc[4, 2].split(" ")[0]),
                        float(data.iloc[4, 3].split(" ")[0]),
                        float(data.iloc[2, 2].split(" ")[0]),
                    ]
                df["Median Perplexity"] = points
                y = "Median Perplexity"
                title = f"ABC LM Task: {model_name}"

        elif framework == "dawinobias":
            df["Pronoun"] = ["Female (F1)", "Male (F1)", "Main Effect (Accuracy)"] * 2
            if task == "coref":
                try:
                    points = [
                        float(data.iloc[4, 1]),
                        float(data.iloc[4, 2]),
                        float(data.iloc[2, 1]),
                        float(data.iloc[4, 3]),
                        float(data.iloc[4, 4]),
                        float(data.iloc[2, 3]),
                    ]
                except:
                    points = [
                        data.loc["F1"][0],
                        data.loc["F1"][1],
                        data.loc["Accuracy"][0],
                        data.loc["F1"][2],
                        data.loc["F1"][3],
                        data.loc["Accuracy"][2],
                    ]
                title = f"DaWinoBias, Coreference Task: {model_name}"
            elif task == "lm":
                try:  # if loaded
                    points = [
                        float(data.iloc[4, 1]),
                        float(data.iloc[4, 2]),
                        float(data.iloc[2, 1]),
                        float(data.iloc[4, 3]),
                        float(data.iloc[4, 4]),
                        float(data.iloc[2, 3]),
                    ]
                except:  # if
                    points = [
                        data.loc["F1"][0],
                        data.loc["F1"][1],
                        data.loc["Accuracy"][0],
                        data.loc["F1"][2],
                        data.loc["F1"][3],
                        data.loc["Accuracy"][2],
                    ]
                title = f"DaWinoBias, LM Task: {model_name}"

            df["Performance"] = points
            df["Condition"] = ["Anti-stereotypical"] * 3 + ["Pro-stereotypical"] * 3
            x = "Condition"
            y = "Performance"
            nuance = "Pronoun"
            markers = ["o", "o", "o"]
            title = title

        elif framework == "ner":
            df["Protected Group"] = ["Majority (F1)", "Minority (F1)"] * 2
            try:
                points = [
                    float(data.iloc[2, 1].split(" ")[0]),
                    float(data.iloc[4, 2].split(" ")[0]),
                    float(data.iloc[2, 3].split(" ")[0]),
                    float(data.iloc[4, 4].split(" ")[0]),
                ]
            except:
                points = [
                    float(data.iloc[4, 0].split(" ")[0]),
                    float(data.iloc[4, 1].split(" ")[0]),
                    float(data.iloc[4, 2].split(" ")[0]),
                    float(data.iloc[4, 3].split(" ")[0]),
                ]

            df["Performance"] = points
            df["Augmentation"] = ["Female Names"] * 2 + ["Male Names"] * 2
            x = "Augmentation"
            y = "Performance"
            nuance = "Protected Group"
            markers = ["o", "o"]
            title = f"NER Task, Augmented DaNe: {model_name}"

        sns.pointplot(
            data=df,
            x=x,
            y=y,
            hue=nuance,
            dodge=True,
            join=True,
            markers=markers,
            scale=1.2,
            linestyles=[":", ":", "-"],
            palette=["sandybrown", "mediumpurple", "darkgrey"],
        ).set_title(title)

        plt.minorticks_on()

        return plt

visualize_results(data, framework, model_name, task=None)

Visualize output from any of the genderbias tests that can be run in this package.

Parameters:

Name Type Description Default
data df

detailed output from any of the tests.

 required
framework str

choose between "ner", "dawinobias" or "abc".

 required
model_name str

model name

 required
task str

choose between "lm" or "coref" if the output is either from the dawinobias or abc framework.

 None

Returns:

Name Type Description
plot plot

seaborn plot visualization.

EXAMPLE

from genda_lens import Visualizer

# initiate visualizer
viz = Visualizer()

# visualize ner results
plot = viz.visualize_results(data = detailed_output_ner, framework = "ner", model_name "my-model-name")

# visualize abc lm results
plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="lm")

# visualize abc coref results
plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="coref")
Source code in genda_lens/genda_lens.py 
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
def visualize_results(self, data, framework, model_name, task=None):
    """Visualize output from any of the genderbias tests that can be run in this package.

    Args:
        data (df): detailed output from any of the tests.
        framework (str): choose between "ner", "dawinobias" or "abc".
        model_name (str): model name
        task (str, optional): choose between "lm" or "coref" if the output is either from the dawinobias or abc framework.
    Returns:
        plot (plot): seaborn plot visualization.

    *EXAMPLE*

       ```python
       from genda_lens import Visualizer

       # initiate visualizer
       viz = Visualizer()

       # visualize ner results
       plot = viz.visualize_results(data = detailed_output_ner, framework = "ner", model_name "my-model-name")

       # visualize abc lm results
       plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="lm")

       # visualize abc coref results
       plot = viz.visualize_results(data = detailed_output_lm, framework = "abc", model_name "my-model-name", task="coref")
       ```
    """
    import matplotlib.pyplot as plt
    import pandas as pd
    import seaborn as sns

    sns.set_style("whitegrid", rc={"lines.linewidth": 10})

    df = pd.DataFrame()

    if framework == "abc":
        df["Stereotypical Occupations"] = [
            "Female occupations",
            "Male occupations",
            "Main Effect",
        ] * 2
        df["Anti-reflexive Pronoun"] = ["Female"] * 3 + ["Male"] * 3
        markers = ["o", "o", "o"]
        x = "Anti-reflexive Pronoun"
        nuance = "Stereotypical Occupations"

        if task == "coref":
            try:
                mean_fem = list(data.loc["Mean Rate of Detected Clusters"])[1]
                mean_male = list(data.loc["Mean Rate of Detected Clusters"])[3]
                fpr_fem_pron_fem_occ = list(data.loc["Rate of Detected Clusters"])[
                    0
                ]
                fpr_fem_pron_male_occ = list(data.loc["Rate of Detected Clusters"])[
                    1
                ]
                fpr_male_pron_fem_occ = list(data.loc["Rate of Detected Clusters"])[
                    2
                ]
                fpr_male_pron_male_occ = list(
                    data.loc["Rate of Detected Clusters"]
                )[3]
            except:
                mean_fem = float(data.iloc[2, 2])
                mean_male = float(data.iloc[2, 4])
                fpr_fem_pron_fem_occ = float(data.iloc[4, 1])
                fpr_fem_pron_male_occ = float(data.iloc[4, 2])
                fpr_male_pron_fem_occ = float(data.iloc[4, 3])
                fpr_male_pron_male_occ = float(data.iloc[4, 4])

            points = [
                fpr_fem_pron_fem_occ,
                fpr_fem_pron_male_occ,
                mean_fem,
                fpr_male_pron_fem_occ,
                fpr_male_pron_male_occ,
                mean_male,
            ]

            df["False Positive Rates"] = points
            y = "False Positive Rates"
            title = f"ABC Coref Task: {model_name}"

        elif task == "lm":
            try:
                points = [
                    float(data.iloc[4, 1].split(" ")[0]),
                    float(data.iloc[4, 2].split(" ")[0]),
                    float(data.iloc[2, 1].split(" ")[0]),
                    float(data.iloc[4, 3].split(" ")[0]),
                    float(data.iloc[4, 4].split(" ")[0]),
                    float(data.iloc[2, 3].split(" ")[0]),
                ]
            except:
                points = [
                    float(data.iloc[4, 0].split(" ")[0]),
                    float(data.iloc[4, 1].split(" ")[0]),
                    float(data.iloc[2, 0].split(" ")[0]),
                    float(data.iloc[4, 2].split(" ")[0]),
                    float(data.iloc[4, 3].split(" ")[0]),
                    float(data.iloc[2, 2].split(" ")[0]),
                ]
            df["Median Perplexity"] = points
            y = "Median Perplexity"
            title = f"ABC LM Task: {model_name}"

    elif framework == "dawinobias":
        df["Pronoun"] = ["Female (F1)", "Male (F1)", "Main Effect (Accuracy)"] * 2
        if task == "coref":
            try:
                points = [
                    float(data.iloc[4, 1]),
                    float(data.iloc[4, 2]),
                    float(data.iloc[2, 1]),
                    float(data.iloc[4, 3]),
                    float(data.iloc[4, 4]),
                    float(data.iloc[2, 3]),
                ]
            except:
                points = [
                    data.loc["F1"][0],
                    data.loc["F1"][1],
                    data.loc["Accuracy"][0],
                    data.loc["F1"][2],
                    data.loc["F1"][3],
                    data.loc["Accuracy"][2],
                ]
            title = f"DaWinoBias, Coreference Task: {model_name}"
        elif task == "lm":
            try:  # if loaded
                points = [
                    float(data.iloc[4, 1]),
                    float(data.iloc[4, 2]),
                    float(data.iloc[2, 1]),
                    float(data.iloc[4, 3]),
                    float(data.iloc[4, 4]),
                    float(data.iloc[2, 3]),
                ]
            except:  # if
                points = [
                    data.loc["F1"][0],
                    data.loc["F1"][1],
                    data.loc["Accuracy"][0],
                    data.loc["F1"][2],
                    data.loc["F1"][3],
                    data.loc["Accuracy"][2],
                ]
            title = f"DaWinoBias, LM Task: {model_name}"

        df["Performance"] = points
        df["Condition"] = ["Anti-stereotypical"] * 3 + ["Pro-stereotypical"] * 3
        x = "Condition"
        y = "Performance"
        nuance = "Pronoun"
        markers = ["o", "o", "o"]
        title = title

    elif framework == "ner":
        df["Protected Group"] = ["Majority (F1)", "Minority (F1)"] * 2
        try:
            points = [
                float(data.iloc[2, 1].split(" ")[0]),
                float(data.iloc[4, 2].split(" ")[0]),
                float(data.iloc[2, 3].split(" ")[0]),
                float(data.iloc[4, 4].split(" ")[0]),
            ]
        except:
            points = [
                float(data.iloc[4, 0].split(" ")[0]),
                float(data.iloc[4, 1].split(" ")[0]),
                float(data.iloc[4, 2].split(" ")[0]),
                float(data.iloc[4, 3].split(" ")[0]),
            ]

        df["Performance"] = points
        df["Augmentation"] = ["Female Names"] * 2 + ["Male Names"] * 2
        x = "Augmentation"
        y = "Performance"
        nuance = "Protected Group"
        markers = ["o", "o"]
        title = f"NER Task, Augmented DaNe: {model_name}"

    sns.pointplot(
        data=df,
        x=x,
        y=y,
        hue=nuance,
        dodge=True,
        join=True,
        markers=markers,
        scale=1.2,
        linestyles=[":", ":", "-"],
        palette=["sandybrown", "mediumpurple", "darkgrey"],
    ).set_title(title)

    plt.minorticks_on()

    return plt