Systems and methods for detecting accessibility failures

1 . A system for evaluating accessibility of one or more software applications, the system comprising:

one or more processors; and

a non-transitory, computer readable medium comprising instructions that, when executed by the one or more processors, causes operations comprising:

monitoring code, entered into a code editor, for a user interface of an updated version of a software application;

generating audio output using a screen reader configured to process textual or visual data of the user interface and the updated version of the software application;

generating predicted textual data by processing the audio output of the user interface using speech recognition;

comparing the predicted textual data to previous textual data, corresponding to a previous version of the software application, to determine a plurality of feature differences present between the updated version of the software application and the previous version of the software application;

identifying that one or more feature differences, of the plurality of feature differences, are unintentional based on comparing the predicted textual data to the previous textual data;

using a machine learning model, trained on code identified with an accessibility failure and code without failure, to identify a code segment, of the code, corresponding to an accessibility failure point in the one or more feature differences based on the one or more feature differences being unintentional based on comparing the predicted textual data to the previous textual data, wherein the accessibility failure point renders content in the user interface inaccessible to a visually impaired user;

determining, based on (1) the code segment and (2) the updated version of the software application, one or more actions for removing the accessibility failure point; and

providing a command for displaying, in the code editor, the one or more actions for removing the accessibility failure point.

2 . The system of claim 1 , wherein the instructions cause the one or more processors to perform operations comprising:

receiving a training dataset comprising indications regarding whether a feature difference corresponds to unintentional or intentional design changes; and

training, based on the training dataset, the machine learning model or a different machine learning model for identifying that the one or more feature differences are unintentional.

3 . The system of claim 1 , wherein determining one or more actions for removing the accessibility failure point comprises:

processing (1) the one or more feature differences and (2) the updated version of the software application using another machine learning model for identifying actions for removing the accessibility failure point.

4 . The system of claim 1 , wherein the predicted textual data comprises (1) a string of text and (2) one or more timestamps and wherein generating the predicted textual data comprises:

generating, for each of a plurality of segments of the string of text, a corresponding timestamp representing a time at which each segment was textualized.

5 . The system of claim 1 , wherein the instructions cause the one or more processors to perform operations comprising:

automatically performing the one or more actions for removing the accessibility failure point; and

transmitting, to a remote device, a notification of successful completion of the one or more actions.

6 . A method, the method comprising:

receiving code, entered into a code editor, for a user interface of an updated version of a software application;

generating audio output using a screen reader configured to process textual or visual data of the user interface and the updated version of the software application;

generating predicted textual data by processing the audio output of the user interface using speech recognition;

comparing the predicted textual data to previous textual data, corresponding to a previous version of the software application, to determine a plurality of feature differences present between the updated version of the software application and the previous version of the software application;

identifying one or more feature differences, of the plurality of feature differences, that are unintentional design changes based on comparing the predicted textual data to the previous textual data;

using a machine learning model, trained on code identified with an accessibility failure and code without failure, to identify a code segment, of the code, corresponding to an accessibility failure point in the one or more feature differences, wherein the accessibility failure point renders content in the user interface inaccessible to a visually impaired user; and

determining, based on (1) the code segment and (2) the updated version of the software application, one or more actions for removing the accessibility failure point.

7 . The method of claim 6 , further comprising transmitting, to a remote device, a command for displaying, at a remote device, the one or more actions for removing the accessibility failure point.

8 . The method of claim 6 , wherein determining one or more actions for removing the accessibility failure point comprises retrieving, from a data structure, the one or more actions based on (1) the one or more feature differences and (2) the updated version of the software application.

9 . The method of claim 6 , further comprising:

receiving a training dataset comprising indications regarding whether a feature difference corresponds to unintentional or intentional design changes; and

training, based on the training dataset, the machine learning model or a different machine learning model for identifying that the one or more feature differences are unintentional.

10 . The method of claim 6 , wherein determining one or more actions for removing the accessibility failure point comprises:

processing (1) the one or more feature differences and (2) the updated version of the software application using a different machine learning model for identifying actions for removing the accessibility failure point.

11 . The method of claim 10 , further comprising:

receiving a training dataset comprising (1) the one or more feature differences, (2) the updated version of the software application, and (3) corresponding actions that were performed to remove accessibility failure points; and

training, based on the training dataset, the different machine learning model for identifying actions for removing accessibility failure points.

12 . The method of claim 6 , wherein the predicted textual data comprises (1) a string of text and (2) one or more timestamps and wherein generating the predicted textual data comprises:

generating, for each of a plurality of segments of the string of text, a corresponding timestamp representing a time at which each segment was textualized.

13 . The method of claim 6 , further comprising:

automatically performing the one or more actions for removing accessibility failure points; and

transmitting, to a remote device, a notification of successful completion of the one or more actions.

14 . One or more non-transitory, computer readable media comprising instructions that, when executed by one or more processors, causes operations comprising:

receiving code, entered into a code editor, for a user interface of an updated version of a software application;

comparing predicted textual data associated with the updated version of the software application to previous textual data, associated with a previous version of the software application, to determine a plurality of feature differences present between the updated version of the software application and the previous version of the software application;

using one or more machine learning models, including a machine learning model trained on code identified with an accessibility failure and code without failure, to identify one or more feature differences, of the plurality of feature differences, that are unintentional based on the predicted textual data being compared to the previous textual data and identify a code segment, of the code, corresponding to an accessibility failure point in the one or more feature differences, wherein the accessibility failure point renders content in the user interface inaccessible to a visually impaired user; and

providing, in the code editor, information based on identifying the code segment corresponding to the accessibility failure point.

15 . The one or more non-transitory, computer readable media of claim 14 , wherein the operations further comprise:

generating audio output using a screen reader configured to process textual or visual data of the user interface and the updated version of the software application; and

generating the predicted textual data by processing the audio output of the user interface using speech recognition.

16 . The one or more non-transitory, computer readable media of claim 15 ,

wherein the predicted textual data comprises (1) a string of text and (2) one or more timestamps, and

wherein generating the predicted textual data comprises:

generating, for each of a plurality of segments of the string of text, a corresponding timestamp representing a time at which each segment was textualized.

17 . The one or more non-transitory, computer readable media of claim 14 , wherein the operations further comprise:

determining one or more actions removing the accessibility failure point.

18 . The one or more non-transitory, computer readable media of claim 14 , wherein the operations further comprise:

receiving a training dataset comprising (1) a plurality of textual data, (2) a plurality of feature differences, and (3) corresponding indications regarding whether a feature difference corresponds to unintentional or intentional design changes; and

training, based on the training dataset, at least one of the one or more machine learning models.

19 . The one or more non-transitory, computer readable media of claim 14 , wherein the operations further comprise:

processing (1) the one or more feature differences and (2) the updated version of the software application using a different machine learning model for identifying actions for removing the accessibility failure point.

20 . The one or more non-transitory, computer readable media of claim 19 , wherein the operations further comprise:

receiving a training dataset comprising (1) the one or more feature differences, (2) code of the updated version of the software application, and (3) corresponding actions that were performed to remove the accessibility failure point; and

training, based on the training dataset, the different machine learning model for identifying the actions for removing the accessibility failure point.