Deep learning bias detection in text
In one embodiment, a method includes obtaining text from a user, applying the text to a deep learning neural network to generate a plurality of bias coordinates defining a point in an embedded space, and, in response to determining that at least one of the plurality of bias coordinates exceeds a threshold, providing an indication of bias to the user.
1. A method comprising:
obtaining text from a user;
applying the text to a deep learning neural network to generate a plurality of bias coordinates defining a point in an embedded space, wherein the embedded space is a multi-dimensional vector space defined by the plurality of bias coordinates, each dimension of the embedded space corresponding to a respective bias and a size of each bias coordinate corresponding to an amount of bias; and
in response to determining that at least one of the plurality of bias coordinates satisfies a bias threshold, providing an indication of bias to the user.
2. The method of claim 1 , wherein obtaining the text from the user includes receiving user input of text into at least one of an instant messaging application, a text messaging application, an e-mail application, or a word processing application.
3. The method of claim 1 , wherein the text includes a plurality of words.
4. The method of claim 1 , wherein the text includes one or more emojis.
5. The method of claim 1 , wherein applying the text to the deep learning neural network includes embedding the text at the point in the embedded space bias.
6. The method of claim 1 , wherein the deep learning neural network includes a plurality of neural network layers.
7. The method of claim 1 , wherein the plurality of bias coordinates includes at least one of a gender bias coordinate, a temporal bias coordinate, a locational bias coordinate, an emotional bias coordinate, or a technological bias coordinate.
8. The method of claim 1 , further comprising training the deep learning neural network with training data by an unsupervised training method and assigning user-comprehensible meaning to one or more of the plurality of bias coordinates.
9. The method of claim 1 , wherein an indication of a respective bias is provided to the user in response to determining that a respective one of the plurality of bias coordinates exceeds a threshold.
10. The method of claim 1 , further comprising, providing a de-biased version of a string to text to the user.
11. The method of claim 1 , further comprising obtaining biometric sensor data regarding the user and further applying the biometric sensor data to the deep learning neural network to generate the plurality of bias coordinates.
12. The method of claim 1 , further comprising parsing the text into a plurality of subsections of text, wherein applying the text to the deep learning neural network includes applying one or more of the plurality of subsections of text to the deep learning neural network to generate one or more respective pluralities of bias coordinates.
13. A system comprising:
one or more processors; and
a non-transitory memory comprising instructions that when executed cause the one or more processors to perform operations comprising:
obtaining text from a user;
applying the text to a deep learning neural network to generate a plurality of bias coordinates defining a point in an embedded space, wherein the embedded space is a multi-dimensional vector space defined by the plurality of bias coordinates, each dimension of the embedded space corresponding to a respective bias and a size of each bias coordinate corresponding to an amount of bias; and
in response to determining that at least one of the plurality of bias coordinates satisfies a bias threshold, providing an indication of bias to the user.
14. The system of claim 13 , wherein obtaining the text from the user includes receiving user input of text into at least one of an instant messaging application, a text messaging application, an e-mail application, or a word processing application executed by the one or more processors.
15. The system of claim 13 , wherein the text includes one or more emojis.
16. The system of claim 13 , wherein the operations further comprise training the deep learning neural network with training data by an unsupervised training method and assigning user-comprehensible meaning to one or more of the plurality of bias coordinates.
17. The system of claim 13 , wherein the operations further comprise providing a de-biased version of a string to text to the user.
18. The system of claim 13 , further comprising obtaining biometric sensor data regarding the user and further applying the biometric sensor data to the deep learning neural network to generate the plurality of bias coordinates.
19. A system comprising:
means for obtaining text from a user;
means for applying the text to a deep learning neural network to generate a plurality of bias coordinates defining a point in an embedded space, wherein the embedded space is a multi-dimensional vector space defined by the plurality of bias coordinates, each dimension of the embedded space corresponding to a respective bias and a size of each bias coordinate corresponding to an amount of bias; and
means for, in response to determining that at least one of the plurality of bias coordinates satisfies a bias threshold, providing an indication of bias to the user.
20. The system of claim 19 , further comprising means for obtaining biometric sensor data regarding the user and further applying the biometric sensor data to the deep learning neural network to generate the plurality of bias coordinates.