Determining states of content characteristics of electronic communications

Patent metadata

Field	Value
Publication number	US-10885279-B2
Application number	US-201816184704-A
Country	US
Kind code	B2
Filing date	Nov 8, 2018
Priority date	Nov 8, 2018
Publication date	Jan 5, 2021
Grant date	Jan 5, 2021

Abstract

Disclosed in some examples are methods, systems, devices, and machine-readable mediums for determining states of content characteristics of electronic messages. In some embodiments, the probability of the states of the content characteristics of electronic messages are determined. Some embodiments determine a scores for states of content characteristics. Some embodiments determine a score for electronic messages for content characteristic diversity and inclusion based on a probability of a gender-bias state, a probability of a gender-neutral state, and a probability of not applicable to gender-bias state or gender-neutral state. In some embodiments the probabilities are determined based on a natural language model that is trained with data structures that relate training phrases to states of content characteristics.

First claim

What is claimed is:

1. A computer-implemented method, the method comprising:
receiving text, the text comprising one or more text phrases;
determining for each of the one or more text phrases a probability of a gender-bias state, a probability of a gender-neutral state, and a probability of a state of not applicable to gender bias or gender neutral, the determining being based on a natural language model trained with data structures, the data structures comprising training phrases indicated as having the gender-bias state, training phrases indicated as having the gender-neutral state, and training phrases indicated as not applicable to the gender-bias state or the gender-neutral state;
determining a score of the text based on the probability of the gender-bias state, the probability of the gender-neutral state, and the probability of the state of not applicable to gender bias or gender neutral for each of the one or more text phrases; and
causing to be displayed on a display of a user an indication of the score of the text.

2. The computer-implemented method of claim 1 , wherein the text is received from an email application and the score is displayed on a user interface of the email application.

3. The computer-implemented method of claim 1 , further comprising:
training the natural language model with the data structures, the data structures comprising the training phrases indicated as having the gender-bias state, the training phrases indicated as having the gender-neutral state, and the training phrases indicated as not applicable to the gender-bias state or the gender-neutral state, wherein the natural language model is a neural network or regression network.

4. The computer-implemented method of claim 1 , wherein the natural language model is a neural network or regression network, and wherein the computer-implemented method further comprises:
training a first neural network or a first regression network with the training phrases indicated as having the gender-bias state;
training a second neural network or a second regression network with the training phrases indicated as having the gender-neutral state; and
training a third neural network or a third regression network with the training phrases indicated as not applicable to the gender-bias state or the gender-neutral state.

5. The computer-implemented method of claim 4 , further comprising:
determining for each of the one or more text phrases the probability of the gender-bias state based on a first matching score of a corresponding text phrase of the one or more text phrases with the first neural network;
determining for each of the one or more text phrases the probability of the gender-neutral state based on a second matching of the corresponding text phrase of the one or more text phrases with the second neural network; and
determining for each of the one or more text phrases the probability of the state of not applicable to gender bias or gender neutral state based on a third matching score of the corresponding text phrase of the one or more text phrases with the third neural network.

6. The computer-implemented method of claim 1 , further comprising:
determining a sentiment score of the text;
determining an adjusted sentiment score of the text based on the sentiment score of the text and the score of the text; and
causing to be displayed on the display of the user an indication of the adjusted sentiment score.

7. The computer-implemented method of claim 1 , further comprising:
determining for each of the one or more text phrases a probability of a fixed mindset state, a probability of a growth mindset state, and a probability of a state of not applicable to fixed mindset or growth mindset, wherein the determining is based on a natural language model trained with second data structures, the second data structures comprising training phrases indicated as having the fixed mindset state, training phrases indicated as having the growth mindset state, and training phrases indicated as not applicable to the fixed mindset state or the growth mindset state;
determining another score of the text based on a probability of the fixed mindset state, a probability of the growth mindset state, and a probability of the state of not applicable to fixed mindset or growth mindset; and
causing to be displayed on the display of the user an indication of the another score of the text.

8. The computer-implemented method of claim 1 , further comprising:
determining the score of the text based on the probability of the gender-bias state, the probability of the gender-neutral state, and the probability of the state of not applicable to gender bias or gender neutral for each of the one or more text phrases, wherein the probability of the gender-bias state is weighted to account more for the score than the probability of the gender-neutral state and the probability of the state of not applicable to gender bias or gender neutral.

9. A non-transitory computer-readable storage medium that stores instructions that when executed by one or more processors of a computing device, cause the one or more processors to perform operations comprising:
receiving text, the text comprising one or more text phrases;
determining for each of the one or more text phrases a probability of a gender-bias state, a probability of a gender-neutral state, and a probability of a state of not applicable to gender bias or gender neutral, the determining being based on a natural language model trained with data structures, the data structures comprising training phrases indicated as having the gender-bias state, training phrases indicated as having the gender-neutral state, and training phrases indicated as not applicable to the gender-bias state or the gender-neutral state;
determining a score of the text based on the probability of the gender-bias state, the probability of the gender-neutral state, and the probability of the state of not applicable to gender bias or gender neutral for each of the one or more text phrases; and
causing to be displayed on a display of a user an indication of the score of the text.

10. The computer-readable storage medium of claim 9 , wherein the text is received from an email application and the score is displayed on a user interface of the email application.

11. The computer-readable storage medium of claim 10 , wherein the operations further comprise:
training the natural language model with the data structures, the data structures comprising the training phrases indicated as having the gender-bias state, the training phrases indicated as having the gender-neutral state, and the training phrases indicated as not applicable to the gender-bias state or the gender-neutral state, wherein the natural language model is a neural network or regression network.

12. The computer-readable storage medium of claim 9 , wherein the natural language model is a neural network or regression network, and wherein the operations further comprise:
training a first neural network or a first regression network with the training phrases indicated as having the gender-bias state;
training a second neural network or a second regression network with the training phrases indicated as having the gender-neutral state; and
training a third neural network or a third regression network with the training phrases indicated as not applicable to the gender-bias state or the gender-neutral state.

13. The computer-readable storage medium of claim 12 , wherein the operations further comprise:
determining for each of the one or more text phrases the probability of the gender-bias state based on a first matching score of a correspo

Assignees

• Microsoft Technology Licensing Llc

Inventors

• Barachha Gajendra Kasturchand

Classifications

• G06N7/01

  Probabilistic graphical models, e.g. probabilistic networks · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G06F40/30 Primary

  Semantic analysis · CPC title