Text correction for dyslexic users on an online social network

• US 10,489,507 B2
• Filed: 01/02/2018
• Issued: 11/26/2019
• Est. Priority Date: 01/02/2018
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Li, Xian, Veliche, Irina-Elena, Sur, Debnil, Wu, Shaomei, Bahl, Amit, Pino, Juan Miguel
• Primary Examiner(s): Flanders, Andrew C
• Application Number:
  US15/860,362
• Publication Number:
  US 20190205372A1
• Time in Patent Office: 693 Days
• CPC Class Codes: G06F 3/0481 based on specific properties of the displayed interaction object or a metaphor-based environment, e.g. interaction with desktop elements like windows or icons, or assisted by a cursor's changing behaviour or appearance
• CPC Class Codes: G06F 40/166 Editing, e.g. inserting or deleting
• CPC Class Codes: G06F 40/232 Orthographic correction, e.g. spell checking or vowelisation
• CPC Class Codes: G06F 40/253 Grammatical analysis; Style critique
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 3/044 Recurrent networks, e.g. Hopfield networks
• CPC Class Codes: G06N 3/045 Combinations of networks
• CPC Class Codes: G06N 5/022 Knowledge engineering; Knowledge acquisition

Abstract

In one embodiment, a method includes identifying a plurality of dyslexic users on an online social network. The plurality of dyslexic users may be identified based on content objects posted by these users over a particular time period, where the content objects may include one or more of word-level errors or sentence-level errors. A machine-learning model may be trained for text correction using a corpus of social network data, which may include at least the content objects with one or more of word-level errors or sentence-level errors, and a corresponding set of corrected content objects. A text string including one or more errors may be received from a client system associated with a first user. The text string may be transformed into a vector representation using an encoder of the machine-learning model. A corrected text string may be generated from the vector representation using a decoder of the machine-learning model.

Claims

1. A method comprising, by one or more computing systems:

identifying a plurality of dyslexic users on an online social network, wherein the plurality of dyslexic users are identified based on a set of content objects posted by the dyslexic users over a particular time period, the content objects posted by the dyslexic users comprising one or more of word-level errors or sentence-level errors;

training a machine-learning model for text correction using a corpus of social network data, the social network data comprising at least the set of content objects posted by the dyslexic users with one or more of word-level errors or sentence-level errors, and a corresponding set of corrected content objects that are posted to replace the posted set of content objects;

receiving, from a client system associated with a first user of an online social network, a text string, the text string comprising one or more errors;

transforming, using an encoder of the machine-learning model, the text string into a vector representation;

generating, using a decoder of the machine-learning model, a corrected text string from the vector representation, wherein the corrected text string has the one or more errors removed; and

sending, to the client system associated with the first user, instructions for presenting the corrected text string.

2. The method of claim 1, wherein the set of content objects posted by the dyslexic users comprises social media posts.

3. The method of claim 2, wherein:

a content object, posted by a dyslexic user from the plurality of dyslexic users, with one or more of word-level errors or sentence-level errors comprises a first social media post; and

a corresponding corrected content object comprises a second social media post submitted by the dyslexic user to correct the first social media post.

4. The method of claim 1, further comprising:

determining that the first user is dyslexic.

5. The method of claim 1, wherein the word-level errors comprises one or more of:

words with swapping letters;

words with omitting letters;

foreshortening words;

words omitting prefixes or suffixes;

or

words with capitalization issues.

6. The method of claim 1, wherein the sentence-level errors comprises one or more of:

grammar errors;

wrong prepositions;

odd or no spacing between words;

sentence fragment issues;

or

wrong punctuations.

7. The method of claim 1, further comprising:

receiving, from the client system associated with the first user of the online social network, a request to post the text string on the online social network, wherein the corrected text string is presented on the client system as a suggestion to the first user in response to receiving the request.

8. The method of claim 7, further comprising:

receiving, from the client system associated with the first user of the online social network, a confirmation from the first user accepting the corrected text string; and

posting the corrected text string on the online social network.

9. The method of claim 1, further comprising:

generating, using the machine-learning model, suggestions for correcting the one or more errors in the text string in real-time.

10. The method of claim 1, wherein transforming the text string into the vector representation preserves the semantic meaning of the text string.

11. The method of claim 1, further comprising:

in response to receiving the text string, analyzing the text; and

categorizing the text, based on the analysis, into one of a non-English text, a well-written text, an understandable but informal text, and a poorly-written text.

12. The method of claim 11, further comprising:

determining that the text is categorized into the poorly-written text.

13. The method of claim 1, wherein the machine-learning model is a sequence-to-sequence model.

14. The method of claim 1, further comprising:

parsing the text string into one or more n-grams, wherein the one or more n-grams comprises the one or more errors.

15. The method of claim 1, wherein the one or more n-grams are one or more of unigrams, bigrams, or trigrams.

16. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

identify a plurality of dyslexic users on an online social network, wherein the plurality of dyslexic users are identified based on a set of content objects posted by the dyslexic users over a particular time period, the content objects posted by the dyslexic users comprising one or more of word-level errors or sentence-level errors;

train a machine-learning model for text correction using a corpus of social network data, the social network data comprising at least the set of content objects posted by the dyslexic users with one or more of word-level errors or sentence-level errors, and a corresponding set of corrected content objects that are posted to replace the posted set of content objects;

receive, from a client system associated with a first user of an online social network, a text string, the text string comprising one or more errors;

transform, using an encoder of the machine-learning model, the text string into a vector representation;

generate, using a decoder of the machine-learning model, a corrected text string from the vector representation, wherein the corrected text string has the one or more errors removed; and

send, to the client system associated with the first user, instructions for presenting the corrected text string.

17. The media of claim 16, wherein the set of content objects posted by the dyslexic users comprises social media posts.

18. The media of claim 17, wherein:

a content object, posted by a dyslexic user from the plurality of dyslexic users, with one or more of word-level errors or sentence-level errors comprises a first social media post; and

a corresponding corrected content object comprises a second social media post submitted by the dyslexic user to correct the first social media post.

19. A system comprising:

one or more processors; and

a non-transitory memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to;

identify a plurality of dyslexic users on an online social network, wherein the plurality of dyslexic users are identified based on a set of content objects posted by the dyslexic users over a particular time period, the content objects posted by the dyslexic users comprising one or more of word-level errors or sentence-level errors;

train a machine-learning model for text correction using a corpus of social network data, the social network data comprising at least the set of content objects posted by the dyslexic users with one or more of word-level errors or sentence-level errors, and a corresponding set of corrected content objects that are posted to replace the posted set of content objects;

receive, from a client system associated with a first user of an online social network, a text string, the text string comprising one or more errors;

transform, using an encoder of the machine-learning model, the text string into a vector representation;

generate, using a decoder of the machine-learning model, a corrected text string from the vector representation, wherein the corrected text string has the one or more errors removed; and

send, to the client system associated with the first user, instructions for presenting the corrected text string.

20. The media of claim 19, wherein the set of content objects posted by the dyslexic users comprises social media posts.