Text formatter
Methods, systems, and computer programs are presented for formatting raw text. One method includes an operation for accessing raw text comprising words corresponding to one or more sentences. The raw text is lowercase text without any punctuation. Further, the method includes operations for creating a plurality of sub-words corresponding to the raw text, and for generating, by a machine-learning (ML) model, an output for each sub-word based on the created sub-words. The output for each sub-word indicates a formatting operation for the corresponding sub-word. The method further includes an operation for generating, based on the formatting operations in the outputs for the sub-words, formatted text corresponding to the raw text. The formatted text is text with correct grammar, proper punctuation, and proper capitalization according to a meaning of words spoken by a speaker associated with the raw text.
1. A method comprising:
accessing, by one or more processors, raw text that includes words;
determining, by the one or more processors, a plurality of parts of the words based on the words in the raw text;
generating, by one or more machine-learning (ML) models, a vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text; and
generating, by the one or more processors, formatted text based on the generated vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text.
2. The method of claim 1 , wherein:
the accessed raw text is unformatted text without any punctuation;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate punctuation to be applied to determined plurality of parts of words; and
the generated formatted text is punctuated correctly and capitalized correctly.
3. The method of claim 1 , wherein:
an ML model among the one or more ML models is trained based on training data that includes paragraphs with correct punctuation and correct capitalization.
4. The method of claim 1 , wherein:
the accessed raw text includes a word spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate joining spelled letters into words; and
the generated formatted text includes a joined word.
5. The method of claim 1 , wherein:
the accessed raw text includes an acronym spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate capitalization of acronyms; and
the generated formatted text includes a capitalized acronym.
6. The method of claim 1 , wherein:
the accessed raw text includes a phone number;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate formatting of phone numbers; and
the generated formatted text includes a formatted phone number.
7. The method of claim 1 , wherein:
the accessed raw text includes a uniform resource locator (URL);
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate formatting of URLs; and
the generated formatted text includes a formatted URL.
8. A system comprising:
one or more processors; and
a memory storing instructions that, when executed by the one or more processors, cause the system to perform operations comprising:
accessing raw text that includes words;
determining a plurality of parts of the words based on the words in the raw text;
generating, by one or more machine-learning (ML) models, a vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text; and
generating formatted text based on the generated vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text.
9. The system of claim 8 , wherein:
the accessed raw text is unformatted text without any punctuation;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate punctuation to be applied to determined plurality of parts of the words in the raw text; and
the generated formatted text is punctuated correctly and capitalized correctly.
10. The system of claim 8 , wherein:
an ML model among the one or more ML models is trained based on training data that includes paragraphs with correct punctuation and correct capitalization.
11. The system of claim 8 , wherein:
the accessed raw text includes a word spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate joining spelled letters into words; and
the generated formatted text includes a joined word.
12. The system of claim 8 , wherein:
the accessed raw text includes an acronym spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate capitalization of acronyms; and
the generated formatted text includes a capitalized acronym.
13. The system of claim 8 , wherein:
the accessed raw text includes a phone number;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate formatting of phone numbers; and
the generated formatted text includes a formatted phone number.
14. The system of claim 8 , wherein:
the accessed raw text includes a uniform resource locator (URL);
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate formatting of URLs; and
the generated formatted text includes a formatted URL.
15. A non-transitory machine-readable medium comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations comprising:
accessing raw text that includes words;
determining a plurality of parts of the words based on the words in the raw text;
generating, by one or more machine-learning (ML) models, a vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text; and
generating formatted text based on the generated vector of multiple values among which each value specifies a corresponding formatting operation for a different corresponding part within the determined plurality of parts of the words in the raw text.
16. The non-transitory machine-readable medium of claim 15 , wherein:
the accessed raw text is unformatted text without any punctuation;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate punctuation to be applied to determined plurality of parts of the words in the raw text; and
the generated formatted text is punctuated correctly and capitalized correctly.
17. The non-transitory machine-readable medium of claim 15 , wherein:
an ML model among the one or more ML models is trained based on training data that includes paragraphs with correct punctuation and correct capitalization.
18. The non-transitory machine-readable medium of claim 15 , wherein:
the accessed raw text includes a word spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate joining spelled letters into words; and
the generated formatted text includes a joined word.
19. The non-transitory machine-readable medium of claim 15 , wherein:
the accessed raw text includes an acronym spelled letter by letter;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate capitalization of acronyms; and
the generated formatted text includes a capitalized acronym.
20. The non-transitory machine-readable medium of claim 15 , wherein:
the accessed raw text includes a phone number;
an ML model among the one or more ML models is trained to output one or more formatting operations that indicate formatting of phone numbers; and
the generated formatted text includes a formatted phone number.