Pictorial symbol prediction
Symbol prediction can be implemented using a multi-task system trained for different tasks. The tasks may include a single symbol prediction, symbol category prediction, and symbol subcategory prediction. Categories of symbols can be generated by clustering sets of training data using a clustering scheme.
1. A method comprising:
identifying one or more text characters input into a client device;
generating a plurality of pictorial symbol classifications from the one or more text characters, the plurality of pictorial symbol classifications including a pictorial symbol category classification for a group of pictorial symbols of a similar type and an individual pictorial symbol classification;
determining that one of the plurality of pictorial symbol classifications satisfies a specified threshold;
displaying, on a display device of the client device, a presentation of a single display element based on the individual pictorial symbol classification exceeding a specified individual pictorial symbol classification threshold; and
displaying, on the display device of the client device, a presentation of two or more display elements that are associated with the pictorial symbol category classification based on the pictorial symbol category classification exceeding a specified a pictorial symbol category classification threshold, the pictorial symbol category classification threshold being less than the specified individual pictorial symbol classification threshold.
2. The method of claim 1 , wherein the generating is accomplished using a machine learning scheme that is a multitask neural network having a first network portion configured to generate individual pictorial symbol classifications and a second network portion configured to generate pictorial symbol category classifications.
3. The method of claim 2 , wherein the machine learning scheme is trained using messages or posts that have a caption and at least one pictorial symbol.
4. The method of claim 2 , wherein the multitask neural network further comprises a third network portion configured to generate pictorial symbol subcategory classifications.
5. The method of claim 2 , wherein the machine learning scheme generates the plurality of pictorial symbol classifications in parallel.
6. The method of claim 2 , further comprising:
generating categories of the pictorial symbol classifications using an additional machine learning scheme configured to cluster similar types of pictorial symbols.
7. The method of claim 6 , wherein the additional machine learning scheme implements k-means clustering to group similar types of pictorial symbols.
8. The method of claim 7 , wherein the additional machine learning scheme is trained using a plurality of network published posts, each of the plurality of network published posts comprising one or more text characters and a pictorial symbol.
9. The method of claim 1 , wherein the one or more text characters are input into the client device by:
receiving voice data at the client device; and
performing speech recognition on the voice data to identify words in the voice data.
10. The method of claim 1 wherein the presentation of two or more display elements comprises a first window for displaying an individual symbol and a second window for displaying at least one symbol in a subcategory or category of symbols.
11. The method of claim 1 , wherein the display elements include one or more of: an emoji, an emoticon, or an image.
12. A system comprising:
one or more processors of a machine; and
a memory storing instructions that, when executed by at least one processor among the one or more processors, causes the machine to perform operations comprising:
identifying one or more text characters input into a client device;
generating a plurality of pictorial symbol classifications from the one or more text characters, the plurality of pictorial symbol classifications including a pictorial symbol category classification for a group of pictorial symbols of a similar type and an individual pictorial symbol classification;
determining that one of the plurality of pictorial symbol classifications satisfies a specified threshold;
displaying, on a display device of the client device, a presentation of a single display element based on the individual pictorial symbol classification exceeding a specified individual pictorial symbol classification threshold; and
displaying, on the display device of the client device, a presentation of two or more display elements that are associated with the pictorial symbol category classification based on the pictorial symbol category classification exceeding a specified a pictorial symbol category classification threshold, the pictorial symbol category classification threshold being less than the specified individual pictorial symbol classification threshold.
13. The system of claim 12 , wherein the generating is accomplished using a machine learning scheme that is a multitask neural network having a first network portion configured to generate individual pictorial symbol classifications and a second network portion configured to generate pictorial symbol category classifications.
14. The method of claim 13 , wherein the machine learning scheme generates the plurality of pictorial symbol classifications in parallel.
15. The system of claim 12 wherein the one or more text characters are input into the client device by:
receiving voice data at the client device; and
performing speech recognition on the voice data to identify words in the voice data.
16. The system of claim 12 wherein the presentation of two or more display elements comprises a first window for displaying an individual symbol and a second window for displaying at least one symbol in a subcategory or category of symbols.
17. A machine-readable storage device embodying instructions that, when executed by a device, cause the device to perform operations comprising:
identifying one or more text characters input into a client device;
generating a plurality of pictorial symbol classifications from the one or more text characters, the plurality of pictorial symbol classifications including a pictorial symbol category classification for a group of pictorial symbols of a similar type and an individual pictorial symbol classification;
determining that one of the plurality of pictorial symbol classifications satisfies a specified threshold;
displaying, on a display device of the client device, a presentation of a single display element based on the individual pictorial symbol classification exceeding a specified individual pictorial symbol classification threshold; and
displaying, on the display device of the client device, a presentation of two or more display elements that are associated with the pictorial symbol category classification based on the pictorial symbol category classification exceeding a specified a pictorial symbol category classification threshold, the pictorial symbol category classification threshold being less than the specified individual pictorial symbol classification threshold.
18. The machine-readable storage device of claim 17 , wherein the generating is accomplished using a machine learning scheme that is a multitask neural network having a first network portion configured to generate individual pictorial symbol classifications and a second network portion configured to generate pictorial symbol category classifications.
19. The machine-readable storage device of claim 17 , wherein the one or more text characters are input into the client device by:
receiving voice data at the client device; and
performing speech recognition on the voice data to identify words in the voice data.
20. The machine-readable storage device of claim 17 , wherein the presentation of two or more display elements comprises a first window for displaying an individual symbol and a second window for displaying at least one symbol in a subcategory or category of symbols.