Transliteration of text entry across scripts

• US 10,643,028 B1
• Filed: 07/19/2019
• Issued: 05/05/2020
• Est. Priority Date: 12/21/2016
• Status: Active Grant

• Current Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Original Assignee: Meta Platforms, Inc. (f/k/a Facebook, Inc.)
• Inventors: Pino, Juan Miguel, Funiak, Stanislav, Malpani, Mridul, Lochan, Gaurav
• Primary Examiner(s): Opsasnick, Michael N
• Application Number:
  US16/516,544
• Time in Patent Office: 291 Days
• CPC Class Codes: G06F 3/02 Input arrangements using manually operated switches, e.g. using keyboards or dials
• CPC Class Codes: G06F 40/129 Handling non-Latin characters, e.g. kana-to-kanji conversion
• CPC Class Codes: G06F 40/216 using statistical methods
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/44 Statistical methods, e.g. probability models
• CPC Class Codes: G06F 40/53 Processing of non-Latin text kana-to-kanji conversion G06F40/129; vowelisation G06F40/232

Abstract

Embodiments are disclosed for transliterating text entries across different script systems. A method according to some embodiments includes steps of: receiving an input string in a first script system input using a keyboard; segmenting, using a probabilistic model, the input string into phonemes that correspond to characters or sets of characters in a second script system; converting the phonemes in the first script system into the characters or sets of characters in the second script system, the characters or sets of characters forming a word or a word prefix in the second script system; and outputting the word or the word prefix in the second script system.

Claims

1. A method for transliterating text entries across different script systems, comprising:

receiving an input string in a first script system input using a keyboard;

identifying segmentation candidates in the input string using a probabilistic model, the segmentation candidates dividing the input string into phonemes that correspond to characters or sets of characters in a script system, the probabilistic model based on a probability distribution of a particular phoneme in the first script system being transliterated into a particular character or set of characters in the second script system;

selecting a segmentation from the segmentation candidates using an additional model, the additional model based on a probability of the word or the word prefix in the second script system appearing in prior texts written in the second script system;

converting the phonemes in the first script system into the character or sets of characters in the second script system, the character or sets of characters forming a word or a word prefix in the second script system; and

outputting the word or the word prefix in the second script system.

2. The method of claim 1, wherein the at least one probabilistic model is generated using a machine learning pipeline.

3. The method of claim 1, further comprising:

using an additional model to select the segmentation of the input string, wherein the additional model comprises a likelihood that the respective segmentation candidate is a valid segmentation of the input string.

4. The method of claim 1, further comprising segmenting the input string using a joint probability distribution, the joint probability distribution to be based on the at least one probabilistic model and the additional model used.

5. The method of claim 4 further comprising segmenting the input string using a conditional probability distribution, wherein the conditional probability distribution is based on a summation of joint probability distributions over different segmentation candidates.

6. The method of claim 1, wherein the converting comprises:

converting the phonemes in the first script system that is a Latin script system into the characters or sets of characters in the second script system, the characters or sets of characters forming the word or the word prefix in the second script system that is a non-Latin script system.

7. The method of claim 1, wherein the receiving comprises:

receiving an input string in a first script system input using a hardware or software keyboard for inputting characters in the first script system.

8. A non-transitory machine-readable storage medium storing instructions configured to be executed by a processor to cause the processor to:

receive an input string in a first script system input using a keyboard;

identify segmentation candidates in the input string using a probabilistic model, the segmentation candidates dividing the input string into phonemes that correspond to characters or sets of characters in a script system, the probabilistic model based on a probability distribution of a particular phoneme in the first script system being transliterated into a particular character or set of characters in the second script system;

select a segmentation from the segmentation candidates using an additional model, the additional model based on a probability of the word or the word prefix in the second script system appearing in prior texts written in the second script system;

convert the phonemes in the first script system into the character or sets of characters in the second script system, the character or sets of characters forming a word or a word prefix in the second script system; and

output the word or the word prefix in the second script system.

9. The non-transitory machine-readable storage medium of claim 8, storing instructions configured to be executed by a processor to cause the processor to:

generate the at least one probabilistic model using a machine learning pipeline.

10. The non-transitory machine-readable storage medium of claim 8, storing instructions configured to be executed by a processor to cause the processor to:

use an additional model to select the segmentation of the input string, wherein the additional model comprises a likelihood that the respective segmentation candidate is a valid segmentation of the input string.

11. The non-transitory machine-readable storage medium of claim 8, storing instructions configured to be executed by a processor to cause the processor to:

segment the input string using a joint probability distribution, the joint probability distribution to be based on the at least one probabilistic model and the additional model used.

12. The non-transitory machine-readable storage medium of claim 11, storing instructions configured to be executed by a processor to cause the processor to:

segment the input string using a conditional probability distribution, wherein the conditional probability distribution is based on a summation of joint probability distributions over different segmentation candidates.

13. The non-transitory machine-readable storage medium of claim 8, storing instructions configured to be executed by a processor to cause the processor to:

convert the phonemes in the first script system that is a Latin script system into the characters or sets of characters in the second script system, the characters or sets of characters forming the word or the word prefix in the second script system that is a non-Latin script system.

14. The non-transitory machine-readable storage medium of claim 8, storing instructions configured to be executed by a processor to cause the processor to:

receive an input string in a first script system input using a hardware or software keyboard for inputting characters in the first script system.

15. A computing device, comprising:

a processor; and

a memory storing instructions configured to be executed by a processor to cause the processor to;

identify segmentation candidates in an input string using a probabilistic model, the segmentation candidates dividing the input string into phonemes that correspond to characters or sets of characters in a script system, the probabilistic model based on a probability distribution of a particular phoneme in the first script system being transliterated into a particular character or set of characters in the second script system; and

select a segmentation from the segmentation candidates using an additional model, the additional model based on a probability of the word or the word prefix in the second script system appearing in prior texts written in the second script system;

convert the phonemes in the first script system into the character or sets of characters in the second script system, a word or a word prefix in the second script system comprising the character; and

store the word or the word prefix in the second script system.

16. The computing device of claim 15, the memory storing instructions configured to be executed by a processor to cause the processor to:

output the word or the word prefix in the second script system.