# US 2019/0295527 A1 · Analyzing messages with typographic errors due to phonemic spellings using text-to-speech and speech-to-text algorithms

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0295527 A1
- Document ID: US20190295527A1
- Status: Published application
- Kind: A1
- Application number: 15/925,888
- Filed: 2018-03-20
- Published: 2019-09-26
- Earliest claimed: 2018-03-20
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Meenal Pore; David Moinina Sengeh
- Web page: https://patentgrep.com/patent/US20190295527A1

## CPC

- G10L 13/086: {Detection of language}
- G06F 40/232: Orthographic correction, e.g. spell checking or vowelisation
- G06F 40/263: Language identification
- G10L 13/00: Speech synthesis; Text to speech systems
- G10L 15/26: Speech to text systems (G10L15/08 takes precedence)
- G10L: SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Correcting typographical errors in electronic text may include converting a text message containing at least one phonemic spelling of a word into speech by running a text-to-speech application programming interface (API) with the text message as input. The converted speech may be input to a speech-to-text API and the speech-to-text API executed to convert the speech to text. A text file comprising the text may be generated and/or output. The text file automatically contains a corrected version of the phonemic spelling of the word in text message.

## Claims

### Claim 1

A computer-implemented method of correcting typographical errors in electronic text, the method executed by at least one hardware processor, the method comprising:
receiving a text message containing at least one phonemic spelling of a word;
converting the text message into speech by running a text-to-speech application programming interface (API) with the text message as input;
generating an audio file comprising at least the speech;
playing the audio file as an input to a speech-to-text API and executing the speech-to-text API to convert the speech to a text corresponding to the played audio file; and
generating a text file comprising at least the text corresponding to the played audio file,
wherein the text corresponding to the played audio file automatically contains a corrected version of the phonemic spelling of the word in the received text message.

### Claim 2 (depends on 1)

The method of claim 1, further comprising removing special vocabulary from the text message prior to inputting the text message into the text-to-speech API.

### Claim 3 (depends on 1)

The method of claim 1, further comprising selecting the text-to-speech API based on a language accent detected in the text message.

### Claim 4 (depends on 3)

The method of claim 3, wherein the selecting of the text-to-speech API based on a language accent detected in the text message comprises at least inputting the text message to a machine learning model to classify the text message into a geographic location and selecting the text-to-speech API that is mapped to the geographic location.

### Claim 5 (depends on 1)

The method of claim 1, further comprising selecting the speech-to-text API based on a language accent detected in the text message.

### Claim 6 (depends on 5)

The method of claim 5, wherein the selecting of speech-to-text API based on a language accent detected in the text message comprises at least inputting the text message to a machine learning model to classify the text message into a geographic location and selecting the text-to-speech API that is mapped to the geographic location.

### Claim 7 (depends on 1)

The method of claim 1, further comprising training a machine learning model with a training data set comprising at least input strings in international phonemic alphabet and mapped locations, the machine learning model trained to classify geographic locations given an input string comprising at least a phonetically spelled word.

### Claim 8 (depends on 7)

The method of claim 7, wherein the machine learning model comprises at least a convolutional neural network.

### Claim 9 (depends on 1)

The method of claim 1, further comprising selecting the text-to-speech API based on context of the text message.

### Claim 10 (depends on 1)

The method of claim 1, further comprising selecting the speech-to-text API based on context of the text message.

### Claim 11

A system of correcting typographical errors in electronic text, comprising:
at least one hardware processor;
a memory device coupled with the at least one hardware processor;
the at least one hardware processor operable to at least:
receive a text message containing at least one phonemic spelling of a word;
convert the text message into speech by running a text-to-speech application programming interface (API) with the text message as input;
generate an audio file comprising at least the speech;
play the audio file as an input to a speech-to-text API and execute the speech-to-text API to convert the speech to a text corresponding to the played audio file; and
generate a text file comprising at least the text corresponding to the played audio file,
wherein the text corresponding to the played audio file automatically contains a corrected version of the phonemic spelling of the word in the received text message.

### Claim 12 (depends on 11)

The system of claim 11, wherein the at least one hardware processor is further operable to remove special vocabulary from the text message prior to inputting the text message into the text-to-speech API.

### Claim 13 (depends on 11)

The system of claim 11, wherein the at least one hardware processor is further operable to select the text-to-speech API based on a language accent detected in the text message.

### Claim 14 (depends on 13)

The system of claim 13, wherein the at least one hardware processor selecting the text-to-speech API based on a language accent detected in the text message comprises at least inputting the text message to a machine learning model to classify the text message into a geographic location and selecting the text-to-speech API that is mapped to the geographic location.

### Claim 15 (depends on 11)

The system of claim 11, wherein the at least one hardware processor is operable to select the speech-to-text API based on a language accent detected in the text message.

### Claim 16 (depends on 15)

The system of claim 15, wherein the at least one hardware processor selecting speech-to-text API based on a language accent detected in the text message comprises at least inputting the text message to a machine learning model to classify the text message into a geographic location and selecting the text-to-speech API that is mapped to the geographic location.

### Claim 17 (depends on 11)

The system of claim 11, wherein the at least one hardware processor is further operable to receive a training data set comprising at least input strings in international phonemic alphabet and corresponding locations, and train a machine learning model by running a machine learning algorithm with the training data set, to classify geographic locations given an input string comprising at least a phonetically spelled word.

### Claim 18 (depends on 17)

The system of claim 17, wherein the machine learning model comprises at least a convolutional neural network.

### Claim 19

A computer program product for correcting typographical errors in electronic text, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, the program instructions readable by a processor to cause the processor to perform a method comprising:
receiving a text message containing at least one phonemic spelling of a word;
converting the text message into speech by running a text-to-speech application programming interface (API) with the text message as input;
generating an audio file comprising at least the speech;
playing the audio file as an input to a speech-to-text API and executing the speech-to-text API to convert the speech to a text corresponding to the played audio file; and
generating a text file comprising at least the text corresponding to the played audio file,
wherein the text corresponding to the played audio file automatically contains a corrected version of the phonemic spelling of the word in the received text message.

### Claim 20 (depends on 19)

The computer program product of claim 19, wherein the method further comprises selecting the text-to-speech API and the speech-to-text API based on a language accent detected in the text message.
