# US 2021/0342530 A1 · Framework for Managing Natural Language Processing Tools

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

- Number: US 2021/0342530 A1
- Document ID: US20210342530A1
- Status: Published application
- Kind: A1
- Application number: 17/361,073
- Filed: 2021-06-28
- Published: 2021-11-04
- Earliest claimed: 2019-12-31
- Assignees: Paypal, Inc.
- Applicants: Paypal, Inc.
- Inventors: Yuehao Wu; Rajesh Munavalli; Junhua Zhao; Xin Chen; Meng Zang
- Web page: https://patentgrep.com/patent/US20210342530A1

## CPC

- G06F 40/20: Natural language analysis (semantic analysis of natural language G06F40/30)
- G06F 40/30: Semantic analysis
- G06F 8/35: model driven
- G06F 8/36: Software reuse
- G06F 40/263: Language identification
- G06F 40/284: Lexical analysis, e.g. tokenisation or collocates
- G06F 40/289: Phrasal analysis, e.g. finite state techniques or chunking
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A system performs operations that include receiving, via first computing environment, a request to process text data using a first natural language processing (NLP) model. The operations further include accessing configuration data associated with the NLP model, where the configuration data generated using a domain specific language that supports a plurality of preprocessing modules in a plurality of programming languages. The operations also include selecting, based on the configuration data, one or more preprocessing modules of the plurality of preprocessing modules, generating, based on the configuration data, a preprocessing pipeline using the one or more preprocessing modules, and generating preprocessed text data by inputting the text data into the preprocessing pipeline. The preprocessed text data is provided to the first NLP model.

## Claims

### Claim 1

A system, comprising:
one or more hardware processors; and
a memory storing computer-executable instructions, that in response to execution by the one or more hardware processors, causes the system to perform operations comprising:
receiving, via an online computing environment, a first request to process first data using a first natural language processing (NLP) model and receiving a second request to process second data using a second NLP model;
accessing first model configuration data corresponding to the first NLP model and second model configuration data corresponding to the second NLP model, the first model configuration data and the second model configuration data being generated via a domain specific language in an offline computing environment that is different than the online computing environment;
responsive to the first request, executing the first NLP model including determining, based on the first model configuration data, a first set of preprocessing modules to preprocess the first data; and
responsive to the second request, executing the second NLP model including determining, based on the second model configuration data, a second set of preprocessing modules to preprocess the second data, the second set of preprocessing modules being different than the first set of preprocessing modules;

### Claim 2 (depends on 1)

The system of claim 1, wherein the first set of preprocessing modules are implemented using one or more libraries of a first programming toolkit, and the second set of preprocessing modules are implemented using one or more libraries of a second programming toolkit that is different than the first programming toolkit.

### Claim 3 (depends on 1)

The system of claim 1, wherein the first set of preprocessing modules and the second set of preprocessing modules are different subsets of a plurality of preprocessing modules, the plurality of preprocessing modules comprising at least one of a language detection module, a sentence detection module, a tokenization module, a cleaning module, an annotation module, a normalization module, or an embedding module.

### Claim 4 (depends on 1)

The system of claim 1, wherein the first model configuration data defines a first preprocessing pipeline for the first set of preprocessing modules, and the second model configuration data defines a second preprocessing pipeline for the second set of preprocessing modules.

### Claim 5 (depends on 4)

The system of claim 4, wherein the first set of preprocessing modules includes the same modules as the second set of preprocessing modules, and wherein the first preprocessing pipeline defines a different preprocessing sequence than the second preprocessing pipeline.

### Claim 6 (depends on 1)

The system of claim 1, wherein the executing the first set of preprocessing modules for the first NLP model further comprises inputting the first model configuration data into a software engine that is programmed in the domain specific language.

### Claim 7 (depends on 6)

The system of claim 6, wherein the operations further comprise:
automatically validating, by the software engine programmed in the domain specific language, the first set of preprocessing modules with sample data.

### Claim 8 (depends on 1)

The system of claim 1, wherein the online computing environment comprises at least one different computing characteristic than the offline computing environment, the at least one different computing characteristic including at least one of different computer hardware, a different operating system, a different software version for one or more software libraries, a different database topography, or access to different data.

### Claim 9 (depends on 1)

The system of claim 1, wherein the operations further comprise:
training the first NLP model and the second NLP model in the offline computing environment.

### Claim 10

A method, comprising:
receiving, via first computing environment, a request to process text data using a first natural language processing (NLP) model;
accessing configuration data associated with the NLP model, the configuration data generated using a domain specific language that supports a plurality of preprocessing modules in a plurality of programming languages;
selecting, based on the configuration data, one or more preprocessing modules of the plurality of preprocessing modules, the one or more preprocessing modules being programmed in a first programming language of the plurality of programming languages;
generating, based on the configuration data, a preprocessing pipeline using the one or more preprocessing modules;
generating preprocessed text data by inputting the text data into the preprocessing pipeline; and
providing the preprocessed text data to the first NLP model.

### Claim 11 (depends on 10)

The method of claim 10, wherein the preprocessing pipeline defines a sequence of the one or more preprocessing modules.

### Claim 12 (depends on 10)

The method of claim 10, wherein the one or more preprocessing modules comprise at least one of a language detection module, a sentence detection module, a tokenization module, a cleaning module, an annotation module, a normalization module, or an embedding module.

### Claim 13 (depends on 10)

The method of claim 10, wherein the configuration data is generated in a second computing environment that is different from the first computing environment.

### Claim 14 (depends on 13)

The method of claim 13, wherein the first computing environment and the second computing environment include a processing engine configured to process the configuration data in the domain specific language.

### Claim 15 (depends on 10)

The method of claim 10, further comprising:
generating a second preprocessing pipeline for a second NLP model based on second configuration data, the second preprocessing pipeline including one or more preprocessing modules that are programmed in a second programming language of the plurality of programming languages.

### Claim 16 (depends on 15)

The method of claim 15, wherein the second preprocessing pipeline includes at least one different type of preprocessing module than the preprocessing pipeline.

### Claim 17

A non-transitory computer readable medium storing computer-executable instructions that in response to execution by one or more hardware processors, causes a service provider system to perform operations comprising:
receiving, via first computing environment, a request to process text data using a first natural language processing (NLP) model;
accessing configuration data associated with the NLP model, the configuration data generated using a domain specific language that supports a plurality of preprocessing modules in a plurality of programming languages;
selecting, based on the configuration data, one or more preprocessing modules of the plurality of preprocessing modules, the one or more preprocessing modules being programmed in a first programming language of the plurality of programming languages;
generating, based on the configuration data, a preprocessing pipeline using the one or more preprocessing modules;
generating preprocessed text data by inputting the text data into the preprocessing pipeline; and
providing the preprocessed text data to the first NLP model.

### Claim 18 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein the one or more preprocessing modules comprise at least one of a language detection module, a sentence detection module, a tokenization module, a cleaning module, an annotation module, a normalization module, or an embedding module.

### Claim 19 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein the operations further comprise:
generating a second preprocessing pipeline for a second NLP model based on second configuration data, the second preprocessing pipeline including one or more preprocessing modules that are programmed in a second programming language of the plurality of programming languages.

### Claim 20 (depends on 17)

The non-transitory computer readable medium of claim 17, wherein the configuration data is generated in a second computing environment that is different from the first computing environment.
