# US 2026/0300230 A1 · Log compression and recovery

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

- Number: US 2026/0300230 A1
- Document ID: US20260300230A1
- Status: Published application
- Kind: A1
- Application number: 19/092,461
- Filed: 2025-03-27
- Published: 2026-10-01
- Earliest claimed: 2025-03-27
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Hui Wang; Mai Zeng; Peng Hui Jiang; Yu Mei Dai; Xiang Yu Xue; Wei Li; Xiao Chen Huang
- Web page: https://patentgrep.com/patent/US20260300230A1

## CPC

- G06F 16/1744: {using compression, e.g. sparse files}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Embodiments receive raw data from an external application, create structured log data by removing noise from the received raw data, perform event merging on the structured log data, perform pattern extraction on the event merged structured log data, create sparse log data by performing field reduction on the pattern extraction, create compressed logs by applying a first set of large language model (LLM) compression on the sparse log data, and create recovered logs by applying a second set of LLM recovery on the compressed logs.

## Claims

### Claim 1

A computer-implemented method, comprising:
receiving raw data from an external application;
creating structured log data by removing noise from the received raw data, wherein the removing the noise includes normalizing the received raw data by calculating a logarithm of each value in the received raw data;
performing event merging on the structured log data by combining similar log events into a combined log, wherein the similarity between the log events is determined based on a predefined threshold;
performing pattern extraction on the event merged structured log data;
creating sparse log data by performing field reduction on the pattern extraction;
reducing a size of the sparse log data by creating compressed logs, wherein the compressed logs are created by applying a first set of large language model (LLM) compression on the sparse log data; and
validating a log structure of the compressed logs.

### Claim 2 (depends on 1)

The method of claim 1, wherein the external application comprises a software application from a mainframe operating system.

### Claim 3 (depends on 1)

The method of claim 1, wherein the received raw data comprises log data.

### Claim 4 (depends on 3)

The method of claim 3, wherein the log data comprises syslog data.

### Claim 5 (depends on 3)

The method of claim 3, wherein the log data comprises operlog data.

### Claim 6 (depends on 1)

The method of claim 1, further comprising extracting at least one of a timestamp, a log level, and a message from the received raw data.

### Claim 7 (depends on 1)

The method of claim 1, wherein the performing the event merging on the structured log data comprises detecting and combining duplicate events within the structured log data.

### Claim 8

(canceled)

### Claim 9 (depends on 1)

The method of claim 1, wherein the performing the pattern extraction comprises identifying recurring log patterns from text within log entries of the event merged structured log data.

### Claim 10 (depends on 1)

The method of claim 1, wherein the performing the field reduction on the pattern extraction comprises removing redundant fields from the pattern extraction.

### Claim 11 (depends on 1)

The method of claim 1, wherein the creating the compressed logs by applying the first set of LLM compression comprises utilizing a first LLM model on the sparse log data.

### Claim 12 (depends on 11)

The method of claim 11, wherein the first LLM model comprises a generalized language model (GLM).

### Claim 13 (depends on 1)

The method of claim 1, further comprising:
creating recovered logs by applying a second set of LLM recovery, wherein the creating the recovered logs comprises utilizing a second LLM model on the compressed logs.

### Claim 14 (depends on 13)

The method of claim 13, wherein the second LLM model comprises a generalized language model (GLM).

### Claim 15

A computer program product comprising:
one or more computer-readable storage media; and
program instructions stored on the one or more computer-readable storage media to perform operations comprising:
receiving raw data from an external application;
creating structured log data by removing noise from the received raw data. wherein the removing the noise includes normalizing the received raw data by calculating a logarithm of each value in the received raw data;
performing event merging on the structured log data by combining similar log events into a combined log, wherein the similarity between the log events is determined based on a predefined threshold;
performing pattern extraction on the event merged structured log data;
creating sparse log data by performing field reduction on the pattern extraction;
reducing a size of the sparse log data by creating compressed logs, wherein the compressed logs are created by applying a first set of large language model (LLM) compression on the sparse log data; and
validating a log structure of the compressed logs.

### Claim 16 (depends on 15)

The computer program product of claim 15, wherein the performing the event merging of the structured log data comprises detecting and combining duplicate events within the structured log data.

### Claim 17

(canceled)

### Claim 18 (depends on 15)

The computer program product of claim 15, wherein:
the performing the pattern extraction comprises identifying recurring log patterns from text within log entries of the event merged structured log data; and
the performing the field reduction on the pattern extraction comprises removing redundant fields from the pattern extraction.

### Claim 19 (depends on 15)

The computer program product of claim 15, further comprising:
utilizing a first LLM model on the sparse log data for creating the compressed logs; and
utilizing a second LLM model on the compressed logs for recovering logs.

### Claim 20

A computer system comprising:
a processor set;
one or more computer-readable storage media; and
program instructions stored on the one or more computer-readable storage media to cause the processor set to perform operations comprising:
receiving raw data from an external application;
creating a first set of structured log data by removing noise from the received raw data;
performing event merging on the first set of structured log data;
performing pattern extraction on the event merged first set of structured log data;
creating sparse log data by performing field reduction on the pattern extraction;
creating compressed logs by applying a first set of large language model (LLM) compression on the sparse log data;
creating a second set of structured log data by extracting key components and organizing data of the compressed logs; and
creating recovered logs by applying a second set of LLM recovery on the second set of structured log data.
