AI Log Processing for Automatic Multi-Line Timestamp Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing log file processing techniques require prior user input specifying timestamp patterns, making them inefficient and prone to human error, especially when dealing with complex apparatuses that generate multi-line log messages.

Innovation Solution

A data-driven processing method using an AI model, such as a large language model (LLM), to automatically detect timestamp formats and regex patterns in log files without prior user input, enabling extraction of timeseries data from single-line and multi-line messages.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If template-based timestamp detection is used, then timestamp extraction can be performed, but prior user input specifying string patterns is required which reduces efficiency and increases complexity

Engineering Contradiction:
Improvetimestamp extraction accuracyVSAvoidprocessing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The processing system automatically detects timestamp formats and regex patterns by analyzing the log file itself, without requiring external user input. The system services itself by autonomously identifying patterns such as timestamps, log messages, and their structures through internal analysis algorithms.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system performs preliminary analysis of the log file to detect timestamp formats and patterns before actual timestamp extraction begins. This preliminary detection phase identifies string patterns, regular expressions, and temporal information structures that are then used for efficient extraction.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If manual pattern specification is used for timestamp detection, then timestamp extraction can be performed, but human error increases and reliability decreases

Engineering Contradiction:
Improvetimestamp detection accuracyVSAvoidprocessing reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The system eliminates human involvement in pattern specification by automatically detecting and identifying timestamp formats, log message structures, and regex patterns through internal analysis, thereby removing the source of human error.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system uses feedback from analyzing multiple log entries to refine and confirm timestamp pattern detection. By examining the structure and content of log messages, the system validates detected patterns against actual data, ensuring accurate and reliable timestamp extraction.

Inventive Principle:
Principle #23Feedback

3Reliability

If log files from complex apparatuses are processed, then comprehensive monitoring is achieved, but log file size increases making analysis challenging

Engineering Contradiction:
Improveapparatus monitoring reliabilityVSAvoidlog file size
Core Design Contradiction:
ReliabilityVSQuantity of substance

Solution Approach 1:

The system extracts only the essential temporal information (timestamps) and relevant log message content from large log files, separating useful data from voluminous raw log data. This extraction process converts large log files into compact timeseries datasets suitable for analysis.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system performs preliminary detection and classification of log message structures before extraction, identifying which portions of large log files contain meaningful temporal and event data versus redundant information, enabling efficient processing of comprehensive monitoring data.

Inventive Principle:
Principle #10Preliminary action

Data Source

PatentEP4597325A1Method and processing system for processing a log file of at least one apparatus
Publication Date: 2025.08.06 CARL ZEISS AG
  • EP4597325A1 patent drawingFigure 1~2
  • EP4597325A1 patent drawingFigure 3
  • EP4597325A1 patent drawingFigure 4~5

AI summary

To process a log file of at least one apparatus (14), a processing system (20) obtains the log file. The processing system (20) identifies one or several multi-line log messages in the log file. The processing system (20) detects timestamps in several log messages of the log file, comprising communicatively interfacing with a data-driven processing model (12) to obtain timestamp-related information. The processing system (20) extracts timeseries data based at least on the timestamp-related information and performs at least one operation based on the timeseries data.