	US 12,105,749 B2
	Method for generating triples from log entries
Georgia Olympia Brikis, Plainsboro, NJ (US); Dmitry Fradkin, Wayne, PA (US); Vladimir Lavrik, Hessen (DE); Serghei Mogoreanu, Munich (DE); and André Scholz, Ansbach (DE)
Assigned to Siemens Aktiengesellschaft, Munich (DE)
Appl. No. 17/782,040
Filed by Siemens Aktiengesellschaft, Munich (DE)
PCT Filed Dec. 10, 2020, PCT No. PCT/EP2020/085402 § 371(c)(1), (2) Date Jun. 2, 2022, PCT Pub. No. WO2021/116240, PCT Pub. Date Jun. 17, 2021.
Claims priority of application No. 19215894 (EP), filed on Dec. 13, 2019.
Prior Publication US 2023/0004591 A1, Jan. 5, 2023
Int. Cl. G06F 16/00 (2019.01); G06F 16/35 (2019.01); G06F 17/40 (2006.01); G06N 5/022 (2023.01)

CPC G06F 16/355 (2019.01) [G06F 17/40 (2013.01); G06N 5/022 (2013.01)]

6 Claims

1. A computer-implemented method for generating triples from log entries, the method comprising:

providing a plurality of log entries from respective log files generated during operation of one or more industrial plants, wherein each log entry of the plurality of log entries comprises at least one text message;

generating at least one template based on the plurality of log entries using unsupervised clustering, wherein the at least one template comprises at least one variable part and at least one fixed part;

assigning each log entry of the plurality of log entries to one respective template based on the generated at least one template using a similarity measure;

extracting the corresponding at least one variable and at least one fixed part of each text message of the plurality of text messages as key/value pairs using the respective assigned at least one template based on the plurality of log entries;

providing the text messages, keys and values as triples; and

loading the triples into a knowledge graph for analyzing the one or more industrial plants.