Shi, Meiling ; Hoffmann, André ; Rüppel, Uwe
Hrsg.: Semenov, Vitaly ; Scherer, Raimar J. (2021)
Metadata based multi-class text classification in engineering project platform.
ECPPM 2021 – eWork and eBusiness in Architecture, Engineering and Construction. Moscow, Russia (15-17 September 20021)
doi: 10.1201/9781003191476-29
Konferenzveröffentlichung, Bibliographie

Kurzbeschreibung (Abstract)

Enterprises usually establish digital platforms to store and share projects as well as business data. Those files in digital platforms usually have metadata that gives information about the life of a document. Metadata is useful to retrieve useful information about the document. They can be generated fully automatically or given manually by users. Sometimes those manually given metadata can be left out, insufficient, or even wrong. Those metadata stay unaware until the files are being searched for a purpose later on. If the number of files with insufficient metadata is enormous, the filtering of files from platform according to specific metadata becomes difficult. Manually correction can be tedious and enormously time intensive. This paper introduces a method for correcting faulty metadata based on other correct metadata with text classification, natural language processing technique, and linguistic synonymy information in the engineering project platform. The proposed method is evaluated on a real-world dataset. Even though the available information from metadata is limited, the method still gives a promising result and has the advantage of its high computing speed.

Frage zum Eintrag

Redaktionelle Details anzeigen