Quick, Correct, and Consistent Text Annotations: An Active Learning-Based Annotation Workflow and Tool for Sequence Labeling of Construction Schedules

Fouad Amer, Hui Yi Koh, Mani Golparvar-Fard

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Today, construction planning and scheduling are still conducted by expert planners whose knowledge is preserved in a portfolio of previous construction schedules and weekly work plans. While research has focused on assisting new planners by creating new artificial intelligence (AI) algorithms that automatically learn patterns of construction activities, a common underlying challenge to these algorithms is the lack of annotated data necessary for their training and testing. Towards this goal, this paper provides a new annotation method and an AI-based tool for part-of-activity tagging, i.e., for decoding the constructional functionalities embedded in a given construction activity name such as the activity's action, object, and location. When annotating, the tool's active learning mechanism allows the tool's AI-engine to suggest the most relevant annotations to expert planners performing the task while learning from their approvals and revisions on the fly. We evaluate the applicability of this algorithm on a dataset of ~1,350 construction schedule activities, and we demonstrate that our method enables fast generation of high-quality ground truth annotations. Furthermore, we extend the tool's capability to provide sequence annotations for any text data and we make it customizable and openly available on GitHub for the benefit of the research community.

Original languageEnglish (US)
Title of host publicationConstruction Research Congress 2020
Subtitle of host publicationComputer Applications - Selected Papers from the Construction Research Congress 2020
EditorsPingbo Tang, David Grau, Mounir El Asmar
PublisherAmerican Society of Civil Engineers
Pages924-933
Number of pages10
ISBN (Electronic)9780784482865
StatePublished - 2020
EventConstruction Research Congress 2020: Computer Applications - Tempe, United States
Duration: Mar 8 2020Mar 10 2020

Publication series

NameConstruction Research Congress 2020: Computer Applications - Selected Papers from the Construction Research Congress 2020

Conference

ConferenceConstruction Research Congress 2020: Computer Applications
CountryUnited States
CityTempe
Period3/8/203/10/20

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Building and Construction

Fingerprint Dive into the research topics of 'Quick, Correct, and Consistent Text Annotations: An Active Learning-Based Annotation Workflow and Tool for Sequence Labeling of Construction Schedules'. Together they form a unique fingerprint.

Cite this