The construction industry is starting to implement intelligent knowledge management systems to support effective project communication. However, current practice in creating construction schedules remains unstandardised, which hinders the effective human-data knowledge exchange and impedes machines from reading and processing the information. This study, therefore, introduces a semi-automatic method to develop a novel ontology that summarises the key constituents of construction activities, and train an online classifier to classify new schedules and extend the schedule dictionary. The outcome is a domain-extensible ontology prototype validated by the selection of 27 completed schedules with a minimum of 500 activities each within a wide range of project types. The experimental results indicate that ontology-based activities improved: (1) schedule’s readability by reducing the cosine similarity of ‘different’ activities reduced from 0.995 to 0.990 (p < 0.01), (2) schedule’s understandability from 75.90% to 85.55%, and (3) schedule’s coherence from -12.04 to -11.44. This approach facilitates quantitative schedule analysis, enables the automatic generation of construction schedules, and improves the human-data exchange of information in the construction industry.