Data quality management and normalization of reference directories of work objects, materials, and defects in construction control information systems to improve reporting comparability

Abstract

The article addresses the problem of low reporting comparability in construction control information systems, arising from semantic heterogeneity and structural fragmentation of reference directories of objects, types of work, materials, and defects when operating heterogeneous digital platforms. It is shown that the accumulation of records without strict validation at the input stage generates high data entropy: duplicates, typos, variable abbreviations, transliteration, and professional slang transform arrays of as-built documentation into a resource that is difficult to use for automated analysis and create a semantic gap during integration with information modeling. Based on a systems analysis of large-scale exports from cloud platforms for 2019-2023 using statistical approaches and set theory methods, a normalization toolkit is proposed and tested, combining fuzzy string matching (edit distance, Jaccard measure) with tokenization and lemmatization adapted to construction morphology; the results of automatic clustering are confirmed by expert verification. Quantitative indicators demonstrate critical redundancy in material nomenclature: for rebar products, the average number of spelling variations per actual item exceeds a hundred, making aggregation by direct queries methodologically incorrect without prior deduplication. For defects, a high proportion of fuzzy semantics and significant values of the semantic entropy index are identified, especially in finishing works, leading to unacceptably high probabilities of errors in automatic classification. Normalization of reference directories and transition to master data provide a substantial increase in the performance of analytical scenarios by replacing string comparisons with operations on identifiers and indexes, drastically reducing response time, computational resource load, and transmitted traffic volume. The managerial and legal significance of data unification is substantiated: iterations of act approvals are reduced, the share of rejections on formal grounds decreases, legal certainty of electronic documents increases, and prerequisites are formed for implementing predictive analytics and industry ontologies necessary for evidence-based construction quality management.