Survey of Smart Infrastructure, Forecasting, and Review-Evidence Systems

Smart infrastructure analytics and review-support systems appear to serve different users, yet both depend on evidence that is time-scoped, entity-bound, and auditable. Infrastructure platforms must combine sensor streams, asset histories, maintenance records, forecasts, and policy constraints before publishing alerts or plans. Review-support systems must combine submissions, citations, reviewer histories, phrase profiles, legal-style precedent features, confidence gates, and explanation records before assisting human judgment. This survey synthesizes hybrid semantic-relational retrieval, long-term forecasting, historical legal profiles, reviewer agreement forecasting, agreement-gated learning, structured extraction, GPU optimization, and prior archive work on judicial decision prediction, reviewer profiles, self-ensembling, smart infrastructure, and enterprise retrieval. We present an evidence lifecycle with six functions: acquisition, historical alignment, profile construction, retrieval, confidence-gated publication, and audit feedback. A comparative coding of representative system families shows that the strongest designs separate broad evidence discovery from narrow publication eligibility. An analytical study illustrates that chronology checks and confidence gates reduce unsupported actions while preserving usable coverage. The conclusion is conservative: evidence-rich systems should optimize and retrieve broadly, but publish only when source scope, time, confidence, and review records can be inspected.