Open-world probabilistic databases: Semantics, algorithms, complexity

Large-scale probabilistic knowledge bases are becoming increasingly important in academia and industry. They continuously extended with new data, powered by modern information extraction tools that associate probabilities base facts. The state of the art to store process such data is founded on databases. Many systems based databases, however, still have certain semantic deficiencies, which limit their potential applications. We revisit semantics argue closed-world assumption i.e., facts not appearing database probability zero, conflicts everyday use large-scale bases. To address this discrepancy, we propose open-world as a model. In model, unknown facts, also called open can be assigned any value from default interval. Our analysis entails our model aligns better many real-world tasks query answering, relational learning, completion, rule mining. make various technical contributions. show complexity dichotomy, between polynomial time , for evaluating unions conjunctive queries databases lifted This result supported an algorithm computes so-called safe efficiently. Based algorithm, prove linear under reasonable assumptions. remains true more restricted class queries. extend beyond queries, obtain host results both classical conclude in-depth investigation combined respective models.