Download presentation
Presentation is loading. Please wait.
1
Trio A System for Data, Uncertainty, and Lineage
Data Model Uncertainty Lineage ULDBs Uncertainty-Lineage Databases Saw (witness,car) (Mary,Honda) ∥ (Mary,Toyota) ∥ (Mary,Mazda) (Susan,Lexus) ID Drives (person,car) 21 (Bob,Honda) ∥ (Hank,Honda) 22 (Jim,Toyota) ∥ (Jim,Acura) ID Saw (witness,car) 11 (Mary,Honda) ∥ (Mary,Toyota) Understandable (we think ) and complete = can represent any set of possible instances ? Suspects = πperson(Saw ⋈ Drives) Alternatives Maybe annotations Confidences ID Suspects (person) 31 Bob ∥ Hank 32 Jim λ(31,1) = (11,1),(21,1); λ(31,2) = (11,1),(21,2) λ(32,1) = (11,2),(22,1) ? ? Drives (person,car) (Bob,Honda): 0.6 ∥ (Hank,Honda): 0.4 (Jim,Toyota): 0.2 ∥ (Jim,Acura): 0.5 See ULDB talk Friday 11:00 AM, room 310 Lineage... Enables simple & consistent representation of uncertain data Correlates uncertainty in query results with uncertainty in the input data Can make computation over uncertain data more efficient ? Formal semantics based on possible instances Query Language TriQL The Trio Query Language Confidences in Query Results Start with SQL … Modify semantics for ULDBs Default probabilistic interpretation Saw (witness,car) (Mary,Honda): 0.6 ∥ (Mary,Toyota): 0.4 Drives (person,car) (Bob,Honda): 0.3 ∥ (Hank,Honda): 0.7 Relational query Q on ULDB D semantics (logical) implementation Create Table Suspects(person) as Select Drives.person From Saw, Drives Where Saw.car = Drives.car Suspects (person) Bob: 0.18 ∥ Hank: 0.42 ? creates Suspects table with lineage to Saw and Drives But query can override implementation of Q D D + Result Select Drives.person, min(conf(Saw),conf(Drives)) as conf From Saw, Drives … … add constructs for querying uncertainty and lineage possible instances representation of instances Q on each instance Suspects (person) Bob: 0.3 ∥ Hank: 0.6 ? D1, D2, …, Dn Q(D1), Q(D2), …, Q(Dn) Built-in functions conf(), lineage(), lineage*() Confidences computed on-demand based on lineage feature System System Structure Demo Database DDL commands TriQL queries Schema browsing Table browsing Explore lineage On-demand confidence computation TrioExplorer (GUI client) Command-line client Trio API and translator (Python) Standard SQL Standard relational DBMS Encoded Data Tables Trio Metadata Lineage Tables Trio Stored Procedures DATA UNCERTAINTY LINEAGE
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.