A curated, searchable collection of semantic query processing research. Follow an idea to its source, then understand the limits of the evidence.
Natural-language specifications for filtering, joining, sorting and aggregation, with optimized implementations evaluated against a reference algorithm.
A SQL-like language for reasoning over images, conversations and reviews, combining foundation models with sampling and execution optimization.
Turns declarative AI analytics into executable plans that trade off output quality, runtime and model cost.
Rewrites complex document workflows by decomposing tasks and data, using agents to propose and evaluate better execution plans.
Composes structured queries with free-text answering and summarization, supporting conversational access to hybrid knowledge sources.
Evaluates semantic engines across text, image and audio scenarios using filters, joins, mappings, ranking and classification.
Organizes Select, Match, Impute, Cluster and Order operators into single-operator and multi-operator workloads over relational databases.
Integrates semantic expressions into SQL and combines logical rewrites with adaptive reordering, fusion and prompt batching.
A DataFusion-based engine using semantic algebra, token histograms and context-aware batching to optimize AI analytics.
Uses decision models for semantic operations, escalates uncertain cases to generative models and prunes semantic-join candidates.
Reuses prior query labels as ranking signals and applies fresh stratified oracle sampling to certify filter precision and recall.
Studies how repeated model decisions propagate into changing query answers and proposes reporting practices for reproducible evaluation.
Compiles suitable semantic operations into local Python functions, routing between compiled and native execution using cost estimates.
Makes model serving aware of semantic operator dependencies, pipelining tuples to reuse KV-cache state under memory constraints.
Adds model-backed scalar and aggregate functions, with PROMPT and MODEL schema objects plus batching and caching optimizations.
Supports in-database ML and LLM inference through a relational predict operator and optimizations across semantic SQL operations.
Balances model-call cost against relational processing cost when choosing semantic operator placement in mixed query plans.
Reorders table rows and fields to increase shared prompt prefixes and improve KV-cache reuse during batch model inference.
Studies physical implementations of semantic ordering and a budget-aware optimizer that chooses among ranking access paths.
Learns per-row evaluation orders for semantic predicates, using either a sequential decision policy or predicted selectivities.
Combines clustering and learned proxies adaptively, using token-aware representations and oracle confidence to reduce expensive filtering calls.
Uses an LLM advisor to route joins between clustering-based candidate pruning and classification into a shared label space.
Evaluates unstructured-document analysis against manually annotated relational views. The arXiv v1 workload contains five datasets and 240 queries covering extraction, filtering, aggregation, joins, and combinations of operators.
Extends TPC-DS-derived relations with return reports, issue bulletins, and release notes to study semantic-join scaling. The JEVDB paper evaluates five queries at XS, S, and M tiers.
Tests whether models generate Snowflake SQL that composes relational operations with AI functions for classification, filtering, sentiment analysis, extraction, similarity, and aggregation.
Benchmarks generation of queries combining relational operations and vector search, spanning three database backends and four source collections. Examines result omissions when SQL filters interact with vector retrieval.
Evaluates agents on 380 tasks requiring both database queries and web evidence. Tasks exercise database-to-search, search-to-database, and parallel reasoning, with an emphasis on preserving constraints across tools.
Studies failures of exact-result scoring for SQL with stochastic AI operators. Proposes evaluating relational logic and semantic operations separately, with experiments on BigQuery and ThalamusDB.
Try a broader term or reset the filters.
sqpe