Text Database Indexing TRIZ Cases
- Document Similarity Search Using Least-Frequent Terms and Proximity
Least-frequent terms and proximity-weighted queries improve similar document retrieval while reducing computational cost in large collections.
- Viewport-Aware Hierarchy Search With Pruned Tree Navigation
On-demand hierarchy provisioning and pruned tree expansion preserve context in limited viewport space while reducing search overhead.
- Value-Directed Parsing for Ambiguous Date Extraction
Scores alternative field values to disambiguate ambiguous dates in unstructured text and produce consistent structured data.
- Nested Classifier Database for Precise Goods and Services Selection
By linking text records to nested subheadings, this classifier database improves the speed and accuracy of goods and services selection.
- Hybrid Positional Posting Lists for Smaller, Faster Inverted Indexes
A hybrid posting list encodes document IDs and term positions in one value to shrink inverted indexes and speed retrieval.
- Hierarchical Duplicate Detection in Document Repositories
Relationship-specific similarity thresholds improve duplicate detection in document repositories by accounting for sibling and ancestor links.
- Hybrid Annotation Framework for Industry-Specific Labeling Accuracy
A hybrid annotation workflow combines automated labeling with manual review and feedback to improve industry-specific accuracy while reducing effort.
- Neural Database Training With Metadata-Augmented Document Chunks
Structured chunking and metadata mapping train a neural database to narrow document search faster with less compute and better targeting.
- LLM Query Response Pipeline With Chunk Ranking for External Knowledge
Synthetic query expansion and chunk ranking help an LLM use external document corpora to answer enterprise-specific questions more accurately.
- Faceted Time-Series Search for Real-Time Ad-Hoc Filtering
Pre-indexed time ranges let faceted search generate filtered time-series analytics in real time across large datasets.
- Incident Resolution Using Embedding Clusters and Resolution Notes
Historical resolution notes are chunked, embedded, and clustered to match new incidents faster and reduce downtime and cost.
- Query Interpretation Using Equation Packages for Accurate Responses
Tokenized query terms are tested across possible meanings, then filtered equation packages improve response accuracy without excessive processing.
- Word Knowledge Curation Using Clarifying Token Interpretation
Clarifying tokens resolve ambiguous word meanings into a single curated knowledge sequence, improving query accuracy and response relevance.
- LLM Retrieval Index Partitioning for Large Dataset Query Latency
Voronoi-cell index partitioning narrows embedding search scope, improving complex-query retrieval speed and resource use on large text datasets.
- Hybrid Document Classification Trees for Unstructured Content
Rule-based and ML nodes are combined in a decision tree to classify ambiguous documents with confidence thresholds and less manual review.