How does Halcyon Search use keyword search terms?
Halcyon Search returns documents that contain an exact match of search terms/phrases. Halcyon Search does not use semantic search (like Google Search or most LLM tools), where a model infers what you probably mean and surfaces conceptually similar results. Halcyon does not do this.
For authoritative energy research, that semantic uncertainty can be problematic. So Halcyon splits the job into two steps: first constrain the universe with filters (publisher, topic, filing type), then search for exact terms within it.
Combining multiple keywords: ANY OF vs ALL OF
When you enter two or more keyword phrases, you choose how they combine:
- ALL OF — returns only documents that contain every phrase (AND logic). Use this to tighten results.
- ANY OF — returns documents that contain any of your phrases (OR logic). Use this to cast a wider net across synonyms or variants.
Each phrase is matched exactly, so “large load tariff” matches that phrase as written, not the three words scattered across a document.
Why we built it this way
Most search comes down to a trade-off between precision (showing exactly the right thing) and recall (not missing anything). Semantic search leans toward recall but is unpredictable — search "data centers in Wisconsin" and a model might also return renewable projects in Minnesota because it has learned the concepts are related. Sometimes that's helpful; often it's noise, and you have little control over how "fuzzy" the results get.
Keyword search gives you the opposite: total, predictable control over precision. Combined with Halcyon's filters, it lets you do something semantic search can't — constrain the universe first, then search within it. There's a big difference between narrowing to Wisconsin and then filtering to the term "data centers," versus searching the entire catalog for the loose phrase "data centers in Wisconsin." Halcyon's filter-first interface is designed around that distinction. (Read more in our CTO's post, Search: Everything Old is New Again.)
Best practices
- Constrain before you keyword. Narrow with Commission, Topic, and Filing Type filters first, then add keywords to pinpoint. This is the single biggest lever on result quality.
- Try variants and synonyms. Because matching is literal, “battery,” “energy storage,” and “BESS” return different sets. Use ANY OF to capture them together.
- Mind singular vs. plural. “data center” and “data centers” are not automatically the same — include both if it matters.
- Spell out acronyms and use proper nouns. Match the language as it actually appears in filings.
- Watch for source typos and OCR errors. If the original document misspells a term, an exact-match search won't find it — try an alternate spelling or a broader term.
- Use exact phrases for multi-word concepts (e.g., “certificate of public convenience and necessity”) to avoid scattered, irrelevant hits.
Got questions or feedback about this page? support@halcyon.io
Powered by