Discover Academic Research

Search through thousands of research papers with intelligent citation tracking and advanced filters

How to Search

CiteQuest supports powerful query types. Hover any card to learn more.

📊

BM25 Search

machine learning

Default multi-term ranked search across all fields.

📊 BM25 Multi-Term Search
Example
machine learning
Behavior
  • Terms scored by BM25 across title, abstract, and body independently.
  • No strict boolean parsing unless query has uppercase AND or NOT.
  • Results sorted by final BM25 score per field.
🔗

Boolean AND

machine AND learning

Both terms must appear in the document.

🔗 Boolean AND
Example
machine AND learning
Behavior
  • AND is a hard filter — doc must contain both terms.
  • BM25 still ranks the filtered docs.
  • Boolean filter applied per field (title, abstract, body).
⚠️ Must be uppercase: AND not and
↔️

Boolean OR

robotics OR vision

Either term can appear in the document.

↔️ Boolean OR
Example
robotics OR vision AND planning
Behavior
  • OR combines groups: (robotics) OR (vision AND planning).
  • Without AND or NOT, behaves like normal BM25 search.
  • Strict boolean mode only activates when AND or NOT is present.
⚠️ Must be uppercase: OR not or
🚫

Boolean NOT

machine AND NOT learning

Exclude documents containing a specific term.

🚫 Boolean NOT
Example
machine AND NOT learning
Behavior
  • NOT excludes docs containing that term from the candidate set.
  • Pure NOT (e.g. NOT neural) enters filter-only mode.
  • Pure NOT results sorted by year desc, score = 0.
⚠️ Must be uppercase: NOT not not
💬

Phrase Search

"neural network"

Search for an exact sequence of words.

💬 Phrase Search
Example
"neural network"
Behavior
  • Quoted text with 2+ tokens treated as a phrase pattern.
  • Phrase matches add a rerank bonus to the score.
  • Not a hard filter — BM25 still scores all terms.
📏

Near Search

"neural #3 network"

Find terms within N words of each other.

📏 Near Search
Example
"neural #3 network"
Behavior
  • Parsed as near pair: (neural, 3, network).
  • Unordered distance check: abs(pos1 − pos2) ≤ 3.
  • Near matches add a rerank bonus, not a hard filter.
🔀

Mixed Query

machine AND "neural network"

Combine boolean operators with phrase search.

🔀 Mixed Query
Example
machine AND "neural network"
Behavior
  • Boolean logic parsed from unquoted text only.
  • Phrase/near patterns parsed from quoted text.
  • Final score = BM25 + phrase bonus + near bonus, after boolean filtering.
🗂️

Field Ranking

title · abstract · body

Results ranked independently across three document fields.

🗂️ Field-Based Ranking
Fields
title · abstract · body
Behavior
  • Ranker returns three independent ranked lists per query.
  • Default (auto): boolean filter applied per field independently.
  • Can force a single field (title / abstract / body) for all ranking.
ℹ️ No wildcard/fuzzy/prefix parser — tokens treated as normal terms.