Semantic Scholar Paper Search
Pricing
from $2.00 / 1,000 paper fetcheds
Search and extract academic research papers from Semantic Scholar's database of over 200 million publications.
Pricing
from $2.00 / 1,000 paper fetcheds
Rating
0.0
(0)
Developer
ryan clinton
Actor stats
0
Bookmarked
7
Total users
3
Monthly active users
a day ago
Last modified
Categories
Share
Record Type
recordType
Optional
Discriminator. one-answer (single compact decision record — emitted by mode=one-answer or always at top of analytics modes), paper (search result), recommendation (similar paper), author-paper, author, batch-result, snippet, citation-edge, search-insights (hero synthesis card), top-picks (curated 5-paper list per criteria), alert (monitoring-driven event), topic-comparison, strategic-choices (head-to-head topic pairs with winner + reason), recommended-actions (read / monitor / ignore / investigate / cite — each with decisionRisk + riskFactors), research-map, blindspots, analysis-pack, deep-analysis-synthesis, contrarian-opportunities, field-dynamics, meta-insights, research-narrative (cross-run timeline, requires monitoring + ≥2 runs), decision-drift (when this run's recommendation differs from the prior monitored run), summary (always last record).
Paper ID
paperId
Optional
Semantic Scholar paper ID — primary key for downstream S2 lookups (recommendations, citation graph, references).
Title
title
Optional
Paper title
Year
year
Optional
Publication year
Publication Date
publicationDate
Optional
ISO publication date (when known)
Citations
citationCount
Optional
Raw citation count
References
referenceCount
Optional
Number of references in this paper
Influential Citations
influentialCitationCount
Optional
Citations Semantic Scholar's ML model classifies as meaningfully building on this paper, not just casually mentioning it. Unique to S2.
Canonical ID
canonicalId
Optional
Stable identifier for cross-mode dedup: doi:... if available, else arxiv:..., else pmid:..., else s2:.... Use this as the dedup key when merging results from multiple runs.
Influence Ratio
influenceRatio
Optional
influentialCitationCount / citationCount. 0.0–1.0. Higher = larger fraction of citations are meaningful (vs casual mentions). Null when citationCount is 0.
Momentum (cites/year)
momentumScore
Optional
Citations per year since publication. Raw — not normalised. Higher = the paper is being picked up faster. Null when year is missing or in the future.
Impact Score (0–100)
impactScore
Optional
Composite: 55% log10(citationCount) + 25% log10(influentialCitationCount) + 20% log10(citationsPerYear). Log-normalised so a 5,000-cite paper isn't 50× a 100-cite paper. Use as a sortable single-number quality signal.
Novelty Score (0–100)
noveltyScore
Optional
Surfaces 'early signal' papers — recent (≤3y), not yet famous (<1000 cites), but rising (decent velocity). 100 = brand new + climbing. 0 = old or already mainstream. Null when year is missing.
Decision Tags
decisionTags
Optional
Stable enum tags downstream automation branches on. foundational (older + high-impact + high-influence), breakout (recent + rising fast), rising (high momentum), declining (high cites but stalled), saturated (mainstream + no new uptake), open-access (free PDF). Empty array when no tag applies.
Why This Matters
whyThisMatters
Optional
Plain-English explanations (≤5 lines, deterministic — no LLM) of why this paper stands out: high impact, high influence ratio, rapid citation growth, foundational status, open-access PDF, etc. Pasteable directly into reports or LLM prompts.
Trajectory
trajectory
Optional
accelerating (high momentum or strong cross-run growth), steady (moderate momentum), plateauing (some momentum but no growth), declining (high cites but stalled), unknown (insufficient signal). Combines momentumScore with cross-run citationDelta when monitoring is on.
Is New (since last run)
isNew
Optional
Cross-run monitoring field. true = paper not seen in the previous snapshot for this monitoringStateKey. false = previously seen. null = no monitoring state OR first run for this key.
Citation Delta
citationDelta
Optional
Cross-run monitoring field. Citation count change since the previous run for this monitoringStateKey. Null when paper is new or no monitoring state.
Previous Citation Count
previousCitationCount
Optional
Cross-run monitoring field. Citation count from the previous snapshot. Null when paper is new or no monitoring state.
Days Since Last Seen
daysSinceLastSeen
Optional
Cross-run monitoring field. Days since this paper was last in a snapshot for the monitoringStateKey.
Open Access
isOpenAccess
Optional
True when a free PDF is available
PDF URL
openAccessPdfUrl
Optional
Direct open-access PDF link
DOI
doi
Optional
Digital Object Identifier
ArXiv ID
arxivId
Optional
ArXiv preprint identifier
PubMed ID
pmid
Optional
PubMed identifier
Venue
venue
Optional
Publication venue
Journal
journalName
Optional
Journal name (when published in a journal vs. conference)
Journal Volume
journalVolume
Optional
Journal volume identifier
Pages
journalPages
Optional
Page range within the journal
Abstract
abstract
Optional
Full paper abstract (when available from Semantic Scholar)
Fields of Study
fieldsOfStudy
Optional
Academic fields the paper is classified under (e.g. Computer Science, Medicine)
Publication Types
publicationTypes
Optional
Publication-type tags (JournalArticle, Conference, Review, etc.)
AI Summary (TLDR)
tldr
Optional
Semantic Scholar's AI-generated 1-sentence summary. Unique to S2.
S2 URL
semanticScholarUrl
Optional
Direct link to the Semantic Scholar page
Extracted At
extractedAt
Optional
Extraction timestamp
Source Paper ID
sourcePaperId
Optional
Provenance — the seed paper this record was derived from. Set on recommendation, author-paper, and citation-edge records.
Relationship
relationship
Optional
How this record relates to its source paper. cites / cited-by / recommended / authored-by.
Snippet Text
snippetText
Optional
Passage extracted from an open-access PDF. Snippet records only.
Snippet Kind
snippetKind
Optional
Snippet classifier (e.g. body, title, abstract). Snippet records only.
Section
section
Optional
PDF section the snippet was found in (e.g. Methods, Discussion). Snippet records only.
Score
score
Optional
Snippet relevance score from Semantic Scholar's snippet retriever.
Author Name
name
Optional
Author's display name. Author records only.
Affiliations
affiliations
Optional
Author's listed affiliations. Author records only.
Homepage
homepage
Optional
Author's homepage URL. Author records only.
Paper Count
paperCount
Optional
Number of papers attributed to this author. Author records only.
h-Index
hIndex
Optional
Author's h-index. Author records only.
Mode
mode
Optional
Mode this run was executed in. Summary records only.
Pushed Count
pushedCount
Optional
Total non-summary records pushed in this run. Summary records only.
Total Available
totalAvailable
Optional
Search-mode only. Total matching papers Semantic Scholar reports (may exceed pushed count if maxResults is lower).
Notes
notes
Optional
Plain-English run notes (filters applied, fallbacks taken, partial failures). Summary records only.
Price Per Event
pricePerEvent
Optional
PPE charge per record. Summary records only.
Estimated Cost (USD)
estimatedCostUsd
Optional
pushedCount × pricePerEvent. Summary records only.
Requested Mode
requestedMode
Optional
The mode value the user submitted. Differs from mode when auto was resolved into a concrete mode. Summary records only.
Auto-Resolved Reason
autoResolvedReason
Optional
Plain-English reason auto mode resolved to the chosen mode (e.g. "seedPaperIds has 3 entries — running the recommendation engine"). Summary records only.
Warnings
warnings
Optional
Pre-flight input warnings (large batch, broad query, capped seeds). Summary records only.
Confidence
confidence
Optional
Run-level confidence 0.0–1.0. Built from cohort size, presence of warnings, and whether any failures occurred. Summary + search-insights records.
Confidence Level
confidenceLevel
Optional
high (≥0.75), medium (≥0.5), low otherwise. Summary + search-insights records.
Confidence Explanation
confidenceExplanation
Optional
Plain-English explanation of the confidence level. Summary + search-insights records.
Failure Type
failureType
Optional
On error, classifies the failure: invalid-input, api-error, rate-limit, timeout, unknown. Null on success. Summary records only.
Headline
headline
Optional
One-line title for hero records (search-insights).
One Line
oneLine
Optional
Ultra-short shareable answer — Slack subject / dashboard tile. Search-insights records.
Key Takeaways
keyTakeaways
Optional
4–6 scannable bullets summarising the result set. Search-insights records.
What to Read
whatToRead
Optional
Role-tagged ranked picks (top-overall / most-influential / breakout / foundational / recent-breakthrough / open-access-pick). Each pick carries a concrete reason. Search-insights records.
Breakout Papers
breakoutPapers
Optional
Top 10% citations AND top 5% velocity (cites/year). Cross-signal synthesis the multi-source aggregator strips. Search-insights records.
Foundational Papers
foundationalPapers
Optional
Old (≥8y) AND still in top 10% velocity — durably influential. Search-insights records.
Recent Breakthroughs
recentBreakthroughs
Optional
≤2y old AND in top 10% velocity — early-stage rising work. Search-insights records.
Cohort Size
cohortSize
Optional
Papers used to compute the search insights. Search-insights records.
Median Citations
medianCitations
Optional
Median citation count across the cohort. Search-insights records.
Median Year
medianYear
Optional
Median publication year across the cohort. Search-insights records.
Suggested Queries
suggestedQueries
Optional
Top keywords from the result set that didn't appear in the user's query — actionable next-search hints. Search-insights records.
Topic Clusters
topicClusters
Optional
Lightweight keyword-based clustering of the result set. Each cluster: { label, keywords, paperCount, medianCitations, samplePaperIds }. Emitted in literature-review mode.
Timeline
timeline
Optional
research-narrative records: ordered shifts { fromRunAt, toRunAt, keyShift, type, evidence }. Also used in literature-review search-insights for per-year timeline.
Query Complexity
queryComplexity
Optional
low (1-2 tokens), medium (3-4), high (≥5 OR explicit AND/OR/NOT operators). Summary records only.
Result Diversity (0-1)
resultDiversityScore
Optional
Shannon diversity over fieldsOfStudy distribution, normalised to 0-1. 1 = papers spread evenly across many fields, 0 = all papers in a single field. Summary records only (search-style modes).
Dominant Field
dominantField
Optional
Most-represented field of study in the result set. Summary records only (search-style modes).
Year Distribution
yearDistribution
Optional
Histogram { year → paperCount }. Summary records only (search-style modes).
Collection Used
collectionUsed
Optional
Curated collection preset that was applied (if any). Summary records only.
Monitoring (cross-run summary)
monitoring
Optional
Cross-run monitoring aggregates: { enabled, stateKey, isFirstRun, priorRunAt, daysSincePriorRun, snapshotSize, newPapersCount, growingPapersCount, citationGrowthSum, trendSummary{newPapers, risingPapers, decliningPapers, spikingPapers, notes[]}, alertsTriggeredCount }. Summary records only when monitoringStateKey is set.
Criteria
criteria
Optional
Top-picks records only. One of: top-impact, top-breakout, top-recent, top-foundational.
Label
label
Optional
Top-picks records: human-readable list label.
Papers
papers
Optional
top-picks records: array of pick items. contrarian-opportunities records: array of contrarian picks.
Alert Type
alertType
Optional
Alert records only. Stable enum: new-breakout-paper, citation-spike, newly-foundational, rising-cluster, declining-cluster.
Severity
severity
Optional
Alert records only. info / notable / high.
Evidence
evidence
Optional
Alert records: plain-English signals that support the alert.
Cluster
cluster
Optional
Cluster-level alerts only — the topic cluster label that triggered the alert.
Comparison Rows
comparison
Optional
topic-comparison records: per-topic { topic, paperCount, avgImpactScore, medianCitations, avgMomentumScore, avgNoveltyScore, breakoutCount, foundationalCount, topPaper, growthRateVsLastRun }.
Leader
leader
Optional
topic-comparison records: which topic ranks first by composite score.
Leader Reason
leaderReason
Optional
topic-comparison records: plain-English reason the leader topic ranks first.
Actions
actions
Optional
recommended-actions records: array of {actionType (read/monitor/ignore/investigate/cite), target (paper/topic/author), paperId, title, topic, reason, priority}.
Action Counts
actionCounts
Optional
recommended-actions records: per-actionType counts.
Core Areas
coreAreas
Optional
research-map records: high-importance, established topic clusters.
Emerging Areas
emergingAreas
Optional
research-map records: clusters growing >25% OR small-but-high-impact.
Declining Areas
decliningAreas
Optional
research-map records: clusters shrinking ≥25% since last monitored run.
Adjacent Opportunities
adjacentOpportunities
Optional
research-map records: cross-disciplinary fields appearing in 5-30% of results — outside your dominant field.
Field Distribution
fieldDistribution
Optional
research-map records: { field → paperCount } histogram across the cohort.
Blindspots
blindspots
Optional
blindspots records: array of {topic, type, reason, suggestedQuery, sampleEvidence}. Types: undersearched-keyword, temporal-gap, adjacent-field, open-access-gap.
Summary
summary
Optional
analysis-pack records: one-paragraph fact-heavy summary of the cohort.
Top Findings
topFindings
Optional
analysis-pack records: 3-5 plain-English finding sentences.
Key Papers
keyPapers
Optional
analysis-pack records: top 5 papers by impactScore with their whyThisMatters reasons.
Trend Narrative
trendNarrative
Optional
analysis-pack records: synthesised narrative about cohort trends.
Risks
risks
Optional
analysis-pack records: list of caveats, blindspots, low-confidence flags.
Recommended Actions (string list)
recommendedActions
Optional
analysis-pack records: paste-ready action lines derived from the recommended-actions record.
Consensus Findings
consensusFindings
Optional
deep-analysis-synthesis records: papers that surfaced across multiple passes (broad search + foundational + emerging).
Conflicting Signals
conflictingSignals
Optional
deep-analysis-synthesis records: tensions between passes (e.g. foundational paper now declining, breakout missing from canon).
High-Confidence Areas
highConfidenceAreas
Optional
deep-analysis-synthesis records: areas where multiple passes converge — well-established AND active.
Modes Run
modesRun
Optional
deep-analysis-synthesis records: list of sub-modes that were executed.
Seed Paper Used
seedPaperUsed
Optional
deep-analysis-synthesis records: the top-impact paper used to seed the citation-graph pass.
User Profile (cross-run summary)
userProfile
Optional
Summary records: cross-run personalisation aggregates: { enabled, profileKey, isFirstRun, runCount, knownQueries, knownClusters, knownPapers, biasNotes[] }. Only when userProfileKey is set.
Deep Analysis Pass Info
deepAnalysis
Optional
Summary records: { modesRun, seedPaperUsed } when mode = deep-analysis.
Confidence Drivers
confidenceDrivers
Optional
Summary records: structured drivers behind the confidence score. Array of { factor, direction (positive/negative/neutral), weight, note }. Lets users audit WHY confidence is high/low instead of trusting an opaque number.
Query Performance
queryPerformance
Optional
Summary records: { coverageScore, diversityScore, missedAreas[], nextBestQueries[] }. Quantifies how well the query covered the result space and proposes follow-ups.
Overall Decision Risk
overallDecisionRisk
Optional
Summary records: low / medium / high — risk of acting on this run's recommendations. Derived from confidence + signal density + cross-pass conflicts.
Overall Decision Risk Factors
overallDecisionRiskFactors
Optional
Summary records: plain-English factors behind the risk classification.
Description
description
Optional
contrarian-opportunities + top-picks records: explanation of the picks.
Overall State
overallState
Optional
field-dynamics records: weighted majority across clusters — emerging / expanding / saturated / declining.
Clusters
clusters
Optional
field-dynamics records: per-cluster { name, state, saturationScore, paperCount, growthRatePct, medianImpactScore, medianNoveltyScore, explanation }.
Patterns
patterns
Optional
meta-insights records: statistical patterns across cohort signals, each with { pattern, evidence, confidence }.
Run Count
runCount
Optional
research-narrative records: number of historical runs in the timeline.
Span (days)
spanDays
Optional
research-narrative records: days between earliest and latest snapshot.
Current Dominant Topic
currentDominantTopic
Optional
research-narrative records: dominant cluster label in the most recent snapshot.
Decision Risk
decisionRisk
Optional
Per-action field on recommended-actions: low / medium / high — risk of acting on this specific recommendation.
Risk Factors
riskFactors
Optional
Per-action field on recommended-actions: plain-English factors that drove the decisionRisk classification.
Answer
answer
Optional
one-answer records: single-sentence-or-two consolidated answer the user can read in 5 seconds.
Decision Summary
decisionSummary
Optional
one-answer + summary records: compressed 4-line summary { primaryInsight, secondaryInsight, biggestRisk, bestOpportunity, quotable }. Pasteable into Slack / email / agent prompts.
Top Papers
topPapers
Optional
one-answer records: 3 top picks by impactScore — each with paperId, title, year, impactScore, semanticScholarUrl.
Previous Decision
previousDecision
Optional
decision-drift records: prior monitored run's primary insight + top paper + top cluster + capturedAt timestamp.
Current Decision
currentDecision
Optional
decision-drift records: this run's primary insight + top paper + top cluster.
Change Detected
changeDetected
Optional
decision-drift records: true when this run's recommendation differs from the prior monitored run.
Change Type
changeType
Optional
decision-drift records: top-paper-changed / top-cluster-changed / insight-direction-changed / no-change.
Choices
choices
Optional
strategic-choices records: array of { optionA, optionB, winner, reason, aPaperCount, bPaperCount, aImportance, bImportance }.
Complexity Level Applied
complexityLevelApplied
Optional
Summary records: the effective complexityLevel for this run (low / medium / high).
Continuous Mode Applied
continuousModeApplied
Optional
Summary records: whether continuous (deltas-only) mode was active.
Suppressed Record Types
suppressedRecordTypes
Optional
Summary records: list of recordTypes that the complexityLevel + continuousMode filter suppressed during this run.
