CNS 2026 Annual Meeting
JW Marriott Parq Vancouver · March 7–10, 2026
High match
Medium match
Low match
0 selected · 0 high-relevance sessions
🧠 Research Profile
S
Sophie Xing Su
Cognitive Science PhD · WashU
🔍 TF-IDF · title 4× · subtitle 3× · talks 2× · abstract 1×
✓ Profile saved
✨ Scoring Method
🔍 Keyword frequency × position weight. Instant, no setup.
🧬 Runs all-MiniLM-L6-v2 fully in your browser — no API key needed. Downloads ~25 MB model once, then cached.
🤖 Uses text-embedding-3-small. Your key never leaves your browser. Est. cost: <$0.01.
Active: TF-IDF
📅 My Schedule
No sessions selected yet.
Click "+ Add" on sessions to build your schedule.
⚡ Top Picks for You
Computing...
✨

How LLM API Integration Would Improve This Scheduler

Optional Upgrade ▼

🔍 Current: Keyword TF-IDF

Runs entirely in-browser. Matches keywords from your profile against session text, weighting by position (title > subtitle > talk-title > abstract).

Cost$0.00
Latency<10ms
Semantic precision~65%
Synonym detection❌ No
Cross-paper matching❌ No
// Example scoring score += title_matches * 4.0 + subtitle_matches * 3.0 + talk_title_matches * 2.0 + abstract_matches * 1.0
✅ Works offline
✅ No API key needed
⚠️ Misses semantic similarity
⚠️ "event boundaries" ≠ "event segmentation" unless both listed

🧬 Better: Sentence Embeddings

Use sentence-transformers/all-MiniLM-L6-v2 or bge-small-en locally. Computes cosine similarity between your research abstract and each session abstract.

Cost$0.00 (local)
Latency~2s setup
Semantic precision~82%
Synonym detection✅ Yes
Cross-paper matching⚠️ Partial
from sentence_transformers import SentenceTransformer model = SentenceTransformer('all-MiniLM-L6-v2') user_emb = model.encode(user_abstract) for session in sessions: # Weight: title 3x, subtitle 2x, abstract 1x text = f"{session.title} {session.title} {session.title} " \ f"{session.subtitle} {session.subtitle} " \ f"{session.abstract}" sess_emb = model.encode(text) score = cosine_similarity(user_emb, sess_emb) # ~$0 cost, 300 sessions ≈ 2s
✅ Understands synonyms
✅ "event segmentation" ≈ "parsing experience"
✅ No API key needed
⚠️ Needs Python backend or WASM

🤖 Best: LLM API (Claude / GPT)

Upload your papers + research statement. LLM deeply understands your specific contributions and finds sessions with genuine intellectual overlap. Generates personalized explanations.

Cost (300 sessions)~$0.30–$0.60
Latency~30s total
Semantic precision~94%
Synonym detection✅✅ Deep
Cross-paper matching✅ Yes
import anthropic client = anthropic.Anthropic() # Batch embed with Anthropic response = client.messages.create( model="claude-opus-4-6", max_tokens=1024, messages=[{ "role": "user", "content": f"""Rate relevance (0-10) of this session to researcher Sophie Su who studies: event segmentation, prediction errors, eye tracking, CLIP, naturalistic fMRI. Session: {session.title} Subtitle: {session.subtitle} Talks: {session.talk_titles} Abstract: {session.abstract[:500]} Return JSON: {{"score": N, "reason": "...", "matching_concepts": [...]}}""" }] ) # Cost: ~$0.002/session × 300 = ~$0.60
✨ Example LLM improvement: Symposium 7 "Neural Time Machine" — Current score: 0.71
LLM score: 0.96 — "Talk 3 by Kanter directly addresses how event structure shapes neural dynamics, a core mechanism in Sophie's event segmentation theory. The temporal organization framework maps precisely onto her JEP:General (2025) paper."
💡 Estimated cost for CNS 2026:
~300 sessions × $0.002 = ~$0.60 total
User sees cost estimate & confirms before running.