File size: 14,839 Bytes
5f58cd5 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 | """
arXiv ingestion module.
Polls the arXiv API for new preprints in ML/AI categories, deduplicates by
arXiv ID, and applies a triage gate: only papers whose abstracts contain
novelty-claim language proceed to extraction.
Categories polled:
cs.LG — Machine Learning
cs.AI — Artificial Intelligence
stat.ML — Machine Learning (Statistics)
cs.CL — Computation and Language (NLP)
Rate limit: arXiv asks for polite delays (one call per 3 seconds).
We use 5s between calls and limit to 50 results per call by default.
"""
from __future__ import annotations
import re
import time
import urllib.parse
import urllib.request
import xml.etree.ElementTree as ET
from dataclasses import dataclass, field
from datetime import datetime, timezone
from typing import Any
from .db import Database, get_db
# ---------------------------------------------------------------------------
# arXiv API constants
# ---------------------------------------------------------------------------
ARXIV_API_BASE = "https://export.arxiv.org/api/query"
DEFAULT_CATEGORIES = ["cs.LG", "cs.AI", "stat.ML", "cs.CL"]
# arXiv API namespaces
_NS = {
"atom": "http://www.w3.org/2005/Atom",
"arxiv": "http://arxiv.org/schemas/atom",
}
# ---------------------------------------------------------------------------
# Triage gate
# ---------------------------------------------------------------------------
# Phrases that suggest the paper claims novelty (pass triage)
_NOVELTY_PATTERNS: list[re.Pattern] = [
re.compile(p, re.IGNORECASE)
for p in [
r"\bnovel\b",
r"\bwe\s+(propose|introduce|present)\b",
r"\bnew\s+(method|architecture|framework|approach|technique|algorithm|model|paradigm)\b",
r"\bstate.of.the.art\b",
r"\boutperforms?\b",
r"\b(unlike|differs?\s+from|in\s+contrast\s+to)\s+(prior|previous|existing|traditional)\b",
r"\badvances?\s+(the\s+)?(state|field)\b",
r"\bfirst\s+(method|approach|architecture|time)\b",
r"\b(breakthrough|groundbreaking|pioneering)\b",
r"\bcontribution\b",
r"\bwe\s+(achieve|obtain|demonstrate)\b",
]
]
# Phrases that suggest the paper claims NO novelty (skip or flag)
_SKIP_PATTERNS: list[re.Pattern] = [
re.compile(p, re.IGNORECASE)
for p in [
r"\b(survey|review|tutorial)\s+(of|on)\b",
r"\bcomprehensive\s+(survey|review)\b",
r"\b(literature\s+review|related\s+work\b)",
r"\b(benchmark|benchmarking)\b",
r"\b(reproduce|replicate|reproduction)\b",
r"\b(dataset|corpus|collection)\s+(release|introduction|description)\b",
r"\b(position\s+paper|opinion|commentary)\b",
r"\b(workshop|competition|challenge)\s+(report|summary|overview)\b",
r"\b(extended\s+abstract|demo|poster)\b",
]
]
def triage(abstract: str) -> tuple[bool, str]:
"""Determine whether an abstract passes the triage gate.
Returns (passed, reason).
"""
# Check skip patterns first (hard no)
for pat in _SKIP_PATTERNS:
if pat.search(abstract):
return False, f"skip_pattern_match: {pat.pattern[:60]}"
# Check novelty patterns (soft yes)
matches: list[str] = []
for pat in _NOVELTY_PATTERNS:
m = pat.search(abstract)
if m:
matches.append(m.group(0))
if matches:
return True, f"novelty_signals: {', '.join(matches[:3])}"
return False, "no_novelty_signals_detected"
# ---------------------------------------------------------------------------
# arXiv API client
# ---------------------------------------------------------------------------
def _fetch_arxiv(
categories: list[str] | None = None,
max_results: int = 50,
start: int = 0,
sort_by: str = "submittedDate",
sort_order: str = "descending",
) -> str:
"""Fetch raw XML from arXiv API. Returns the XML as a string.
Tries urllib first, falls back to requests if urllib fails (some
container environments have DNS/config issues with urllib).
"""
if categories is None:
categories = DEFAULT_CATEGORIES
cat_query = "+OR+".join(f"cat:{c}" for c in categories)
params = {
"search_query": cat_query,
"start": str(start),
"max_results": str(max_results),
"sortBy": sort_by,
"sortOrder": sort_order,
}
url = f"{ARXIV_API_BASE}?{urllib.parse.urlencode(params)}"
headers = {"User-Agent": "DifferanceEngine/0.1 (mailto:chris@graziul.io)"}
# Strategy 1: urllib
try:
req = urllib.request.Request(url, headers=headers)
with urllib.request.urlopen(req, timeout=30) as resp:
return resp.read().decode("utf-8")
except Exception as e:
print(f" [ingest] urllib fetch failed ({type(e).__name__}: {str(e)[:100]}), trying requests...")
# Strategy 2: requests (more robust in containerized environments)
import requests as _requests
resp = _requests.get(url, headers=headers, timeout=30)
resp.raise_for_status()
return resp.text
def parse_arxiv_xml(xml_str: str) -> list[dict]:
"""Parse arXiv API Atom XML into a list of paper dicts."""
root = ET.fromstring(xml_str)
papers: list[dict] = []
for entry in root.findall("atom:entry", _NS):
arxiv_id_full = entry.find("atom:id", _NS).text or ""
# Strip the "http://arxiv.org/abs/" prefix to get the canonical ID
arxiv_id = arxiv_id_full.split("/abs/")[-1] if "/abs/" in arxiv_id_full else arxiv_id_full
title = " ".join((entry.find("atom:title", _NS).text or "").split())
abstract = " ".join((entry.find("atom:summary", _NS).text or "").split())
# Authors
authors: list[str] = []
for author_elem in entry.findall("atom:author", _NS):
name_elem = author_elem.find("atom:name", _NS)
if name_elem is not None and name_elem.text:
authors.append(name_elem.text.strip())
# Categories
categories: list[str] = []
for cat_elem in entry.findall("atom:category", _NS):
term = cat_elem.get("term", "")
if term:
categories.append(term)
# Dates
published = entry.find("atom:published", _NS)
published_str = published.text if published is not None else ""
updated = entry.find("atom:updated", _NS)
updated_str = updated.text if updated is not None else ""
# PDF link
pdf_url = ""
for link in entry.findall("atom:link", _NS):
if link.get("title") == "pdf":
pdf_url = link.get("href", "")
break
papers.append({
"arxiv_id": arxiv_id,
"title": title,
"abstract": abstract,
"authors": authors,
"categories": categories,
"published": published_str,
"updated": updated_str,
"pdf_url": pdf_url,
})
return papers
# ---------------------------------------------------------------------------
# Ingestion runner
# ---------------------------------------------------------------------------
@dataclass
class IngestResult:
ingested: int = 0
triaged_in: int = 0
triaged_out: int = 0
skipped_existing: int = 0
def ingest_daily(
db: Database | None = None,
categories: list[str] | None = None,
max_results: int = 50,
max_pages: int = 2,
) -> IngestResult:
"""Run daily ingestion: fetch new papers, deduplicate, triage, store.
Pages through results up to max_pages * max_results papers.
"""
if db is None:
db = get_db()
db.connect()
if categories is None:
categories = DEFAULT_CATEGORIES
result = IngestResult()
for page in range(max_pages):
start = page * max_results
try:
xml_str = _fetch_arxiv(
categories=categories,
max_results=max_results,
start=start,
)
except Exception as e:
print(f" [ingest] arXiv API error (page {page}, start={start}): {e}")
if page == 0:
raise # Fail hard on first page error; tolerate subsequent pages
break
papers = parse_arxiv_xml(xml_str)
if not papers:
break # No more results
for paper in papers:
# Deduplicate
if db.paper_exists(paper["arxiv_id"]):
result.skipped_existing += 1
continue
# Triage
passed, reason = triage(paper["abstract"])
# Store
db.insert_paper(paper)
db.update_triage(paper["arxiv_id"], passed, reason)
result.ingested += 1
if passed:
result.triaged_in += 1
else:
result.triaged_out += 1
# Respect arXiv rate limit
if page < max_pages - 1:
time.sleep(5)
return result
def _fetch_paper_via_hf_hub(arxiv_id: str) -> dict | None:
"""Fallback: try to fetch paper metadata via huggingface_hub papers API.
HF Hub mirrors arXiv metadata and may be reachable when arXiv is not.
Returns a paper dict matching the arXiv parse format, or None.
Uses the HF_TOKEN from the environment (automatically available in Spaces).
The list_papers API was added in huggingface_hub 0.26+; if unavailable,
falls back to searching daily papers.
"""
try:
from huggingface_hub import HfApi
api = HfApi()
# Try the dedicated list_papers API (huggingface_hub >= 0.26)
if hasattr(api, "list_papers"):
papers = api.list_papers(query=arxiv_id, limit=1)
paper_list = list(papers)
if paper_list:
p = paper_list[0]
return {
"arxiv_id": p.id or arxiv_id,
"title": p.title or "",
"abstract": p.summary or "",
"authors": p.authors or [],
"categories": p.tags or [],
"published": p.published_at.isoformat() if p.published_at else "",
"updated": p.updated_at.isoformat() if getattr(p, "updated_at", None) else "",
"pdf_url": p.url_pdf or "",
}
# Fallback: search daily papers
if hasattr(api, "search_papers"):
papers = api.search_papers(query=arxiv_id, limit=1)
paper_list = list(papers)
if paper_list:
p = paper_list[0]
return {
"arxiv_id": p.id or arxiv_id,
"title": p.title or "",
"abstract": p.summary or "",
"authors": p.authors or [],
"categories": p.tags or [],
"published": p.published_at.isoformat() if p.published_at else "",
"updated": p.updated_at.isoformat() if getattr(p, "updated_at", None) else "",
"pdf_url": p.url_pdf or "",
}
print(f" [ingest] HF Hub paper API not available in this huggingface_hub version")
return None
except Exception as e:
print(f" [ingest] HF Hub paper fallback also failed: {type(e).__name__}: {str(e)[:120]}")
return None
def _try_fetch_citations(arxiv_id: str, db: Database | None = None):
"""Fetch citation count from Semantic Scholar and update the DB.
Runs synchronously but catches all errors — this is best-effort enrichment,
not mission-critical. If it fails, the paper is still ingested.
"""
import json as _json
try:
url = f"https://api.semanticscholar.org/graph/v1/paper/ArXiv:{arxiv_id}?fields=citationCount"
req = urllib.request.Request(url, headers={"User-Agent": "DifferanceEngine/1.0"})
with urllib.request.urlopen(req, timeout=10) as resp:
data = _json.loads(resp.read())
count = data.get("citationCount", 0)
if count and db:
db.update_citation(arxiv_id, count)
return count
except Exception:
return 0
def ingest_single(
arxiv_id: str,
db: Database | None = None,
) -> dict | None:
"""Ingest a single paper by arXiv ID.
Returns the paper dict if found and ingested, None if not found.
Tries arXiv API first, falls back to huggingface_hub papers API.
"""
if db is None:
db = get_db()
db.connect()
# Check if already ingested (version-aware) — if so, return it
existing = db.find_paper(arxiv_id)
if existing:
return existing
paper = None
errors: list[str] = []
# Strategy 1: arXiv API (primary)
try:
params = {
"id_list": arxiv_id,
"max_results": "1",
}
url = f"{ARXIV_API_BASE}?{urllib.parse.urlencode(params)}"
req = urllib.request.Request(url)
req.add_header("User-Agent", "DifferanceEngine/0.1 (mailto:chris@graziul.io)")
with urllib.request.urlopen(req, timeout=30) as resp:
xml_str = resp.read().decode("utf-8")
papers = parse_arxiv_xml(xml_str)
if papers:
paper = papers[0]
except Exception as e:
err_msg = f"arXiv API: {type(e).__name__}: {str(e)[:120]}"
errors.append(err_msg)
print(f" [ingest] {err_msg}")
# Strategy 2: Also try with requests (sometimes urllib fails on weird network configs)
if paper is None:
try:
import requests as _requests
params = {
"id_list": arxiv_id,
"max_results": "1",
}
url = f"{ARXIV_API_BASE}?{urllib.parse.urlencode(params)}"
resp = _requests.get(
url,
headers={"User-Agent": "DifferanceEngine/0.1 (mailto:chris@graziul.io)"},
timeout=30,
)
resp.raise_for_status()
papers = parse_arxiv_xml(resp.text)
if papers:
paper = papers[0]
except Exception as e:
err_msg = f"arXiv via requests: {type(e).__name__}: {str(e)[:120]}"
errors.append(err_msg)
print(f" [ingest] {err_msg}")
# Strategy 3: huggingface_hub papers API (last resort)
if paper is None:
print(f" [ingest] arXiv fetch failed, trying HF Hub papers API...")
paper = _fetch_paper_via_hf_hub(arxiv_id)
if paper is None:
print(f" [ingest] All strategies failed for {arxiv_id}: {'; '.join(errors)}")
return None
passed, reason = triage(paper["abstract"])
db.insert_paper(paper)
db.update_triage(paper["arxiv_id"], passed, reason)
# Fetch citation count from Semantic Scholar in background
_try_fetch_citations(paper["arxiv_id"], db)
return paper
|