MCPcopy Create free account
hub / github.com/917Dhj/DeepPaperNote / extract_negative_claims

Function extract_negative_claims

scripts/common.py:2243–2333  ·  view source on GitHub ↗
(text: str, *, limit: int = 6)

Source from the content-addressed store, hash-verified

2241
2242
2243def extract_negative_claims(text: str, *, limit: int = 6) -> list[str]:
2244 claims: list[str] = []
2245 seen = set()
2246 explicit_negative_tokens = [
2247 "worse",
2248 "degrade",
2249 "degraded",
2250 "drop",
2251 "dropped",
2252 "decrease",
2253 "decreased",
2254 "unstable",
2255 "instability",
2256 "fail",
2257 "failed",
2258 "fails",
2259 "collapse",
2260 "collapsed",
2261 "underperform",
2262 "underperformed",
2263 "sensitive",
2264 "sensitivity",
2265 "trade-off",
2266 "tradeoff",
2267 "hurt performance",
2268 "hurts performance",
2269 ]
2270 ablation_tokens = [
2271 "without",
2272 "w/o",
2273 "remove",
2274 "removed",
2275 "removing",
2276 "omit",
2277 "omits",
2278 "omitted",
2279 "omitting",
2280 "ablation",
2281 ]
2282 performance_tokens = [
2283 "accuracy",
2284 "f1",
2285 "auc",
2286 "auprc",
2287 "mae",
2288 "rmse",
2289 "score",
2290 "performance",
2291 "%",
2292 "result",
2293 "results",
2294 "training",
2295 "converge",
2296 "convergence",
2297 "stable",
2298 "stability",
2299 "baseline",
2300 ]

Calls 3

split_sentencesFunction · 0.85
normalize_whitespaceFunction · 0.85
normalize_titleFunction · 0.85