File size: 5,155 Bytes
71d49e7 | 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 | import re
FEATURE_NAMES = [
"n_words",
"n_sents",
"n_punct",
"avg_sent_len",
"n_code_blocks",
"uppercase_ratio",
"n_questions",
"has_code",
"has_arch",
"has_math",
"has_debug",
"has_research",
"n_complex_verbs",
"n_scale_adjs",
"n_hard_nouns",
"has_simple",
"has_realtime",
"has_long_output",
"avg_word_len",
]
_WORD_RE = re.compile(r"[A-Za-z0-9_]+")
_SENT_RE = re.compile(r"[^.!?\n]+[.!?]?")
_PUNCT_RE = re.compile(r"[^\w\s]")
CODE_PAT = re.compile(
r"(```|~~~|`[^`]+`|\b(def |class |import |from .* import|function |const |let |"
r"var |SELECT |INSERT |UPDATE |DELETE |public static|#include|fn main|=>"
r"|println!|System\.out))"
)
ARCH_WORDS = {
"architecture",
"architect",
"microservice",
"microservices",
"distributed",
"pipeline",
"infrastructure",
"kubernetes",
"cluster",
"scalab",
"deploy",
"migration",
"monolith",
"shard",
"replica",
"consensus",
"raft",
"paxos",
}
MATH_PAT = re.compile(
r"[=<>+\-*/^√∫∑∏∂∇≤≥≠≈∞]|\b(theorem|proof|prove|lemma|equation|"
r"induction|derivative|integral|matrix|eigen|probability|qed)\b"
)
DEBUG_WORDS = {
"debug",
"debugging",
"error",
"bug",
"traceback",
"stacktrace",
"exception",
"fix",
"broken",
"crash",
"fail",
"failing",
"patch",
"hotfix",
}
RESEARCH_WORDS = {
"research",
"paper",
"study",
"survey",
"literature",
"benchmark",
"dataset",
"baseline",
"ablation",
"sota",
"state-of-the-art",
}
COMPLEX_VERBS = {
"analyze",
"analyse",
"design",
"implement",
"optimize",
"optimise",
"architect",
"prove",
"derive",
"synthesize",
"evaluate",
"compare",
"migrate",
"deploy",
"scale",
"secure",
"refactor",
"integrate",
"orchestrate",
"formalize",
"generalize",
"debug",
}
SCALE_ADJS = {
"distributed",
"large-scale",
"large scale",
"massive",
"enterprise",
"global",
"high-throughput",
"high throughput",
"fault-tolerant",
"fault tolerant",
"real-time",
"realtime",
"petabyte",
"million",
"billion",
"end-to-end",
"end to end",
"production-grade",
"production grade",
}
HARD_NOUNS = {
"kubernetes",
"microservices",
"redis",
"postgres",
"theorem",
"proof",
"induction",
"quantum",
"compiler",
"kernel",
"cryptography",
"blockchain",
"consensus",
"neural",
"transformer",
"gradient",
"eigen",
"mlops",
"terraform",
"kafka",
"spark",
"hadoop",
"docker",
"inference",
}
SIMPLE_PAT = re.compile(
r"\b(simple|simply|easy|basic|basics|hello|hi\b|what is|who is|where is"
r"|how do i say)\b"
)
REALTIME_PAT = re.compile(r"\b(real-?time|live|streaming|latency|deadline|instant)\b")
LONGOUT_PAT = re.compile(
r"\b(essay|report|thesis|detailed|comprehensive|in-?depth|"
r"step-?by-?step|thorough|elaborate|paragraphs|pages?)\b"
)
def _count_phrases(text_low: str, phrases: tuple[str, ...]) -> int:
return sum(text_low.count(phrase) for phrase in phrases)
def featurize(text: str) -> list[float]:
source = text if isinstance(text, str) else ""
low = source.lower()
words = _WORD_RE.findall(source)
n_words = len(words)
sentences = [part for part in _SENT_RE.findall(source) if part.strip()]
n_sents = max(len(sentences), 1)
n_punct = len(_PUNCT_RE.findall(source))
avg_sent_len = n_words / n_sents
n_code_blocks = len(re.findall(r"```|~~~", source)) // 2 + len(
re.findall(r"(?m)^(?: |\t)\S", source)
)
alpha = [char for char in source if char.isalpha()]
uppercase_ratio = (
(sum(char.isupper() for char in alpha) / len(alpha)) if alpha else 0.0
)
n_questions = source.count("?")
has_code = 1.0 if CODE_PAT.search(source) else 0.0
has_arch = 1.0 if any(word in low for word in ARCH_WORDS) else 0.0
has_math = 1.0 if MATH_PAT.search(low) else 0.0
has_debug = 1.0 if any(word in low for word in DEBUG_WORDS) else 0.0
has_research = 1.0 if any(word in low for word in RESEARCH_WORDS) else 0.0
n_complex_verbs = float(_count_phrases(low, COMPLEX_VERBS))
n_scale_adjs = float(_count_phrases(low, SCALE_ADJS))
n_hard_nouns = float(_count_phrases(low, HARD_NOUNS))
has_simple = 1.0 if SIMPLE_PAT.search(low) else 0.0
has_realtime = 1.0 if REALTIME_PAT.search(low) else 0.0
has_long_output = 1.0 if LONGOUT_PAT.search(low) else 0.0
avg_word_len = (sum(len(word) for word in words) / n_words) if n_words else 0.0
return [
n_words,
n_sents,
n_punct,
avg_sent_len,
n_code_blocks,
uppercase_ratio,
n_questions,
has_code,
has_arch,
has_math,
has_debug,
has_research,
n_complex_verbs,
n_scale_adjs,
n_hard_nouns,
has_simple,
has_realtime,
has_long_output,
avg_word_len,
]
|