File size: 10,107 Bytes
9e39004 6a467c4 9e39004 |
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 |
import os
import json
import stanza
# -----------------------------------------------------------------------------
# Stanza Pipeline (English, tokenize + POS + lemma)
# -----------------------------------------------------------------------------
# We initialise this **once** at import‑time so that every subsequent call to
# `annotate_morpheme()` re‑uses the same pipeline (avoids re‑loading models).
# -----------------------------------------------------------------------------
nlp = stanza.Pipeline(
lang="en",
processors="tokenize,pos,lemma",
tokenize_pretokenized=False,
)
# -----------------------------------------------------------------------------
# Canonical suffix sets for *inflectional* morphemes (unchanged)
# -----------------------------------------------------------------------------
_EXPECTED_SUFFIXES = {
"Plural": {"s", "es"},
"Possessive": {"'s", "s"},
"Comparative": {"er"},
"Superlative": {"est"},
"3rd Person Singular": {"s", "es"},
"Past Tense": {"ed"},
"Past Participle": {"ed", "en", "n"},
"Progressive": {"ing"},
"Gerund": {"ing"},
}
# -----------------------------------------------------------------------------
# *New* : Contraction particles (clitics)
# -----------------------------------------------------------------------------
# Mapping from particle → canonical meaning. These are added as a *new* type
# "Contraction" in the output list.
# -----------------------------------------------------------------------------
_CONTRACTION_PARTICLES = {
"'ll": "will", # we'll, he'll …
"'d": "would/had", # I'd, she'd …
"'ve": "have", # we've, they've …
"'re": "are", # you're, they're …
"'m": "am", # I'm …
"n't": "not", # isn't, didn't …
"'s": "is/has", # what's, she's, Tom's(?) – see disambiguation below
}
_S_TOKENS = {"'s", "’s"} # keep both apostrophe forms
# -----------------------------------------------------------------------------
# Helper functions
# -----------------------------------------------------------------------------
def is_possessive_candidate(tok):
"""Return True if token text is 's / ’s and UD tag == PART."""
return tok.text in _S_TOKENS and tok.upos == "PART"
def lcp(a: str, b: str) -> str:
"""Longest common prefix (case‑insensitive)."""
i = 0
while i < min(len(a), len(b)) and a[i].lower() == b[i].lower():
i += 1
return a[:i]
def strip_doubling(lemma: str, suf: str) -> str:
"""Remove doubled final consonant when the suffix repeats it (stop + p + ing)."""
if suf and len(suf) >= 2 and suf[0] == lemma[-1]:
cand = suf[1:]
if any(cand in v for v in _EXPECTED_SUFFIXES.values()):
return cand
return suf
def get_suffix(lemma: str, surface: str) -> str:
"""Return raw suffix after common prefix is stripped and doubling handled."""
return strip_doubling(lemma, surface[len(lcp(lemma, surface)):])
def normalize_suffix(lemma: str, raw_suf: str, expected_set: set) -> str | None:
"""Bring irregular spelling variants back to canonical form (e.g. ies → s)."""
if raw_suf in expected_set:
return raw_suf
if lemma.lower().endswith("y") and raw_suf.startswith("i"):
alt = raw_suf[1:]
if alt in expected_set:
return alt
return None
# -----------------------------------------------------------------------------
# Core extractor
# -----------------------------------------------------------------------------
def extract_inflectional_morphemes(text: str):
"""Return list of inflectional & contraction morpheme annotations for *text*."""
doc = nlp(text)
results = []
for sent in doc.sentences:
words = sent.words
i = 0
while i < len(words):
w = words[i]
surf, lem, pos = w.text, w.lemma, w.upos
feats = {k: v for k, v in (f.split("=", 1) for f in (w.feats or "").split("|") if "=" in f)}
low_txt = surf.lower()
# -----------------------------------------------------------------
# 1) 's : Disambiguate Possessive vs Contraction
# -----------------------------------------------------------------
if is_possessive_candidate(w) and i > 0:
prev = words[i - 1]
# If the previous token is a NOUN/PROPN we *assume* possessive,
# otherwise treat it as a contraction for *is/has*.
if prev.upos in {"NOUN", "PROPN"}:
results.append({
"word": prev.text + surf,
"lemma": prev.lemma,
"index": i - 1,
"inflectional_morpheme": "Possessive",
"morpheme_form": "'/s",
})
else: # Contraction: what’s / she’s / it’s …
results.append({
"word": prev.text + surf,
"lemma": prev.lemma,
"index": i - 1,
"inflectional_morpheme": "Contraction",
"morpheme_form": "'/s",
})
i += 1
continue
# -----------------------------------------------------------------
# 2) Other contraction particles ( 'll, 're, 'm, 've, 'd, n't )
# -----------------------------------------------------------------
if low_txt in _CONTRACTION_PARTICLES and i > 0:
prev = words[i - 1]
results.append({
"word": prev.text + surf,
"lemma": prev.lemma,
"index": i - 1,
"inflectional_morpheme": "Contraction",
"morpheme_form": low_txt,
})
i += 1
continue
# -----------------------------------------------------------------
# 3) Possessive pronouns / determiners (his, yours …)
# -----------------------------------------------------------------
if feats.get("Poss") == "Yes" and pos in {"PRON", "DET"}:
low_lem, low_surf = lem.lower(), surf.lower()
suf = get_suffix(low_lem, low_surf)
morpheme_form = "/s" if suf in {"s", "es"} and low_lem + suf == low_surf else "<IRR>"
results.append({
"word": surf,
"lemma": lem,
"index": i,
"inflectional_morpheme": "Possessive",
"morpheme_form": morpheme_form,
})
i += 1
continue
# -----------------------------------------------------------------
# 4) Standard inflectional endings (plural, tense, degree …)
# -----------------------------------------------------------------
inflect_type = None
if pos == "NOUN" and feats.get("Number") == "Plur":
inflect_type = "Plural"
elif pos == "ADJ" and feats.get("Degree") == "Cmp":
inflect_type = "Comparative"
elif pos == "ADJ" and feats.get("Degree") == "Sup":
inflect_type = "Superlative"
elif pos == "VERB" and feats.get("VerbForm") == "Fin" and feats.get("Tense") == "Pres" and feats.get("Person") == "3":
inflect_type = "3rd Person Singular"
elif pos == "VERB" and feats.get("VerbForm") == "Fin" and feats.get("Tense") == "Past":
inflect_type = "Past Tense"
elif pos == "VERB" and feats.get("VerbForm") == "Part":
if feats.get("Tense") == "Past" or w.xpos == "VBN":
inflect_type = "Past Participle"
elif feats.get("Tense") == "Pres" or w.xpos == "VBG":
inflect_type = "Progressive"
if inflect_type:
raw_suffix = get_suffix(lem, low_txt)
canon = normalize_suffix(lem, raw_suffix, _EXPECTED_SUFFIXES[inflect_type])
morpheme_form = f"/{canon}" if canon else "<IRR>"
results.append({
"word": surf,
"lemma": lem,
"index": i,
"inflectional_morpheme": inflect_type,
"morpheme_form": morpheme_form,
})
i += 1
return results
# -----------------------------------------------------------------------------
# Pipeline entry‑point used by main_socket / other modules
# -----------------------------------------------------------------------------
def annotate_morpheme(session_id, base_dir="session_data"):
"""Annotate `{session_id}_transcriptionCW.json` with morpheme information."""
base_dir = base_dir or os.getcwd()
json_file = os.path.join(base_dir, f"{session_id}/{session_id}_transcriptionCW.json")
if not os.path.exists(json_file):
raise FileNotFoundError(f"{json_file} not found – make sure transcription step ran first.")
with open(json_file, "r", encoding="utf-8") as f:
data = json.load(f)
# Support both list‑of‑segments or {segments: [...]} formats
segments = data.get("segments", data) if isinstance(data, dict) else data
for seg in segments:
text = seg.get("text", "")
seg["morphemes"] = extract_inflectional_morphemes(text)
with open(json_file, "w", encoding="utf-8") as f:
json.dump(data, f, ensure_ascii=False, indent=2)
# Example usage inside main_socket.py
# -----------------------------------------------------------------------------
# from morpheme import annotate_morpheme
# def handle_session(session_id: str):
# ... # other processing steps
# annotate_morpheme(session_id, base_dir=session_data_dir)
# ... # return/serve updated JSON
|