Update app.py
Browse files
app.py
CHANGED
|
@@ -164,12 +164,15 @@ def is_content_token(t) -> bool:
|
|
| 164 |
if t.dep_ in ("discourse", "intj", "vocative", "dep"):
|
| 165 |
return False
|
| 166 |
low = t.lower_.strip("¿?¡!.,;:()[]{}\"'").lower()
|
| 167 |
-
|
|
|
|
| 168 |
"como","cómo","que","qué","quien","quién","donde","dónde","cuando","cuándo",
|
| 169 |
"porqué","por","por qué","cuanto","cuánto",
|
| 170 |
"which","what","who","where","when","why","how",
|
| 171 |
-
|
| 172 |
-
|
|
|
|
|
|
|
| 173 |
return False
|
| 174 |
return True
|
| 175 |
|
|
@@ -298,16 +301,22 @@ def realize_minimax(doc, src_lang: str, drop_articles=True, zero_copula=True, se
|
|
| 298 |
S = realize_np(subs)
|
| 299 |
O = realize_np(objs) + realize_np(obls)
|
| 300 |
ADV=[]
|
|
|
|
| 301 |
for a in advs:
|
| 302 |
if not USE_SPACY or is_content_token(a):
|
| 303 |
lem = lemma_of(a, src_lang) if USE_SPACY else a.text
|
| 304 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 305 |
|
| 306 |
if zero_copula and not semi_lossless and vlem in ("ser","estar","be") and tense=="Pres" and not is_neg and not is_q:
|
| 307 |
parts = S + O + ADV
|
| 308 |
else:
|
| 309 |
parts = [vcode] + S + O + ADV
|
| 310 |
-
|
|
|
|
| 311 |
|
| 312 |
def realize_komin(doc, src_lang: str, drop_articles=True, zero_copula=True, semi_lossless=False, person_hint="2s"):
|
| 313 |
root, subs, objs, obls, advs = extract_core(doc)
|
|
@@ -354,7 +363,7 @@ def realize_komin(doc, src_lang: str, drop_articles=True, zero_copula=True, semi
|
|
| 354 |
return out
|
| 355 |
|
| 356 |
# ------------ Lossless (Base85 comprimido) ------------
|
| 357 |
-
SIDECAR_B85_RE = re.compile(r"\s?§
|
| 358 |
|
| 359 |
def b85_enc_raw(s: str) -> str:
|
| 360 |
comp = zlib.compress(s.encode("utf-8"), 9)
|
|
@@ -530,6 +539,11 @@ def decode_simple(text: str, source: str, tgt_lang: str) -> str:
|
|
| 530 |
parts = [p for p in [S, v_conj] + O_ADV if p]
|
| 531 |
out_text = " ".join(parts)
|
| 532 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 533 |
# Pregunta
|
| 534 |
if has_q:
|
| 535 |
start_q = "¿" if tgt_lang == "Español" else ""
|
|
@@ -789,4 +803,3 @@ if __name__ == "__main__":
|
|
| 789 |
|
| 790 |
|
| 791 |
|
| 792 |
-
|
|
|
|
| 164 |
if t.dep_ in ("discourse", "intj", "vocative", "dep"):
|
| 165 |
return False
|
| 166 |
low = t.lower_.strip("¿?¡!.,;:()[]{}\"'").lower()
|
| 167 |
+
# Permite wh en preguntas (advmod/obl)
|
| 168 |
+
is_wh = t.tag_.startswith("W") or low in {
|
| 169 |
"como","cómo","que","qué","quien","quién","donde","dónde","cuando","cuándo",
|
| 170 |
"porqué","por","por qué","cuanto","cuánto",
|
| 171 |
"which","what","who","where","when","why","how",
|
| 172 |
+
}
|
| 173 |
+
if is_wh and t.dep_ not in ("advmod", "obl") and "?" not in t.doc.text:
|
| 174 |
+
return False
|
| 175 |
+
if low in {"hola","hello","hi","hey","adios","adiós","ciao"}:
|
| 176 |
return False
|
| 177 |
return True
|
| 178 |
|
|
|
|
| 301 |
S = realize_np(subs)
|
| 302 |
O = realize_np(objs) + realize_np(obls)
|
| 303 |
ADV=[]
|
| 304 |
+
wh_adv = [] # Para wh en Q
|
| 305 |
for a in advs:
|
| 306 |
if not USE_SPACY or is_content_token(a):
|
| 307 |
lem = lemma_of(a, src_lang) if USE_SPACY else a.text
|
| 308 |
+
code = code_es(lem, "Minimax-ASCII") if src_lang=="Español" else code_en(lem, "Minimax-ASCII")
|
| 309 |
+
if is_q and a.dep_ == "advmod" and a.tag_.startswith("W"):
|
| 310 |
+
wh_adv.append(code)
|
| 311 |
+
else:
|
| 312 |
+
ADV.append(code)
|
| 313 |
|
| 314 |
if zero_copula and not semi_lossless and vlem in ("ser","estar","be") and tense=="Pres" and not is_neg and not is_q:
|
| 315 |
parts = S + O + ADV
|
| 316 |
else:
|
| 317 |
parts = [vcode] + S + O + ADV
|
| 318 |
+
full_parts = wh_adv + parts # Wh al frente si Q
|
| 319 |
+
return " ".join(p for p in full_parts if p)
|
| 320 |
|
| 321 |
def realize_komin(doc, src_lang: str, drop_articles=True, zero_copula=True, semi_lossless=False, person_hint="2s"):
|
| 322 |
root, subs, objs, obls, advs = extract_core(doc)
|
|
|
|
| 363 |
return out
|
| 364 |
|
| 365 |
# ------------ Lossless (Base85 comprimido) ------------
|
| 366 |
+
SIDECAR_B85_RE = re.compile(r"\s?§\((?P<b85>[A-Za-z0-9!#$%&()*+\-;<=>?@^_`{|}~]+)\)$")
|
| 367 |
|
| 368 |
def b85_enc_raw(s: str) -> str:
|
| 369 |
comp = zlib.compress(s.encode("utf-8"), 9)
|
|
|
|
| 539 |
parts = [p for p in [S, v_conj] + O_ADV if p]
|
| 540 |
out_text = " ".join(parts)
|
| 541 |
|
| 542 |
+
# Wh en Q: si primer token es wh, muévelo al frente
|
| 543 |
+
if has_q and lemma_tokens and lemma_tokens[0].lower() in {"como", "cómo", "what", "how"}:
|
| 544 |
+
wh = lemma_tokens.pop(0)
|
| 545 |
+
out_text = f"{wh} {out_text}"
|
| 546 |
+
|
| 547 |
# Pregunta
|
| 548 |
if has_q:
|
| 549 |
start_q = "¿" if tgt_lang == "Español" else ""
|
|
|
|
| 803 |
|
| 804 |
|
| 805 |
|
|
|