Update app.py
Browse files
app.py
CHANGED
|
@@ -1,5 +1,5 @@
|
|
| 1 |
import os, re, json, base64
|
| 2 |
-
from typing import Dict, List
|
| 3 |
import gradio as gr
|
| 4 |
|
| 5 |
# ------------ Archivos ya existentes ------------
|
|
@@ -9,10 +9,11 @@ LEX_MASTER = "lexicon_master.json" # {entries:[{lemma_es,lemma_en,minimax,kom
|
|
| 9 |
|
| 10 |
# ------------ Normalización ------------
|
| 11 |
WORD_RE = re.compile(r"[A-Za-zÁÉÍÓÚÜÑáéíóúüñ]+", re.UNICODE)
|
| 12 |
-
PUNCT_SPLIT = re.compile(r"([^\w\s]+)")
|
| 13 |
STRIP = str.maketrans("ÁÉÍÓÚÜÑáéíóúüñ", "AEIOUUNaeiouun")
|
|
|
|
| 14 |
def norm_es(w: str) -> str:
|
| 15 |
return re.sub(r"[^a-záéíóúüñ]", "", (w or "").lower()).translate(STRIP)
|
|
|
|
| 16 |
def norm_en(w: str) -> str:
|
| 17 |
return re.sub(r"[^a-z]", "", (w or "").lower())
|
| 18 |
|
|
@@ -32,20 +33,36 @@ def load_lexicons():
|
|
| 32 |
mini2es = {v:k for k,v in es2mini.items()}
|
| 33 |
komi2es = {v:k for k,v in es2komi.items()}
|
| 34 |
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
if isinstance(master, dict) and "entries" in master:
|
| 37 |
for e in master["entries"]:
|
| 38 |
es = norm_es(str(e.get("lemma_es","")))
|
| 39 |
en = norm_en(str(e.get("lemma_en","")))
|
| 40 |
mi = str(e.get("minimax",""))
|
| 41 |
ko = str(e.get("komin",""))
|
| 42 |
-
if
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
|
|
|
| 47 |
|
| 48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
# ------------ OOV reversible ------------
|
| 51 |
ALPHA_MINI64 = "@ptkmnslraeiouy0123456789><=:/!?.+-_*#bcdfghjvqwxzACEGHIJKLMNOPRS"[:64]
|
|
@@ -110,6 +127,17 @@ def lemma_of(tok, src_lang: str) -> str:
|
|
| 110 |
else:
|
| 111 |
return norm_en(tok.lemma_ if tok.lemma_ else tok.text)
|
| 112 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
# ------------ mapeo lema→código (usa tus JSON) ------------
|
| 114 |
def code_es(lemma: str, target: str) -> str:
|
| 115 |
lemma = norm_es(lemma)
|
|
@@ -127,7 +155,7 @@ def code_en(lemma: str, target: str) -> str:
|
|
| 127 |
if EN2KOMI: return EN2KOMI.get(lemma) or enc_oov_komin(lemma)
|
| 128 |
return enc_oov_komin(lemma)
|
| 129 |
|
| 130 |
-
# ------------ reglas compactas ------------
|
| 131 |
TAM_MINI = {"Pres":"P", "Past":"T", "Fut":"F", "UNK":"P"}
|
| 132 |
TAM_KOMI = {"Pres":"Ⓟ", "Past":"Ⓣ", "Fut":"Ⓕ", "UNK":"Ⓟ"}
|
| 133 |
|
|
@@ -143,7 +171,7 @@ def detect_neg(doc) -> bool:
|
|
| 143 |
def detect_tense(root) -> str:
|
| 144 |
m = root.morph
|
| 145 |
if "Tense=Past" in m: return "Past"
|
| 146 |
-
if "Tense=Fut"
|
| 147 |
if "Tense=Pres" in m: return "Pres"
|
| 148 |
for c in root.children:
|
| 149 |
if c.pos_ == "AUX":
|
|
@@ -169,19 +197,21 @@ def realize_minimax(doc, src_lang: str, drop_articles=True, zero_copula=True):
|
|
| 169 |
root, subs, objs, obls, advs = extract_core(doc)
|
| 170 |
tense, is_q, is_neg = detect_tense(root), detect_polarity(doc), detect_neg(doc)
|
| 171 |
|
| 172 |
-
vlem
|
| 173 |
vcode = code_es(vlem, "Minimax-ASCII") if src_lang=="Español" else code_en(vlem, "Minimax-ASCII")
|
| 174 |
|
| 175 |
tail = TAM_MINI.get(tense, "P")
|
| 176 |
if is_neg: tail += "N"
|
| 177 |
if is_q: tail += "Q"
|
| 178 |
if tail:
|
| 179 |
-
vcode = f"{vcode}·{tail}"
|
| 180 |
|
| 181 |
def realize_np(tokens):
|
| 182 |
outs=[]
|
| 183 |
for t in tokens:
|
| 184 |
-
|
|
|
|
|
|
|
| 185 |
code = code_es(lem, "Minimax-ASCII") if src_lang=="Español" else code_en(lem, "Minimax-ASCII")
|
| 186 |
outs.append(code)
|
| 187 |
return outs
|
|
@@ -190,7 +220,9 @@ def realize_minimax(doc, src_lang: str, drop_articles=True, zero_copula=True):
|
|
| 190 |
O = realize_np(objs) + realize_np(obls)
|
| 191 |
ADV=[]
|
| 192 |
for a in advs:
|
| 193 |
-
|
|
|
|
|
|
|
| 194 |
ADV.append(code_es(lem, "Minimax-ASCII") if src_lang=="Español" else code_en(lem, "Minimax-ASCII"))
|
| 195 |
|
| 196 |
if zero_copula and vlem in ("ser","estar","be") and tense=="Pres" and not is_neg and not is_q:
|
|
@@ -214,7 +246,9 @@ def realize_komin(doc, src_lang: str, drop_articles=True, zero_copula=True):
|
|
| 214 |
def realize_np(tokens, particle):
|
| 215 |
outs=[]
|
| 216 |
for t in tokens:
|
| 217 |
-
|
|
|
|
|
|
|
| 218 |
code = code_es(lem, "Kōmín-CJK") if src_lang=="Español" else code_en(lem, "Kōmín-CJK")
|
| 219 |
outs.append(code + particle)
|
| 220 |
return outs
|
|
@@ -223,7 +257,9 @@ def realize_komin(doc, src_lang: str, drop_articles=True, zero_copula=True):
|
|
| 223 |
O = realize_np(objs + obls, P_OBJ)
|
| 224 |
ADV=[]
|
| 225 |
for a in advs:
|
| 226 |
-
|
|
|
|
|
|
|
| 227 |
ADV.append(code_es(lem, "Kōmín-CJK") if src_lang=="Español" else code_en(lem, "Kōmín-CJK"))
|
| 228 |
|
| 229 |
v_form = vcode + TAM + (NEG_M if is_neg else "")
|
|
@@ -270,8 +306,8 @@ def decode_simple(text: str, source: str, tgt_lang: str) -> str:
|
|
| 270 |
code2en = MINI2EN if source=="Minimax-ASCII" else KOMI2EN
|
| 271 |
|
| 272 |
mini_tail_re = re.compile(r"^(?P<stem>.+?)·(?P<tail>[PTFNQ]+)$")
|
| 273 |
-
komi_np_re = re.compile(r"(?P<stem>.+?)(?:[ᵖᵒ])$")
|
| 274 |
-
komi_tam_re = re.compile(r"^(?P<stem>.+?)[ⓅⓉⒻ]+(?:\u0306)?$")
|
| 275 |
|
| 276 |
tokens = re.split(r"(\s+|[^\w\s]+)", text)
|
| 277 |
out=[]
|
|
@@ -345,26 +381,118 @@ def decode_simple(text: str, source: str, tgt_lang: str) -> str:
|
|
| 345 |
|
| 346 |
return "".join(out)
|
| 347 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 348 |
# ------------ UI y explicaciones ------------
|
| 349 |
EXPLAIN_ES = """
|
| 350 |
-
**
|
| 351 |
-
- **Minimax (VSO)**: verbo
|
| 352 |
-
- **Kōmín (SOV)**: partículas **ᵖ/ᵒ**
|
| 353 |
-
-
|
|
|
|
| 354 |
"""
|
|
|
|
| 355 |
EXPLAIN_EN = """
|
| 356 |
-
**
|
| 357 |
-
- **Minimax (VSO)**: verb carries TAM/neg/question
|
| 358 |
-
- **Kōmín (SOV)**: particles **ᵖ/ᵒ**
|
| 359 |
-
-
|
|
|
|
| 360 |
"""
|
| 361 |
|
| 362 |
-
with gr.Blocks(title="
|
| 363 |
-
gr.Markdown("# Minimax-ASCII / Kōmín-CJK
|
| 364 |
with gr.Row():
|
| 365 |
gr.Markdown(EXPLAIN_ES)
|
| 366 |
gr.Markdown(EXPLAIN_EN)
|
| 367 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 368 |
with gr.Tab("Construir frase (ES/EN → Conlang)"):
|
| 369 |
with gr.Row():
|
| 370 |
src_lang = gr.Dropdown(["Español","English"], value="Español", label="Fuente")
|
|
@@ -376,7 +504,8 @@ with gr.Blocks(title="Conlangs · Reversible Phrase Builder") as demo:
|
|
| 376 |
out = gr.Textbox(lines=6, label="Salida")
|
| 377 |
gr.Button("Construir").click(build_sentence, [text_in, src_lang, target, drop_articles, zero_copula], [out])
|
| 378 |
|
| 379 |
-
|
|
|
|
| 380 |
with gr.Row():
|
| 381 |
src_lang2 = gr.Dropdown(["Español","English"], value="Español", label="Fuente")
|
| 382 |
target2 = gr.Dropdown(["Minimax-ASCII","Kōmín-CJK"], value="Minimax-ASCII", label="Conlang")
|
|
@@ -392,6 +521,23 @@ with gr.Blocks(title="Conlangs · Reversible Phrase Builder") as demo:
|
|
| 392 |
out3 = gr.Textbox(lines=6, label="Salida")
|
| 393 |
gr.Button("Decodificar").click(decode_simple, [code_in, src_code, tgt_lang], [out3])
|
| 394 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 395 |
if __name__ == "__main__":
|
| 396 |
demo.launch()
|
| 397 |
|
|
@@ -401,3 +547,4 @@ if __name__ == "__main__":
|
|
| 401 |
|
| 402 |
|
| 403 |
|
|
|
|
|
|
| 1 |
import os, re, json, base64
|
| 2 |
+
from typing import Dict, List, Tuple
|
| 3 |
import gradio as gr
|
| 4 |
|
| 5 |
# ------------ Archivos ya existentes ------------
|
|
|
|
| 9 |
|
| 10 |
# ------------ Normalización ------------
|
| 11 |
WORD_RE = re.compile(r"[A-Za-zÁÉÍÓÚÜÑáéíóúüñ]+", re.UNICODE)
|
|
|
|
| 12 |
STRIP = str.maketrans("ÁÉÍÓÚÜÑáéíóúüñ", "AEIOUUNaeiouun")
|
| 13 |
+
|
| 14 |
def norm_es(w: str) -> str:
|
| 15 |
return re.sub(r"[^a-záéíóúüñ]", "", (w or "").lower()).translate(STRIP)
|
| 16 |
+
|
| 17 |
def norm_en(w: str) -> str:
|
| 18 |
return re.sub(r"[^a-z]", "", (w or "").lower())
|
| 19 |
|
|
|
|
| 33 |
mini2es = {v:k for k,v in es2mini.items()}
|
| 34 |
komi2es = {v:k for k,v in es2komi.items()}
|
| 35 |
|
| 36 |
+
# Tablas naturales ↔ naturales a partir del master
|
| 37 |
+
es2en_lemma: Dict[str,str] = {}
|
| 38 |
+
en2es_lemma: Dict[str,str] = {}
|
| 39 |
+
|
| 40 |
+
en2mini, en2komi = {}, {}
|
| 41 |
+
mini2en, komi2en = {}, {}
|
| 42 |
+
|
| 43 |
if isinstance(master, dict) and "entries" in master:
|
| 44 |
for e in master["entries"]:
|
| 45 |
es = norm_es(str(e.get("lemma_es","")))
|
| 46 |
en = norm_en(str(e.get("lemma_en","")))
|
| 47 |
mi = str(e.get("minimax",""))
|
| 48 |
ko = str(e.get("komin",""))
|
| 49 |
+
if es and en:
|
| 50 |
+
# guardamos el primero estable
|
| 51 |
+
es2en_lemma.setdefault(es, en)
|
| 52 |
+
en2es_lemma.setdefault(en, es)
|
| 53 |
+
if en and mi: en2mini.setdefault(en, mi)
|
| 54 |
+
if en and ko: en2komi.setdefault(en, ko)
|
| 55 |
|
| 56 |
+
mini2en = {v:k for k,v in en2mini.items()}
|
| 57 |
+
komi2en = {v:k for k,v in en2komi.items()}
|
| 58 |
+
|
| 59 |
+
return (es2mini, es2komi, mini2es, komi2es,
|
| 60 |
+
en2mini, en2komi, mini2en, komi2en,
|
| 61 |
+
es2en_lemma, en2es_lemma)
|
| 62 |
+
|
| 63 |
+
(ES2MINI, ES2KOMI, MINI2ES, KOMI2ES,
|
| 64 |
+
EN2MINI, EN2KOMI, MINI2EN, KOMI2EN,
|
| 65 |
+
ES2EN_LEMMA, EN2ES_LEMMA) = load_lexicons()
|
| 66 |
|
| 67 |
# ------------ OOV reversible ------------
|
| 68 |
ALPHA_MINI64 = "@ptkmnslraeiouy0123456789><=:/!?.+-_*#bcdfghjvqwxzACEGHIJKLMNOPRS"[:64]
|
|
|
|
| 127 |
else:
|
| 128 |
return norm_en(tok.lemma_ if tok.lemma_ else tok.text)
|
| 129 |
|
| 130 |
+
# ------------ filtros de contenido (evita “hola”, “cómo”, etc. en el núcleo) ------------
|
| 131 |
+
def is_content_token(t) -> bool:
|
| 132 |
+
if t.pos_ in ("INTJ", "DET", "ADP", "SCONJ", "PART", "SYM", "PUNCT"):
|
| 133 |
+
return False
|
| 134 |
+
if t.tag_.startswith("W") or t.lower_ in (
|
| 135 |
+
"como","cómo","que","qué","quien","quién","donde","dónde","cuando","cuándo",
|
| 136 |
+
"por","porqué","por qué","how","what","who","where","when","why"
|
| 137 |
+
):
|
| 138 |
+
return False
|
| 139 |
+
return True
|
| 140 |
+
|
| 141 |
# ------------ mapeo lema→código (usa tus JSON) ------------
|
| 142 |
def code_es(lemma: str, target: str) -> str:
|
| 143 |
lemma = norm_es(lemma)
|
|
|
|
| 155 |
if EN2KOMI: return EN2KOMI.get(lemma) or enc_oov_komin(lemma)
|
| 156 |
return enc_oov_komin(lemma)
|
| 157 |
|
| 158 |
+
# ------------ reglas compactas (fraseador) ------------
|
| 159 |
TAM_MINI = {"Pres":"P", "Past":"T", "Fut":"F", "UNK":"P"}
|
| 160 |
TAM_KOMI = {"Pres":"Ⓟ", "Past":"Ⓣ", "Fut":"Ⓕ", "UNK":"Ⓟ"}
|
| 161 |
|
|
|
|
| 171 |
def detect_tense(root) -> str:
|
| 172 |
m = root.morph
|
| 173 |
if "Tense=Past" in m: return "Past"
|
| 174 |
+
if "Tense=Fut" in m: return "Fut"
|
| 175 |
if "Tense=Pres" in m: return "Pres"
|
| 176 |
for c in root.children:
|
| 177 |
if c.pos_ == "AUX":
|
|
|
|
| 197 |
root, subs, objs, obls, advs = extract_core(doc)
|
| 198 |
tense, is_q, is_neg = detect_tense(root), detect_polarity(doc), detect_neg(doc)
|
| 199 |
|
| 200 |
+
vlem = lemma_of(root, src_lang)
|
| 201 |
vcode = code_es(vlem, "Minimax-ASCII") if src_lang=="Español" else code_en(vlem, "Minimax-ASCII")
|
| 202 |
|
| 203 |
tail = TAM_MINI.get(tense, "P")
|
| 204 |
if is_neg: tail += "N"
|
| 205 |
if is_q: tail += "Q"
|
| 206 |
if tail:
|
| 207 |
+
vcode = f"{vcode}·{tail}" # separador para decodificar
|
| 208 |
|
| 209 |
def realize_np(tokens):
|
| 210 |
outs=[]
|
| 211 |
for t in tokens:
|
| 212 |
+
if not is_content_token(t):
|
| 213 |
+
continue
|
| 214 |
+
lem = lemma_of(t, src_lang)
|
| 215 |
code = code_es(lem, "Minimax-ASCII") if src_lang=="Español" else code_en(lem, "Minimax-ASCII")
|
| 216 |
outs.append(code)
|
| 217 |
return outs
|
|
|
|
| 220 |
O = realize_np(objs) + realize_np(obls)
|
| 221 |
ADV=[]
|
| 222 |
for a in advs:
|
| 223 |
+
if not is_content_token(a):
|
| 224 |
+
continue
|
| 225 |
+
lem = lemma_of(a, src_lang)
|
| 226 |
ADV.append(code_es(lem, "Minimax-ASCII") if src_lang=="Español" else code_en(lem, "Minimax-ASCII"))
|
| 227 |
|
| 228 |
if zero_copula and vlem in ("ser","estar","be") and tense=="Pres" and not is_neg and not is_q:
|
|
|
|
| 246 |
def realize_np(tokens, particle):
|
| 247 |
outs=[]
|
| 248 |
for t in tokens:
|
| 249 |
+
if not is_content_token(t):
|
| 250 |
+
continue
|
| 251 |
+
lem = lemma_of(t, src_lang)
|
| 252 |
code = code_es(lem, "Kōmín-CJK") if src_lang=="Español" else code_en(lem, "Kōmín-CJK")
|
| 253 |
outs.append(code + particle)
|
| 254 |
return outs
|
|
|
|
| 257 |
O = realize_np(objs + obls, P_OBJ)
|
| 258 |
ADV=[]
|
| 259 |
for a in advs:
|
| 260 |
+
if not is_content_token(a):
|
| 261 |
+
continue
|
| 262 |
+
lem = lemma_of(a, src_lang)
|
| 263 |
ADV.append(code_es(lem, "Kōmín-CJK") if src_lang=="Español" else code_en(lem, "Kōmín-CJK"))
|
| 264 |
|
| 265 |
v_form = vcode + TAM + (NEG_M if is_neg else "")
|
|
|
|
| 306 |
code2en = MINI2EN if source=="Minimax-ASCII" else KOMI2EN
|
| 307 |
|
| 308 |
mini_tail_re = re.compile(r"^(?P<stem>.+?)·(?P<tail>[PTFNQ]+)$")
|
| 309 |
+
komi_np_re = re.compile(r"(?P<stem>.+?)(?:[ᵖᵒ])$")
|
| 310 |
+
komi_tam_re = re.compile(r"^(?P<stem>.+?)[ⓅⓉⒻ]+(?:\u0306)?$")
|
| 311 |
|
| 312 |
tokens = re.split(r"(\s+|[^\w\s]+)", text)
|
| 313 |
out=[]
|
|
|
|
| 381 |
|
| 382 |
return "".join(out)
|
| 383 |
|
| 384 |
+
# ------------ Natural ↔ Natural (lemas) ------------
|
| 385 |
+
def translate_natural(text: str, src_lang: str, tgt_lang: str) -> str:
|
| 386 |
+
"""
|
| 387 |
+
Traducción de ES<->EN a nivel de lema usando lexicon_master.json.
|
| 388 |
+
Devuelve lemas (no inflexiona).
|
| 389 |
+
"""
|
| 390 |
+
if not text.strip(): return ""
|
| 391 |
+
if not USE_SPACY: return "[Error] spaCy no está cargado (revisa requirements.txt)."
|
| 392 |
+
nlp = nlp_es if src_lang=="Español" else nlp_en
|
| 393 |
+
doc = nlp(text)
|
| 394 |
+
out=[]
|
| 395 |
+
for t in doc:
|
| 396 |
+
if not t.is_alpha:
|
| 397 |
+
out.append(t.text)
|
| 398 |
+
continue
|
| 399 |
+
lem = lemma_of(t, src_lang)
|
| 400 |
+
if src_lang=="Español": # ES -> EN
|
| 401 |
+
tr = ES2EN_LEMMA.get(lem)
|
| 402 |
+
out.append(tr if tr else lem)
|
| 403 |
+
else: # EN -> ES
|
| 404 |
+
tr = EN2ES_LEMMA.get(lem)
|
| 405 |
+
out.append(tr if tr else lem)
|
| 406 |
+
return " ".join(out)
|
| 407 |
+
|
| 408 |
+
# ------------ Round-trip (ida→vuelta) ------------
|
| 409 |
+
def round_trip(text: str, src_lang: str, target: str, drop_articles: bool, zero_copula: bool):
|
| 410 |
+
if not text.strip():
|
| 411 |
+
return "", ""
|
| 412 |
+
conlang_out = build_sentence(text, src_lang, target, drop_articles, zero_copula)
|
| 413 |
+
back_out = decode_simple(conlang_out, target, src_lang)
|
| 414 |
+
return conlang_out, back_out
|
| 415 |
+
|
| 416 |
+
# ------------ Traductor universal (4-way) ------------
|
| 417 |
+
ALL_LANGS = ["Español","English","Minimax-ASCII","Kōmín-CJK"]
|
| 418 |
+
|
| 419 |
+
def universal_translate(text: str, src: str, tgt: str,
|
| 420 |
+
drop_articles: bool, zero_copula: bool) -> str:
|
| 421 |
+
if not text.strip(): return ""
|
| 422 |
+
if src == tgt: return text
|
| 423 |
+
|
| 424 |
+
# Natural → Conlang
|
| 425 |
+
if src in ("Español","English") and tgt in ("Minimax-ASCII","Kōmín-CJK"):
|
| 426 |
+
return build_sentence(text, src, tgt, drop_articles, zero_copula)
|
| 427 |
+
|
| 428 |
+
# Conlang → Natural
|
| 429 |
+
if src in ("Minimax-ASCII","Kōmín-CJK") and tgt in ("Español","English"):
|
| 430 |
+
return decode_simple(text, src, tgt)
|
| 431 |
+
|
| 432 |
+
# Natural ↔ Natural (lemas)
|
| 433 |
+
if src in ("Español","English") and tgt in ("Español","English"):
|
| 434 |
+
return translate_natural(text, src, tgt)
|
| 435 |
+
|
| 436 |
+
# Conlang ↔ Conlang: Conlang_src -> ES (lemmas) -> Conlang_tgt (lexical)
|
| 437 |
+
if src in ("Minimax-ASCII","Kōmín-CJK") and tgt in ("Minimax-ASCII","Kōmín-CJK"):
|
| 438 |
+
# 1) decodifica a Español (preferimos ES porque el léxico base es ES)
|
| 439 |
+
es_lemmas = decode_simple(text, src, "Español")
|
| 440 |
+
# 2) lexical encode (simple) al conlang destino
|
| 441 |
+
# nota: aquí no fraseamos; mapeamos palabra a palabra
|
| 442 |
+
words = re.findall(r"\w+|[^\w\s]+", es_lemmas)
|
| 443 |
+
out=[]
|
| 444 |
+
for w in words:
|
| 445 |
+
if re.fullmatch(r"\w+", w):
|
| 446 |
+
code = ES2MINI.get(norm_es(w)) if tgt=="Minimax-ASCII" else ES2KOMI.get(norm_es(w))
|
| 447 |
+
if not code:
|
| 448 |
+
code = enc_oov_minimax(w) if tgt=="Minimax-ASCII" else enc_oov_komin(w)
|
| 449 |
+
out.append(code)
|
| 450 |
+
else:
|
| 451 |
+
out.append(w)
|
| 452 |
+
return " ".join(out)
|
| 453 |
+
|
| 454 |
+
return "[No soportado]"
|
| 455 |
+
|
| 456 |
# ------------ UI y explicaciones ------------
|
| 457 |
EXPLAIN_ES = """
|
| 458 |
+
**Reversibilidad garantizada (ES/EN ↔ conlangs).**
|
| 459 |
+
- **Minimax (VSO)**: verbo con sufijos **TAM/neg/pregunta** separados por **`·`** del código léxico (p. ej. `código·PTFNQ`).
|
| 460 |
+
- **Kōmín (SOV)**: partículas **ᵖ/ᵒ**, marcas **Ⓟ/Ⓣ/Ⓕ**, negación breve y **「?」** final. El decodificador **pela** todo antes de buscar el código base.
|
| 461 |
+
- **OOV**: tokens fuera de diccionario se envuelven de forma **reversible** (`~…` / `「…」`).
|
| 462 |
+
- **Naturales**: ES↔EN se hace a nivel de **lema** usando `lexicon_master.json` (no hay inflexión).
|
| 463 |
"""
|
| 464 |
+
|
| 465 |
EXPLAIN_EN = """
|
| 466 |
+
**Full reversibility (EN/ES ↔ conlangs).**
|
| 467 |
+
- **Minimax (VSO)**: verb carries compact **TAM/neg/question** after a reserved separator **`·`** (e.g. `code·PTFNQ`).
|
| 468 |
+
- **Kōmín (SOV)**: particles **ᵖ/ᵒ**, verb **Ⓟ/Ⓣ/Ⓕ**, short neg mark, final **「?」**. Decoder strips them before lookup.
|
| 469 |
+
- **OOV**: unknown tokens are wrapped **reversibly** (`~…` / `「…」`).
|
| 470 |
+
- **Natural languages**: EN↔ES is lemma-level via `lexicon_master.json` (no inflection).
|
| 471 |
"""
|
| 472 |
|
| 473 |
+
with gr.Blocks(title="Universal Conlang Translator (Reversible)") as demo:
|
| 474 |
+
gr.Markdown("# Universal Conlang Translator · Minimax-ASCII / Kōmín-CJK")
|
| 475 |
with gr.Row():
|
| 476 |
gr.Markdown(EXPLAIN_ES)
|
| 477 |
gr.Markdown(EXPLAIN_EN)
|
| 478 |
|
| 479 |
+
# --- Traducir (universal) ---
|
| 480 |
+
with gr.Tab("Traducir (universal)"):
|
| 481 |
+
with gr.Row():
|
| 482 |
+
uni_src = gr.Dropdown(ALL_LANGS, value="Español", label="Fuente")
|
| 483 |
+
uni_tgt = gr.Dropdown(ALL_LANGS, value="Minimax-ASCII", label="Destino")
|
| 484 |
+
uni_text = gr.Textbox(lines=3, label="Texto", value="Hola, ¿cómo estás?")
|
| 485 |
+
with gr.Row():
|
| 486 |
+
uni_drop = gr.Checkbox(value=True, label="Omitir artículos (ES/EN→conlang)")
|
| 487 |
+
uni_zero = gr.Checkbox(value=True, label="Cópula cero (presente afirm.) (ES/EN→conlang)")
|
| 488 |
+
uni_out = gr.Textbox(lines=6, label="Traducción")
|
| 489 |
+
gr.Button("Traducir").click(
|
| 490 |
+
universal_translate,
|
| 491 |
+
[uni_text, uni_src, uni_tgt, uni_drop, uni_zero],
|
| 492 |
+
[uni_out]
|
| 493 |
+
)
|
| 494 |
+
|
| 495 |
+
# --- Construir frase (natural → conlang) ---
|
| 496 |
with gr.Tab("Construir frase (ES/EN → Conlang)"):
|
| 497 |
with gr.Row():
|
| 498 |
src_lang = gr.Dropdown(["Español","English"], value="Español", label="Fuente")
|
|
|
|
| 504 |
out = gr.Textbox(lines=6, label="Salida")
|
| 505 |
gr.Button("Construir").click(build_sentence, [text_in, src_lang, target, drop_articles, zero_copula], [out])
|
| 506 |
|
| 507 |
+
# --- Codificar / Decodificar simples ---
|
| 508 |
+
with gr.Tab("Codificar simple (sólo léxico)"):
|
| 509 |
with gr.Row():
|
| 510 |
src_lang2 = gr.Dropdown(["Español","English"], value="Español", label="Fuente")
|
| 511 |
target2 = gr.Dropdown(["Minimax-ASCII","Kōmín-CJK"], value="Minimax-ASCII", label="Conlang")
|
|
|
|
| 521 |
out3 = gr.Textbox(lines=6, label="Salida")
|
| 522 |
gr.Button("Decodificar").click(decode_simple, [code_in, src_code, tgt_lang], [out3])
|
| 523 |
|
| 524 |
+
# --- Round-trip (ida→vuelta) ---
|
| 525 |
+
with gr.Tab("Prueba de reversibilidad / Round-trip"):
|
| 526 |
+
with gr.Row():
|
| 527 |
+
rt_src = gr.Dropdown(["Español","English"], value="Español", label="Fuente (ida y vuelta)")
|
| 528 |
+
rt_tgt = gr.Dropdown(["Minimax-ASCII","Kōmín-CJK"], value="Minimax-ASCII", label="Conlang")
|
| 529 |
+
rt_text = gr.Textbox(lines=3, label="Frase de prueba", value="Hola, ¿cómo estás?")
|
| 530 |
+
with gr.Row():
|
| 531 |
+
rt_drop = gr.Checkbox(value=True, label="Omitir artículos")
|
| 532 |
+
rt_zero = gr.Checkbox(value=True, label="Cópula cero (presente afirm.)")
|
| 533 |
+
rt_out_conlang = gr.Textbox(lines=3, label="Conlang (ida)")
|
| 534 |
+
rt_out_back = gr.Textbox(lines=3, label="Vuelta (lemas en idioma fuente)")
|
| 535 |
+
gr.Button("Probar ida → vuelta", variant="primary").click(
|
| 536 |
+
round_trip,
|
| 537 |
+
[rt_text, rt_src, rt_tgt, rt_drop, rt_zero],
|
| 538 |
+
[rt_out_conlang, rt_out_back]
|
| 539 |
+
)
|
| 540 |
+
|
| 541 |
if __name__ == "__main__":
|
| 542 |
demo.launch()
|
| 543 |
|
|
|
|
| 547 |
|
| 548 |
|
| 549 |
|
| 550 |
+
|