Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,6 +2,7 @@
|
|
2 |
"""
|
3 |
Enhanced Dialect Bengali Translator with Semantic Search
|
4 |
Uses both text similarity and semantic pattern matching
|
|
|
5 |
"""
|
6 |
|
7 |
import difflib
|
@@ -10,15 +11,16 @@ import gradio as gr
|
|
10 |
from collections import defaultdict
|
11 |
import re
|
12 |
|
13 |
-
# === Phrase data: [Dialect
|
14 |
phrases_data = [
|
15 |
-
|
16 |
-
["
|
17 |
-
["
|
18 |
-
["
|
19 |
-
["
|
20 |
-
["
|
21 |
-
["oise
|
|
|
22 |
["jaite ni", "জাইতে নি", "যাবে কি?", "jabe ki?"],
|
23 |
["or ni", "ওর নি", "হচ্ছে কি?", "hocche ki?"],
|
24 |
["or", "ওর", "হচ্ছে", "hocche"],
|
@@ -26,13 +28,7 @@ phrases_data = [
|
|
26 |
["bala ni", "বালা নি", "ভালো কি?", "bhalo ki?"],
|
27 |
["or je", "ওর যে", "হচ্ছে যে", "hocche je"],
|
28 |
["jaibe ni", "জাইবে নি", "যাবে কি?", "jabe ki?"],
|
29 |
-
["jare ni", "
|
30 |
-
["ami jaimu", "আমি জাইমু", "আমি যাব", "ami jabo"],
|
31 |
-
["jaimu", "জাইমু", "যাব", "jabo"],
|
32 |
-
["jaibo", "জাইবো", "যাবে", "jabe"],
|
33 |
-
["oibo", "ওইবো", "হবে", "hobe"],
|
34 |
-
["oibo jen", "ওইবো জেন", "হবে যে", "hobe je"],
|
35 |
-
["Goto kali", "গোতো কালি", "গত কাল", "goto kal"],
|
36 |
["Kita kobor?", "কিতা খবর?", "কি খবর?", "ki khobor?"],
|
37 |
["Kita korde?", "কিতা কোর্দে?", "কি করছে?", "ki korchho?"],
|
38 |
["acha oibo-tik ase", "আচা ওইবো-তিক আসে", "ঠিক আছে", "thik ache"],
|
@@ -42,43 +38,112 @@ phrases_data = [
|
|
42 |
["se hole", "সে হলে", "তাহলে", "tahole"],
|
43 |
["Sob bala asoin ni", "সব বালা আসইন নি", "সব ভালো আছে কি?", "sob bhalo ache ki?"],
|
44 |
["Sob bala ase", "সব বালা আসে", "সব ভালো আছে", "sob bhalo ache"],
|
45 |
-
["Sob bala", "সব বালা", "সব ভালো", "sob bhalo"],
|
46 |
["asoini", "আসইনি", "আছে কি?", "ache ki?"],
|
47 |
-
["ase ni", "আছে নি", "আছে কি?", "ache ki?"],
|
48 |
["ase", "আসে", "আছে", "ache"],
|
49 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
]
|
51 |
|
52 |
-
# Semantic mapping of dialect patterns to meanings
|
53 |
semantic_patterns = {
|
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 |
# Precompute data structures for matching
|
@@ -86,7 +151,7 @@ dialects = [p[0] for p in phrases_data]
|
|
86 |
dialects_lower = [d.lower() for d in dialects]
|
87 |
actual_bengali_list = [p[2] for p in phrases_data]
|
88 |
|
89 |
-
# Create a mapping from dialect to
|
90 |
dialect_to_all = {p[0].lower(): p for p in phrases_data}
|
91 |
|
92 |
def semantic_analysis(user_input):
|
@@ -94,39 +159,45 @@ def semantic_analysis(user_input):
|
|
94 |
user_lower = user_input.lower()
|
95 |
detected_patterns = []
|
96 |
meaning_components = []
|
97 |
-
|
98 |
-
#
|
99 |
for pattern, info in semantic_patterns.items():
|
100 |
-
|
101 |
-
|
102 |
-
|
103 |
-
|
|
|
|
|
|
|
|
|
104 |
return detected_patterns, meaning_components
|
105 |
|
106 |
-
def find_semantic_matches(user_input, threshold=0.
|
107 |
-
"""Find matches based on semantic similarity"""
|
108 |
user_lower = user_input.lower()
|
109 |
matches = []
|
110 |
-
|
111 |
# Get semantic patterns from user input
|
112 |
detected_patterns, meaning_components = semantic_analysis(user_input)
|
113 |
-
|
114 |
# If we found semantic patterns, look for phrases with similar meanings
|
115 |
if meaning_components:
|
116 |
for i, (dialect, dialect_bengali, actual, benglish) in enumerate(phrases_data):
|
117 |
-
|
118 |
-
|
119 |
for meaning in meaning_components:
|
120 |
if meaning in actual:
|
121 |
-
match_score += 0.
|
122 |
-
|
123 |
-
|
|
|
|
|
124 |
text_similarity = difflib.SequenceMatcher(None, user_lower, dialect.lower()).ratio()
|
125 |
-
total_score = match_score + (text_similarity * 0.
|
126 |
-
|
127 |
if total_score > threshold:
|
128 |
matches.append((i, total_score, "semantic"))
|
129 |
-
|
130 |
return matches
|
131 |
|
132 |
def format_suggestions_from_indices(indices, match_type="text", scores=None):
|
@@ -134,16 +205,16 @@ def format_suggestions_from_indices(indices, match_type="text", scores=None):
|
|
134 |
lines = []
|
135 |
for i, idx in enumerate(indices):
|
136 |
d, dialect_bengali, actual, benglish = phrases_data[idx]
|
137 |
-
|
138 |
score_str = ""
|
139 |
if scores is not None and i < len(scores):
|
140 |
s_pct = int(scores[i] * 100)
|
141 |
score_str = f" ({match_type}-match: {s_pct}%)"
|
142 |
-
|
143 |
lines.append(f"• {d}{score_str}\n Dialect Bengali: {dialect_bengali}\n Actual Bengali: {actual}\n Benglish: {benglish}")
|
144 |
return "\n\n".join(lines)
|
145 |
|
146 |
-
def translate_text(user_text, top_k: int =
|
147 |
"""
|
148 |
Returns: (dialect_out, actual_out, benglish_out, suggestions_out)
|
149 |
"""
|
@@ -159,71 +230,62 @@ def translate_text(user_text, top_k: int = 5):
|
|
159 |
if q_lower == dialect.lower():
|
160 |
return dialect_bengali, actual, benglish, "✅ EXACT MATCH (100%)"
|
161 |
|
162 |
-
# 2)
|
163 |
-
potential_phrases = re.split(r'[.,;!?]\s*', q)
|
164 |
if len(potential_phrases) > 1:
|
165 |
results = []
|
166 |
for phrase in potential_phrases:
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
if results:
|
177 |
-
return "", "", "", "Multiple phrases detected:\n\n" + "\n\n".join(results)
|
178 |
|
179 |
# 3) Semantic matches
|
180 |
semantic_matches = find_semantic_matches(q)
|
181 |
if semantic_matches:
|
|
|
182 |
semantic_matches.sort(key=lambda x: x[1], reverse=True)
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
# Format suggestions
|
187 |
-
indices = [idx for idx, score, match_type in semantic_matches[:top_k]]
|
188 |
-
scores = [score for idx, score, match_type in semantic_matches[:top_k]]
|
189 |
suggestions = "🔍 Semantic matches found:\n\n" + format_suggestions_from_indices(indices, "semantic", scores)
|
|
|
|
|
|
|
190 |
return dialect_bengali, actual, benglish, suggestions
|
191 |
|
192 |
-
# 4) Partial matches in dialect
|
193 |
partial_matches = []
|
194 |
for i, (dialect, dialect_bengali, actual, benglish) in enumerate(phrases_data):
|
195 |
if q_lower in dialect.lower() or dialect.lower() in q_lower:
|
196 |
similarity = difflib.SequenceMatcher(None, q_lower, dialect.lower()).ratio()
|
197 |
partial_matches.append((i, similarity))
|
198 |
-
|
199 |
if partial_matches:
|
200 |
partial_matches.sort(key=lambda x: x[1], reverse=True)
|
201 |
-
best_idx = partial_matches[0][0]
|
202 |
-
d, dialect_bengali, actual, benglish = phrases_data[best_idx]
|
203 |
-
|
204 |
-
# Format suggestions
|
205 |
indices = [idx for idx, score in partial_matches[:top_k]]
|
206 |
scores = [score for idx, score in partial_matches[:top_k]]
|
|
|
|
|
207 |
suggestions = "🔍 Partial matches in dialect:\n\n" + format_suggestions_from_indices(indices, "text", scores)
|
208 |
return dialect_bengali, actual, benglish, suggestions
|
209 |
|
210 |
-
# 5)
|
211 |
close_matches = difflib.get_close_matches(q_lower, dialects_lower, n=top_k, cutoff=0.3)
|
212 |
if close_matches:
|
213 |
-
best_text = close_matches[0]
|
214 |
-
idx = dialects_lower.index(best_text)
|
215 |
-
d, dialect_bengali, actual, benglish = phrases_data[idx]
|
216 |
-
|
217 |
-
text_sim_scores = []
|
218 |
-
for m in close_matches:
|
219 |
-
score = difflib.SequenceMatcher(None, q_lower, m).ratio()
|
220 |
-
text_sim_scores.append(score)
|
221 |
indices = [dialects_lower.index(m) for m in close_matches]
|
|
|
|
|
|
|
222 |
suggestions = "🔍 Similar dialect phrases:\n\n" + format_suggestions_from_indices(indices, "text", text_sim_scores)
|
223 |
return dialect_bengali, actual, benglish, suggestions
|
224 |
|
225 |
-
# 6) Nothing found
|
226 |
-
sample_phrases = [p[0] for p in phrases_data[:
|
227 |
return "", "", "", "❓ NO MATCH FOUND\n\nTry these sample phrases:\n" + "\n".join([f"• {ph}" for ph in sample_phrases])
|
228 |
|
229 |
except Exception as ex:
|
@@ -236,7 +298,7 @@ def show_semantic_analysis(user_text):
|
|
236 |
return ""
|
237 |
patterns, meanings = semantic_analysis(user_text)
|
238 |
if patterns:
|
239 |
-
return f"Detected patterns: {', '.join([f'{p}→{m}' for p, m, t in patterns])}"
|
240 |
return "No specific patterns detected"
|
241 |
|
242 |
# Custom CSS for a softer, less blinding color scheme
|
@@ -261,10 +323,10 @@ body {
|
|
261 |
with gr.Blocks(title="Enhanced Dialect Translator", css=css, theme=gr.themes.Soft()) as demo:
|
262 |
gr.Markdown("# 🌍 Dialect Bengali → Actual Bengali → Benglish")
|
263 |
gr.Markdown("Type a phrase in your dialect. The app uses both text and semantic matching to find similar phrases.")
|
264 |
-
|
265 |
# Define input component first
|
266 |
inp = gr.Textbox(label="Type phrase in Dialect Bengali", placeholder="e.g. Kita kobor? Sob bala asoin ni")
|
267 |
-
|
268 |
with gr.Row():
|
269 |
with gr.Column(scale=1):
|
270 |
gr.Markdown("### Examples to try:")
|
@@ -275,24 +337,24 @@ with gr.Blocks(title="Enhanced Dialect Translator", css=css, theme=gr.themes.Sof
|
|
275 |
)
|
276 |
with gr.Column(scale=2):
|
277 |
btn = gr.Button("Translate / Find", variant="primary")
|
278 |
-
|
279 |
with gr.Row():
|
280 |
out_dialect = gr.Textbox(label="Dialect Bengali (Bengali Script)")
|
281 |
out_actual = gr.Textbox(label="Actual Bengali (Standard)")
|
282 |
out_benglish = gr.Textbox(label="Benglish (Phonetic English)")
|
283 |
-
|
284 |
with gr.Row():
|
285 |
semantic_info = gr.Textbox(label="Semantic Analysis", lines=2)
|
286 |
-
|
287 |
suggestions = gr.Textbox(label="Status / Suggestions / Top Candidates", lines=8)
|
288 |
|
289 |
# Set up event handlers
|
290 |
btn.click(
|
291 |
-
fn=translate_text,
|
292 |
-
inputs=[inp],
|
293 |
outputs=[out_dialect, out_actual, out_benglish, suggestions]
|
294 |
)
|
295 |
-
|
296 |
inp.change(
|
297 |
fn=show_semantic_analysis,
|
298 |
inputs=[inp],
|
|
|
2 |
"""
|
3 |
Enhanced Dialect Bengali Translator with Semantic Search
|
4 |
Uses both text similarity and semantic pattern matching
|
5 |
+
Updated to include new dialect patterns and polite/negative 'des/dis' behavior
|
6 |
"""
|
7 |
|
8 |
import difflib
|
|
|
11 |
from collections import defaultdict
|
12 |
import re
|
13 |
|
14 |
+
# === Phrase data: [Dialect Latin, Dialect Bengali Script, Actual Bengali (Std), Benglish] ===
|
15 |
phrases_data = [
|
16 |
+
# Questions / common
|
17 |
+
["gesle ni", "গেসলে নি", "গিয়েছিলে কি?", "giese chile ki?"],
|
18 |
+
["oislo ni", "ওইস্লো নি", "হয়েছে কি?", "hoyeche ki?"],
|
19 |
+
["oigese ni", "ওইগেসে নি", "হয়ে গেছে কি?", "hoyegese ki?"],
|
20 |
+
["oise", "ওইসে", "হয়েছে", "hoyeche"],
|
21 |
+
["bala oise", "বালা ওইসে", "ভালো হয়েছে", "bhalo hoyeche"],
|
22 |
+
["kub bala oise", "কুব বালা ওইসে", "অনেক ভালো হয়েছে", "onek bhalo hoyeche"],
|
23 |
+
["oise jen", "ওইসে জেন", "হয়েছিল যে", "hoyechilo je"],
|
24 |
["jaite ni", "জাইতে নি", "যাবে কি?", "jabe ki?"],
|
25 |
["or ni", "ওর নি", "হচ্ছে কি?", "hocche ki?"],
|
26 |
["or", "ওর", "হচ্ছে", "hocche"],
|
|
|
28 |
["bala ni", "বালা নি", "ভালো কি?", "bhalo ki?"],
|
29 |
["or je", "ওর যে", "হচ্ছে যে", "hocche je"],
|
30 |
["jaibe ni", "জাইবে নি", "যাবে কি?", "jabe ki?"],
|
31 |
+
["jare ni", "জারে নি", "যাচ্ছো কি?", "jaccho ki?"],
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
["Kita kobor?", "কিতা খবর?", "কি খবর?", "ki khobor?"],
|
33 |
["Kita korde?", "কিতা কোর্দে?", "কি করছে?", "ki korchho?"],
|
34 |
["acha oibo-tik ase", "আচা ওইবো-তিক আসে", "ঠিক আছে", "thik ache"],
|
|
|
38 |
["se hole", "সে হলে", "তাহলে", "tahole"],
|
39 |
["Sob bala asoin ni", "সব বালা আসইন নি", "সব ভালো আছে কি?", "sob bhalo ache ki?"],
|
40 |
["Sob bala ase", "সব বালা আসে", "সব ভালো আছে", "sob bhalo ache"],
|
|
|
41 |
["asoini", "আসইনি", "আছে কি?", "ache ki?"],
|
|
|
42 |
["ase", "আসে", "আছে", "ache"],
|
43 |
+
|
44 |
+
# Future / Present / Past core verbs (ja / de / fawa / ka)
|
45 |
+
["jaimu", "জাইমু", "যাব", "jabo"],
|
46 |
+
["jaibay", "জাইবে", "তুমি যাবে", "tumi jabe (dialect)"],
|
47 |
+
["jaibe", "জাইবে", "তুমি যাবে (friend)", "tumi jabe (friend form)"],
|
48 |
+
["jaibo", "জাইবো", "যাবে", "jabe"],
|
49 |
+
["jaiba", "জাইবা", "তারা যাবে", "tara jabe"],
|
50 |
+
["oibo", "ওইবো", "হবে", "hobe"],
|
51 |
+
["oibo jen", "ওইবো জেন", "হবে যে", "hobe je"],
|
52 |
+
|
53 |
+
["ami jaimu", "আমি জাইমু", "আমি যাব", "ami jabo"],
|
54 |
+
["Ami bazaro jaimu", "আমি বাজারো জাইমু", "আমি বাজারে যাব", "ami bazar e jabo"],
|
55 |
+
["He rit aise", "হে রিত আসে", "সে রাতে এসেছে", "se rate esheche"],
|
56 |
+
|
57 |
+
# Give (de) family
|
58 |
+
["des", "দেস", "দাও (মৃদু)", "des (give, friendly)"],
|
59 |
+
["des na", "দেস না", "দাও (দয়া করে, মৃদু অনুরোধ)", "des na (please give)"],
|
60 |
+
["dis", "দিস", "না দাও / নিষেধ", "dis (don't give)"],
|
61 |
+
["dis na", "দিস না", "দেও না", "dis na (don't give)"],
|
62 |
+
["dilaisi", "দিলাইসি", "দিয়েছি", "diyechi (I gave)"],
|
63 |
+
["dilaise", "দিলাইসে", "দিয়েছে", "diyeche (he gave)"],
|
64 |
+
["dilaisoin", "দিলাইসইন", "দিয়েছেন (সম্মানভাষা)", "diyechen (honorific)"],
|
65 |
+
["dise na", "দিসে না", "দেয়নি", "deni (didn't give)"],
|
66 |
+
["dibo", "দিবো", "দেব", "debo (will give)"],
|
67 |
+
["der amare", "দের আমিরে", "সে আমাকে দেয়", "se amake dey"],
|
68 |
+
["dibo amare", "দিবো আমিরে", "সে আমাকে দেবে", "se amake debe"],
|
69 |
+
|
70 |
+
# Get / receive (fawa) family
|
71 |
+
["faisi", "ফাইসি", "পেয়েছি", "peyechi (I got)"],
|
72 |
+
["faisi na", "ফাইসি না", "পাইনি", "pelam na (didn't get)"],
|
73 |
+
["faisot ni", "ফাইসোট নি", "পেলে কি?", "pele ki?"],
|
74 |
+
["faislo", "ফাইসলো", "পেয়ে গেল/লাভ করল (3sg past)", "pelo (he got)"],
|
75 |
+
["faislam", "ফাইসলাম", "পেয়েছিলাম", "pelam (I got past)"],
|
76 |
+
["faisla", "ফাইসলা", "পেয়েছিল (they)", "pela (they got)"],
|
77 |
+
["faisly", "ফাইসলাই", "তুমি পেয়েছ", "tumi pele (you got)"],
|
78 |
+
["faimu", "ফাইমু", "পাব", "pabo (I will get)"],
|
79 |
+
["faibay", "ফাইবে", "তুমি পাবে (dialect)", "tumi pabe"],
|
80 |
+
["faibe", "ফাইবে", "তুমি পাবে (friend)", "tumi pabe (friend)"],
|
81 |
+
["faibo", "ফাইবো", "সে পাবে", "se pabe"],
|
82 |
+
["faiba", "ফাইবা", "তারা পাবে", "tara pabe"],
|
83 |
+
|
84 |
+
# Eat (ka) family
|
85 |
+
["kaimu", "কাইমু", "খাব", "khaimu (I will eat)"],
|
86 |
+
["kaibay", "কাইবে", "তুমি খাব (dialect)", "tumi khabe"],
|
87 |
+
["kaibe", "কাইবে", "তুমি খাব (friend)", "tumi khabe (friend)"],
|
88 |
+
["kaibo", "কাইবো", "সে খাবে", "se khabe"],
|
89 |
+
["kaiba", "কাইবা", "তারা খাবে", "tara khabe"],
|
90 |
+
|
91 |
+
# Other sample sentences from user's corpus
|
92 |
+
["Ami faisi ekta notun jinish", "আমি ফাইসি একটা নতুন জিনিস", "আমি একটা নতুন জিনিস পেয়েছি", "ami ekta notun jinish peyechi"],
|
93 |
+
["Tumi taka faiso ni", "তুমি টাকা ফাইসো নি", "তুমি টাকা পেয়েছ কি?", "tumi taka peyecho ki?"],
|
94 |
+
["He sobsomoy amare teka dey", "হে সবসময় আমিারে তেকা দেয়", "সে সবসময় আমাকে টাকা দেয়", "se shobshomoy amake taka dey"],
|
95 |
+
["Ami bazaro jaimu", "আমি বাজারো জাইমু", "আমি বাজারে যাব", "ami bazar e jabo"],
|
96 |
+
["Tara bazaro bohut jinish faisoin", "তারা বাজারো বহুত জিনিস ফাইসইন", "তারা বাজারে অনেক জিনিস পেয়েছে", "tara bazar e onek jinish peyechhe"],
|
97 |
+
["Tumi boi diso ni", "তুমি বই দিসো নি", "আপনি কি বই দিয়েছেন?", "apni boi diyechen?"],
|
98 |
+
["Tuin boi disot ni", "তুইন বই দিসট নি", "তুই বই দিয়েছ কি?", "tui boi diyechish?"],
|
99 |
+
["Bifodo asi", "বিফোডো আছি", "বিপদে আছি", "bipode achi"],
|
100 |
+
["Kotobil bade fawa gese", "কোটবিল বাদে ফাওয়া গেসে", "অনেকদিন পরে পেয়েছি", "got after long time got"]
|
101 |
]
|
102 |
|
103 |
+
# Semantic mapping of dialect patterns to meanings + types
|
104 |
semantic_patterns = {
|
105 |
+
# question/particles
|
106 |
+
r"\bni\b": {"meaning": "কি", "type": "question"},
|
107 |
+
r"\bni\b$": {"meaning": "কি", "type": "question"},
|
108 |
+
# verbs / roots
|
109 |
+
r"\bor\b": {"meaning": "হচ্ছে", "type": "verb"},
|
110 |
+
r"\boise\b": {"meaning": "হয়েছে", "type": "verb"},
|
111 |
+
r"\boibo\b": {"meaning": "হবে", "type": "verb"},
|
112 |
+
r"\bjaimu\b": {"meaning": "যাব", "type": "verb"},
|
113 |
+
r"\bjaib[aey]\b": {"meaning": "যাবে", "type": "verb"},
|
114 |
+
r"\bkobor\b": {"meaning": "খবর", "type": "noun"},
|
115 |
+
r"\bkorde\b": {"meaning": "করছে", "type": "verb"},
|
116 |
+
r"\bacha\b": {"meaning": "ঠিক", "type": "adjective"},
|
117 |
+
r"\bbala\b": {"meaning": "ভালো", "type": "adjective"},
|
118 |
+
r"\bkub\b": {"meaning": "অনেক", "type": "adverb"},
|
119 |
+
r"\bgesle\b": {"meaning": "গিয়েছিলে", "type": "verb"},
|
120 |
+
r"\boislo\b": {"meaning": "হয়েছে", "type": "verb"},
|
121 |
+
r"\boigese\b": {"meaning": "হয়েগেছে", "type": "verb"},
|
122 |
+
r"\bjen\b": {"meaning": "যে", "type": "conjunction"},
|
123 |
+
r"\bje\b": {"meaning": "যে", "type": "conjunction"},
|
124 |
+
r"\btik\b": {"meaning": "ঠিক", "type": "adjective"},
|
125 |
+
r"\base\b": {"meaning": "আছে", "type": "verb"},
|
126 |
+
r"\basoin\b": {"meaning": "আছে", "type": "verb"},
|
127 |
+
r"\basoini\b": {"meaning": "আছে কি", "type": "verb+question"},
|
128 |
+
r"\bGoto\b": {"meaning": "গত", "type": "adjective"},
|
129 |
+
r"\bkali\b": {"meaning": "কাল", "type": "noun"},
|
130 |
+
r"\bkita\b": {"meaning": "কি", "type": "question"},
|
131 |
+
r"\btew\b": {"meaning": "তাহলে", "type": "conjunction"},
|
132 |
+
# give/get polarity (important dialect contrast)
|
133 |
+
r"\bdes\b": {"meaning": "দান/দাও (বন্ধু-মৃদু)", "type": "give_positive"},
|
134 |
+
r"\bdes\s+na\b": {"meaning": "মৃদু অনুরোধ: দাও", "type": "give_positive"},
|
135 |
+
r"\bdis\b": {"meaning": "না দাও / নিষেধ", "type": "give_negative"},
|
136 |
+
r"\bdis\s+na\b": {"meaning": "না দাও (নিষেধ)", "type": "give_negative"},
|
137 |
+
# fawa/get variants
|
138 |
+
r"\bfaisi\b": {"meaning": "পেয়েছি", "type": "verb"},
|
139 |
+
r"\bfaisl[ao]m\b": {"meaning": "পেয়েছিলাম/পেয়েছি(past)", "type": "verb"},
|
140 |
+
r"\bfaimu\b": {"meaning": "পাব", "type": "verb"},
|
141 |
+
r"\bfaib[ae]y?\b": {"meaning": "পাবে", "type": "verb"},
|
142 |
+
# future pattern markers
|
143 |
+
r"\bmu\b": {"meaning": "ভবিষ্যৎ: 1sg", "type": "tense_future"},
|
144 |
+
r"\bbay\b": {"meaning": "ভবিষ্যৎ: 2sg (tumi)", "type": "tense_future"},
|
145 |
+
r"\bbo\b": {"meaning": "ভবিষ্যৎ: 3sg", "type": "tense_future"},
|
146 |
+
r"\bba\b": {"meaning": "ভবিষ্যৎ: plural/3pl", "type": "tense_future"},
|
147 |
}
|
148 |
|
149 |
# Precompute data structures for matching
|
|
|
151 |
dialects_lower = [d.lower() for d in dialects]
|
152 |
actual_bengali_list = [p[2] for p in phrases_data]
|
153 |
|
154 |
+
# Create a mapping from dialect to full row
|
155 |
dialect_to_all = {p[0].lower(): p for p in phrases_data}
|
156 |
|
157 |
def semantic_analysis(user_input):
|
|
|
159 |
user_lower = user_input.lower()
|
160 |
detected_patterns = []
|
161 |
meaning_components = []
|
162 |
+
|
163 |
+
# Use regex-based whole-word matching for patterns
|
164 |
for pattern, info in semantic_patterns.items():
|
165 |
+
try:
|
166 |
+
if re.search(pattern, user_lower):
|
167 |
+
detected_patterns.append((pattern, info["meaning"], info["type"]))
|
168 |
+
meaning_components.append(info["meaning"])
|
169 |
+
except re.error:
|
170 |
+
# If pattern is bad, skip it safely
|
171 |
+
continue
|
172 |
+
|
173 |
return detected_patterns, meaning_components
|
174 |
|
175 |
+
def find_semantic_matches(user_input, threshold=0.35):
|
176 |
+
"""Find matches based on semantic similarity + text similarity"""
|
177 |
user_lower = user_input.lower()
|
178 |
matches = []
|
179 |
+
|
180 |
# Get semantic patterns from user input
|
181 |
detected_patterns, meaning_components = semantic_analysis(user_input)
|
182 |
+
|
183 |
# If we found semantic patterns, look for phrases with similar meanings
|
184 |
if meaning_components:
|
185 |
for i, (dialect, dialect_bengali, actual, benglish) in enumerate(phrases_data):
|
186 |
+
match_score = 0.0
|
187 |
+
# boost if any of the meaning_components appear in actual or dialect
|
188 |
for meaning in meaning_components:
|
189 |
if meaning in actual:
|
190 |
+
match_score += 0.35
|
191 |
+
if meaning in dialect.lower():
|
192 |
+
match_score += 0.25
|
193 |
+
|
194 |
+
# text similarity between user and dialect form
|
195 |
text_similarity = difflib.SequenceMatcher(None, user_lower, dialect.lower()).ratio()
|
196 |
+
total_score = match_score + (text_similarity * 0.5)
|
197 |
+
|
198 |
if total_score > threshold:
|
199 |
matches.append((i, total_score, "semantic"))
|
200 |
+
|
201 |
return matches
|
202 |
|
203 |
def format_suggestions_from_indices(indices, match_type="text", scores=None):
|
|
|
205 |
lines = []
|
206 |
for i, idx in enumerate(indices):
|
207 |
d, dialect_bengali, actual, benglish = phrases_data[idx]
|
208 |
+
|
209 |
score_str = ""
|
210 |
if scores is not None and i < len(scores):
|
211 |
s_pct = int(scores[i] * 100)
|
212 |
score_str = f" ({match_type}-match: {s_pct}%)"
|
213 |
+
|
214 |
lines.append(f"• {d}{score_str}\n Dialect Bengali: {dialect_bengali}\n Actual Bengali: {actual}\n Benglish: {benglish}")
|
215 |
return "\n\n".join(lines)
|
216 |
|
217 |
+
def translate_text(user_text, top_k: int = 6):
|
218 |
"""
|
219 |
Returns: (dialect_out, actual_out, benglish_out, suggestions_out)
|
220 |
"""
|
|
|
230 |
if q_lower == dialect.lower():
|
231 |
return dialect_bengali, actual, benglish, "✅ EXACT MATCH (100%)"
|
232 |
|
233 |
+
# 2) If input contains multiple phrases separated by punctuation
|
234 |
+
potential_phrases = [p.strip() for p in re.split(r'[.,;!?]\s*', q) if p.strip()]
|
235 |
if len(potential_phrases) > 1:
|
236 |
results = []
|
237 |
for phrase in potential_phrases:
|
238 |
+
matched = False
|
239 |
+
for d, dialect_bengali, actual, benglish in phrases_data:
|
240 |
+
if phrase.lower() == d.lower():
|
241 |
+
results.append(f"{dialect_bengali} → {actual} → {benglish}")
|
242 |
+
matched = True
|
243 |
+
break
|
244 |
+
if not matched:
|
245 |
+
results.append(f"'{phrase}' → No match found")
|
246 |
+
return "", "", "", "Multiple phrases detected:\n\n" + "\n\n".join(results)
|
|
|
|
|
247 |
|
248 |
# 3) Semantic matches
|
249 |
semantic_matches = find_semantic_matches(q)
|
250 |
if semantic_matches:
|
251 |
+
# sort and return top semantic candidates
|
252 |
semantic_matches.sort(key=lambda x: x[1], reverse=True)
|
253 |
+
indices = [idx for idx, score, mt in semantic_matches[:top_k]]
|
254 |
+
scores = [score for idx, score, mt in semantic_matches[:top_k]]
|
|
|
|
|
|
|
|
|
255 |
suggestions = "🔍 Semantic matches found:\n\n" + format_suggestions_from_indices(indices, "semantic", scores)
|
256 |
+
# Return best match as primary output
|
257 |
+
best_idx = indices[0]
|
258 |
+
d, dialect_bengali, actual, benglish = phrases_data[best_idx]
|
259 |
return dialect_bengali, actual, benglish, suggestions
|
260 |
|
261 |
+
# 4) Partial matches in dialect strings
|
262 |
partial_matches = []
|
263 |
for i, (dialect, dialect_bengali, actual, benglish) in enumerate(phrases_data):
|
264 |
if q_lower in dialect.lower() or dialect.lower() in q_lower:
|
265 |
similarity = difflib.SequenceMatcher(None, q_lower, dialect.lower()).ratio()
|
266 |
partial_matches.append((i, similarity))
|
267 |
+
|
268 |
if partial_matches:
|
269 |
partial_matches.sort(key=lambda x: x[1], reverse=True)
|
|
|
|
|
|
|
|
|
270 |
indices = [idx for idx, score in partial_matches[:top_k]]
|
271 |
scores = [score for idx, score in partial_matches[:top_k]]
|
272 |
+
best_idx = indices[0]
|
273 |
+
d, dialect_bengali, actual, benglish = phrases_data[best_idx]
|
274 |
suggestions = "🔍 Partial matches in dialect:\n\n" + format_suggestions_from_indices(indices, "text", scores)
|
275 |
return dialect_bengali, actual, benglish, suggestions
|
276 |
|
277 |
+
# 5) Close textual matches using difflib
|
278 |
close_matches = difflib.get_close_matches(q_lower, dialects_lower, n=top_k, cutoff=0.3)
|
279 |
if close_matches:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
280 |
indices = [dialects_lower.index(m) for m in close_matches]
|
281 |
+
text_sim_scores = [difflib.SequenceMatcher(None, q_lower, m).ratio() for m in close_matches]
|
282 |
+
best_idx = indices[0]
|
283 |
+
d, dialect_bengali, actual, benglish = phrases_data[best_idx]
|
284 |
suggestions = "🔍 Similar dialect phrases:\n\n" + format_suggestions_from_indices(indices, "text", text_sim_scores)
|
285 |
return dialect_bengali, actual, benglish, suggestions
|
286 |
|
287 |
+
# 6) Nothing found — give sample suggestions
|
288 |
+
sample_phrases = [p[0] for p in phrases_data[:10]]
|
289 |
return "", "", "", "❓ NO MATCH FOUND\n\nTry these sample phrases:\n" + "\n".join([f"• {ph}" for ph in sample_phrases])
|
290 |
|
291 |
except Exception as ex:
|
|
|
298 |
return ""
|
299 |
patterns, meanings = semantic_analysis(user_text)
|
300 |
if patterns:
|
301 |
+
return f"Detected patterns: {', '.join([f'{p} → {m}' for p, m, t in patterns])}"
|
302 |
return "No specific patterns detected"
|
303 |
|
304 |
# Custom CSS for a softer, less blinding color scheme
|
|
|
323 |
with gr.Blocks(title="Enhanced Dialect Translator", css=css, theme=gr.themes.Soft()) as demo:
|
324 |
gr.Markdown("# 🌍 Dialect Bengali → Actual Bengali → Benglish")
|
325 |
gr.Markdown("Type a phrase in your dialect. The app uses both text and semantic matching to find similar phrases.")
|
326 |
+
|
327 |
# Define input component first
|
328 |
inp = gr.Textbox(label="Type phrase in Dialect Bengali", placeholder="e.g. Kita kobor? Sob bala asoin ni")
|
329 |
+
|
330 |
with gr.Row():
|
331 |
with gr.Column(scale=1):
|
332 |
gr.Markdown("### Examples to try:")
|
|
|
337 |
)
|
338 |
with gr.Column(scale=2):
|
339 |
btn = gr.Button("Translate / Find", variant="primary")
|
340 |
+
|
341 |
with gr.Row():
|
342 |
out_dialect = gr.Textbox(label="Dialect Bengali (Bengali Script)")
|
343 |
out_actual = gr.Textbox(label="Actual Bengali (Standard)")
|
344 |
out_benglish = gr.Textbox(label="Benglish (Phonetic English)")
|
345 |
+
|
346 |
with gr.Row():
|
347 |
semantic_info = gr.Textbox(label="Semantic Analysis", lines=2)
|
348 |
+
|
349 |
suggestions = gr.Textbox(label="Status / Suggestions / Top Candidates", lines=8)
|
350 |
|
351 |
# Set up event handlers
|
352 |
btn.click(
|
353 |
+
fn=translate_text,
|
354 |
+
inputs=[inp],
|
355 |
outputs=[out_dialect, out_actual, out_benglish, suggestions]
|
356 |
)
|
357 |
+
|
358 |
inp.change(
|
359 |
fn=show_semantic_analysis,
|
360 |
inputs=[inp],
|