Initial: initial commit

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+cleaned_stories.txt filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,27 @@
+import gradio as gr
+from ngram_text_gen import (
+    generate_output as ngram_text_gen,
+    NGRAM_TEXT_GEN_EXAMPLES,
+)
+
+ngram_generate = gr.Interface(
+    fn=ngram_text_gen,
+    inputs=[
+        gr.Slider(minimum=2, maximum=4, value=2, step=1, label="N-gram"),
+        gr.Textbox(label="Input text"),
+        gr.Number(label="Token length", value=10),
+    ],
+    outputs=[
+        gr.Textbox(label="Generated text"),
+        gr.HTML(label="Other possible outputs"),
+    ],
+    examples=NGRAM_TEXT_GEN_EXAMPLES,
+    title="N-gram Text Generator",
+    description="Generate text using N-grams.",
+    allow_flagging="never",
+)
+
+with gr.Blocks() as demo:
+    ngram_generate.render()
+
+demo.launch()

cleaned_stories.txt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12c40b2fce072f7664e5fbe6b4ab1908cf49bfdad1796814a5176799bfb682fc
+size 158068184

ngram_text_gen.py

@@ -0,0 +1,84 @@
+import random
+
+
+def make_markov_model(cleaned_stories, n_gram=2):
+    markov_model = {}
+
+    # Create n-grams and count transitions
+    for i in range(len(cleaned_stories) - n_gram):
+        curr_state = tuple(cleaned_stories[i : i + n_gram])
+        next_state = cleaned_stories[i + n_gram]
+
+        if curr_state not in markov_model:
+            markov_model[curr_state] = {}
+
+        if next_state in markov_model[curr_state]:
+            markov_model[curr_state][next_state] += 1
+        else:
+            markov_model[curr_state][next_state] = 1
+
+    # Calculate transition probabilities
+    for curr_state, transitions in markov_model.items():
+        total = sum(transitions.values())
+        for next_state in transitions:
+            markov_model[curr_state][next_state] /= total
+
+    return markov_model
+
+
+def generate_text(markov_model, n_gram=2, max_length=100, start=None):
+    if start is None:
+        current_state = random.choice(list(markov_model.keys()))
+    else:
+        current_state = tuple(start.split())
+
+    generated_text = list(current_state)
+
+    while len(generated_text) < max_length:
+        if current_state in markov_model:
+            next_state = random.choices(
+                list(markov_model[current_state].keys()),
+                list(markov_model[current_state].values()),
+            )[0]
+            generated_text.append(next_state)
+            current_state = tuple(generated_text[-n_gram:])
+        else:
+            break
+
+    return " ".join(generated_text)
+
+
+# Function to generate a random story with a specified starting seed and limit
+def generate_random_story(markov_model, start=None, n_gram=2, limit=100):
+    return generate_text(markov_model, n_gram, max_length=limit, start=start)
+
+
+# Load the cleaned stories
+with open("cleaned_stories.txt", "r", encoding="utf-8") as file:
+    cleaned_stories = [line.strip() for line in file]
+
+
+def generate_output(n_gram, input, token_length):
+    markov_model = make_markov_model(cleaned_stories, n_gram)
+
+    text_outputs = []
+    for _ in range(5):
+        text = generate_random_story(markov_model, input, n_gram, token_length)
+
+        generated = " ".join(text.split())
+
+        text_outputs.append(generated)
+
+        other_outputs = "".join(text + "<br />" for text in text_outputs[1:])
+
+    return text_outputs[0], other_outputs
+
+
+NGRAM_TEXT_GEN_EXAMPLES = [
+    [2, "မႂ် သုင်", 10],
+    [2, "မႂ် သုင်", 50],
+    [3, "မႂ် သုင် ၶႃႈ", 10],
+    [2, "ၵိၼ် ၶဝ်ႈ", 10],
+    [3, "လိၵ်ႈ လၢႆး တႆး", 10],
+    [4, "ပွႆး ပီ မႂ်ႇ တႆး", 20],
+]

requirements.txt

@@ -0,0 +1 @@
+gradio