Spaces:
Runtime error
Runtime error
File size: 2,634 Bytes
8e079ed 7d5172b 71c8fbc 7d5172b 71c8fbc 7d5172b 8e079ed 1b948dc 8e079ed 7d5172b 8e079ed | 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 | import gradio as gr
import stanza
import torch
import numpy
# Allowlist numpy.ndarray for weights-only loading
torch.serialization.add_safe_globals([numpy.ndarray])
# Allowlist the required global for weights-only loading
# torch.serialization.add_safe_globals([numpy.core.multiarray._reconstruct])
# Download and initialize the Stanza pipeline
stanza.download('en')
nlp = stanza.Pipeline(lang='en', processors='tokenize,pos,constituency')
def generate_reordering_rule(english, reordered):
tree = nlp(english).sentences[0].constituency
reordered_tokens = reordered.split()
rules = []
def extract(node):
if not hasattr(node, 'children') or all(isinstance(c, str) for c in node.children):
return
child_labels = tuple(c.label if hasattr(c, 'label') else c for c in node.children)
rule_key = (node.label, child_labels)
# Get leaf tokens for each child
child_tokens = []
for c in node.children:
if hasattr(c, 'leaf_labels'):
child_tokens.append(' '.join(c.leaf_labels()))
else:
child_tokens.append(c)
# Try to find token positions in reordered sentence
positions = []
for tok in child_tokens:
first_word = tok.split()[0]
try:
pos = reordered_tokens.index(first_word)
except ValueError:
pos = -1
positions.append(pos)
# Infer reordering function
reordered_indices = sorted(range(len(child_tokens)), key=lambda i: positions[i])
rule_func = f"lambda {', '.join(f'c{i}' for i in range(len(child_tokens)))}: " \
f"{' + '.join(f'c{i}' for i in reordered_indices)}"
rules.append((rule_key, rule_func))
for c in node.children:
if hasattr(c, 'children'):
extract(c)
extract(tree)
# Format rules for display
if not rules:
return "No reordering rules could be inferred."
return "\n".join([f"{key}: {func}" for key, func in rules])
# Gradio interface
demo = gr.Interface(
fn=generate_reordering_rule,
inputs=[
gr.Textbox(lines=2, label="English Sentence", placeholder="e.g. I want to eat the cake."),
gr.Textbox(lines=2, label="Reordered Sentence", placeholder="e.g. I cake the eat to want.")
],
outputs="text",
title="Reordering Rule Generator",
description="Enter an English sentence and its reordered version. This app extracts syntactic transformation rules using Stanza's constituency parser."
)
if __name__ == "__main__":
demo.launch()
|