2.11.0

README copy.md

@@ -0,0 +1,12 @@
+---
+title: Grammar Corrector
+emoji: 🏢
+colorFrom: pink
+colorTo: blue
+sdk: gradio
+sdk_version: 3.39.0
+app_file: app.py
+pinned: false
+---
+
+Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

app.py

@@ -0,0 +1,135 @@
+import gradio as gr
+import requests
+import os
+import subprocess
+subprocess.run("python3 -m spacy download en".split(" "))
+data = {"sentences":["I am an good boy.", "I wanted to going to supermarket."]}
+llama2_url = os.environ['url']
+
+# UJ
+import json
+import errant
+from tqdm import tqdm
+import random
+from difflib import Differ
+
+annotator = errant.load('en')
+
+
+error_categories = [
+    'ADJ',
+    'ADJ:FORM',
+    'ADV',
+    'CONJ',
+    'CONTR',
+    'DET',
+    'MORPH',
+    'NOUN',
+    'NOUN:INFL',
+    'NOUN:NUM',
+    'NOUN:POSS',
+    'ORTH',
+    'OTHER',
+    'PART',
+    'PREP',
+    'PRON',
+    'PUNCT',
+    'SPELL',
+    'UNK',
+    'VERB',
+    'VERB:FORM',
+    'VERB:INFL',
+    'VERB:SVA',
+    'VERB:TENSE',
+    'WO',
+]
+eng2zh = {
+    'M': '漏掉',
+    'R': '換成',
+    'U': '多餘的',
+    'ADJ': '形容詞',
+    'ADJ:FORM': '形容詞形（比較級或最高級）',
+    'ADV': '副詞',
+    'CONJ': '連接詞',
+    'CONTR': '縮寫',
+    'DET': ' 限定詞（冠詞、指示詞、所有格）',
+    'MORPH': '語尾變化（詞性、單複數、拼字）',
+    'NOUN': '名詞',
+    'NOUN:INFL': '名詞語尾變化（可不可數、單複數、拼字）',
+    'NOUN:NUM': '名詞單複數',
+    'NOUN:POSS': '名詞所有格',
+    'ORTH': '大小寫',
+    'OTHER': '換其它的用法',
+    'PART': '介副詞',
+    'PREP': '介詞',
+    'PRON': '代名詞',
+    'PUNCT': '標點',
+    'SPELL': '拼字',
+    'UNK': '難以歸類',
+    'VERB': '動詞',
+    'VERB:FORM': '動詞形',
+    'VERB:INFL': '動詞詞語尾變化',
+    'VERB:SVA': '主詞動詞一致',
+    'VERB:TENSE': '動詞時態',
+    'WO': '詞序',
+}
+color_map = {}
+for pre in ['M', 'R', 'U']:
+    for err in error_categories:
+        color_map[f'{pre}:{err}'] = {'M': 'red', 'R': 'blue', 'U': 'green'}[pre]
+def comp(s1, s2):
+    global annotator
+    orig = annotator.parse(s1, tokenise=True)
+    cor = annotator.parse(s2, tokenise=True)
+    edits = annotator.annotate(orig, cor, merging="all-equal")
+    ori_anno = {
+        'text': s1,
+        'entities': []
+    }
+    cor_anno = {
+        'text': s2,
+        'entities': []
+    }
+    for e in edits:
+        typ, content = e.type[0], e.type[2:]
+        print(e.type, typ, content)
+        if typ in eng2zh and content in eng2zh:
+            new_statement = eng2zh[typ]+':' + eng2zh[content]
+        ori_anno['entities'].append({
+            'entity': new_statement,
+            'start': orig[e.o_start:e.o_end].start_char,
+            'end': orig[e.o_start:e.o_end].end_char
+        })
+        cor_anno['entities'].append({
+            'entity': new_statement,
+            'start': cor[e.c_start:e.c_end].start_char,
+            'end': cor[e.c_start:e.c_end].end_char
+        })
+
+    return ori_anno, cor_anno
+def llama2_all(text):
+    corr = llama2_cor(text)
+    ori_anno, corr_anno = comp(text.strip(), corr.strip())
+    return corr, ori_anno, corr_anno
+def llama2_cor(text):
+    data = {"sentences": text}
+    r = requests.post(f"{llama2_url}/llama2", json=data)
+    try:
+        json_res = r.json()
+        return json_res['sentences']
+    except: return "Please retry or reboot the LLM server."
+with gr.Blocks() as demo:
+    with gr.Tab("Llama-2-13b-chat"):
+        with gr.Row():
+            text_input = gr.Textbox(lines=5, label="Input", placeholder="Please enter sentences line by line.")
+            text_output = gr.Textbox(lines=5, label="Output")
+        with gr.Row():
+            text_output01 = gr.HighlightedText(label="Original Text", combine_adjacent=True, line=10).style(color_map=color_map)
+            text_output02 = gr.HighlightedText(label="Corrected Text", combine_adjacent=True, line=10).style(color_map=color_map)
+        with gr.Row():
+            text_button = gr.Button("Correct and Compare")
+            text_button1 = gr.Button("Correct")
+    text_button.click(llama2_all, inputs=text_input, outputs=[text_output, text_output01, text_output02])
+    text_button1.click(llama2_cor, inputs=text_input, outputs=text_output)
+
+demo.launch(enable_queue=True)

requirements.txt

@@ -0,0 +1 @@
+errant