Initial Commit

InferenceServer.py

@@ -0,0 +1,48 @@
+import os
+
+print("Installing dependencies...")
+
+os.system("git clone https://github.com/PrithivirajDamodaran/neuspell.git")
+os.chdir('neuspell')
+os.system('pip install -e .[elmo]')
+os.system('pip install -e .[spacy]')
+os.system("python -m spacy download en_core_web_sm")
+
+import neuspell
+from neuspell import BertsclstmChecker, ElmosclstmChecker, CnnlstmChecker
+bl_checker = BertsclstmChecker()
+el_checker = ElmosclstmChecker()
+cl_checker = CnnlstmChecker()
+
+print("Loading Models...")
+bl_checker.from_pretrained()
+el_checker.from_pretrained()
+cl_checker.from_pretrained()
+print("Dummy run", bl_checker.correct("I luk foward to receving your reply"))
+print("Dummy run", el_checker.correct("I luk foward to receving your reply"))
+print("Dummy run", cl_checker.correct("I luk foward to receving your reply"))
+
+import uvicorn
+from fastapi import File
+from fastapi import FastAPI
+import sys
+
+
+app = FastAPI()
+print("Models loaded !")
+
+
+@app.get("/")
+def read_root():
+    return {"Neuspell !"}
+
+@app.get("/{correct}")
+def get_correction(input_sentence, model):
+    if model == "BERT-LSTM":
+        return {"corrected_sentence": bl_checker.correct(input_sentence)}
+    elif model == "ELMo-LSTM":  
+        return {"corrected_sentence": el_checker.correct(input_sentence)}
+    else:
+        return {"corrected_sentence": cl_checker.correct(input_sentence)}
+
+

app.py

@@ -0,0 +1,140 @@
+import streamlit as st
+from annotated_text import annotated_text
+from multiprocessing import Process
+import math
+import re
+import json
+import difflib
+import requests
+import time
+import os
+
+def start_server():   
+    os.system("uvicorn InferenceServer:app --port 8080 --host 0.0.0.0 --workers 1")
+
+def load_models():
+    if not is_port_in_use(8080):
+        with st.spinner(text="Loading models, please wait..."):
+            proc = Process(target=start_server, args=(), daemon=True)
+            proc.start()
+            while not is_port_in_use(8080):
+                time.sleep(1)
+            st.success("Model server started.")
+    else:
+        st.success("Model server already running...")
+    st.session_state['models_loaded'] = True
+
+def is_port_in_use(port):
+    import socket
+    with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
+        return s.connect_ex(('0.0.0.0', port)) == 0
+
+if 'models_loaded' not in st.session_state:
+    st.session_state['models_loaded'] = False
+
+
+
+def get_correction(input_text, model):
+    correct_request = "http://0.0.0.0:8080/correct?input_sentence="+input_text + "&model=" + model
+    correct_response = requests.get(correct_request)
+    correct_json = json.loads(correct_response.text)
+    corrected_sentence = correct_json["corrected_sentence"]
+    diff = diff_strings(input_text, corrected_sentence)
+    st.markdown(f'##### Corrected text:')
+    st.write('')
+    annotated_text(*diff)
+
+
+def diff_strings(a, b):
+    # take from team-writing-assistant/grammar-correction
+    result = []
+    diff = difflib.Differ().compare(a.split(), b.split())
+    replacement = ""
+    for line in diff:
+        if line.startswith("  "):
+            if len(replacement) == 0:
+                result.append(" ")
+                result.append(line[2:])
+            else:
+                result.append(" ")
+                result.append(("", replacement, "#39ff14"))
+                replacement = ""
+                result.append(line[2:])
+        if line.startswith("- "):
+            if len(replacement) == 0:
+                replacement = line[2:]
+            else:
+                result.append(" ")
+                result.append(("", replacement, "#39ff14"))
+                replacement = ""
+        elif line.startswith("+ "):
+            if len(replacement) == 0:
+                result.append((line[2:], "", "#39ff14"))
+            else:
+                result.append(" ")
+                result.append((line[2:], replacement, "#39ff14"))
+                replacement = ""
+    return result    
+        
+if __name__ == "__main__":
+    
+        st.title('Neuspell - A Python library')
+        st.subheader('For fast & accurate spell correction')
+        st.markdown("Added to HuggingFace hub and spaces with 💙  by Prithivi Da", unsafe_allow_html=True)
+        st.markdown("<p style='color:blue; display:inline'> Integrate with your app with just few lines of code </p>", unsafe_allow_html=True)
+        st.markdown("""
+                    ```python 
+                    import neuspell
+                    from neuspell import BertsclstmChecker
+                    checker = BertsclstmChecker()
+                    checker.from_pretrained()
+
+                    checker.correct("Oh! I loovee it when it starts to rain, its smells awesomee")
+                    #  (or)
+                    checker.correct_strings(["sent1", "sent2"])
+                    #  (or)
+                    checker.correct_from_file(src="typo_ridden_texts.txt")
+                    ```    
+                    """)
+
+        models = [
+                    "BERT-LSTM",
+                    "ELMo-LSTM",
+                    "CNN-LSTM"
+                    ]
+                    
+
+        examples = [
+                    "what is the reazon for everyone to laeve the comapny?",
+                    "I loovee it when it starts to rain, its smells awesomee",
+                    "Feel free to rech out to me",
+                    "Life is shart, so live freely",
+                    "We know the boy actually stol the books",
+                    " We all ate the foood and then made desert",
+                    ]
+
+        if not st.session_state['models_loaded']:
+            load_models()                     
+
+        st.markdown(f'##### Models:')
+        selected_model = st.selectbox(
+        label="Choose a model: (Test and integrate the best that suits your needs)",
+        options=models
+        )
+
+        st.markdown(f'##### Try it now:')
+        input_text = st.selectbox(
+        label="Choose an example",
+        options=examples
+        )
+        st.write("(or)")
+        input_text = st.text_input(
+            label="Bring your own sentence",
+            value=input_text
+        )
+
+        if input_text.strip(): 
+            get_correction(input_text, selected_model)
+
+
+

requirements.txt

@@ -0,0 +1,4 @@
+st-annotated-text
+fastapi
+uvicorn
+huggingface_hub