|
|
|
|
|
"""app.ipynb |
|
|
|
|
|
Automatically generated by Colab. |
|
|
|
|
|
Original file is located at |
|
|
https://colab.research.google.com/drive/1neCwOaMM4pn-eYYwZ6zYIdhhNdiwIG5K |
|
|
""" |
|
|
|
|
|
import gradio as gr |
|
|
from openai import OpenAI |
|
|
import os |
|
|
|
|
|
|
|
|
|
|
|
client = OpenAI( |
|
|
api_key=os.environ["GROQ_API_KEY"], |
|
|
base_url="https://api.groq.com/openai/v1" |
|
|
) |
|
|
|
|
|
|
|
|
def lcpp_llm(prompt, max_tokens=512, temperature=0.3, stop=None): |
|
|
response = client.chat.completions.create( |
|
|
model="llama3-8b-8192", |
|
|
messages=[{"role": "user", "content": prompt}], |
|
|
max_tokens=max_tokens, |
|
|
temperature=temperature, |
|
|
stop=stop |
|
|
) |
|
|
return response.choices[0].message.content.strip() |
|
|
|
|
|
|
|
|
def pos_function(user_query): |
|
|
POS_SYSTEM_MESSAGE=""" |
|
|
SYSTEM: |
|
|
You are a linguistics expert and NLP model trained to analyze the grammatical structure of English text. |
|
|
Your task is to perform Part-of-Speech (POS) tagging for the following customer review. |
|
|
For each word, return its POS tag in a clean, aligned format. |
|
|
|
|
|
USER: """ |
|
|
prompt=POS_SYSTEM_MESSAGE+user_query |
|
|
output = lcpp_llm(prompt, max_tokens=512) |
|
|
return output |
|
|
|
|
|
def ner_function(user_query): |
|
|
NER_SYSTEM_MESSAGE=""" |
|
|
SYSTEM: |
|
|
You are an Information Extraction Specialist AI that extracts meaningful entities from customer feedback. |
|
|
Identify all named entities in the review and classify them into the appropriate types (e.g., PERSON, ORGANIZATION, PRODUCT, LOCATION, DATE, etc.). |
|
|
Return your output in a structured table with two columns: Entity and Label. Do not ask any follow up questions. |
|
|
|
|
|
USER: |
|
|
""" |
|
|
prompt=NER_SYSTEM_MESSAGE+user_query |
|
|
output = lcpp_llm(prompt, max_tokens=512) |
|
|
return output |
|
|
|
|
|
def analysis_function(user_query): |
|
|
ANALYSIS_SYSTEM_MESSAGE=""" |
|
|
SYSTEM: |
|
|
You are a Customer Experience Analyst AI. Analyze the customer review below and create a concise table with three columns: |
|
|
1. Category (Product or Service) mentioned in comment |
|
|
2. Sentiment (Positive, Negative, or Mixed) |
|
|
3. Insight (What is going well or needs improvement) |
|
|
|
|
|
Only output the table Focus on clear business insights. Do not ask any follow up questions. |
|
|
USER: |
|
|
""" |
|
|
prompt=ANALYSIS_SYSTEM_MESSAGE+user_query |
|
|
output= lcpp_llm(prompt, max_tokens=512) |
|
|
|
|
|
if output.count("| Category | Sentiment | Insight |") > 1: |
|
|
first_table = output.split("| Category | Sentiment | Insight |", 1)[1] |
|
|
first_table = "| Category | Sentiment | Insight |\n" + first_table.split("SYSTEM:")[0].strip() |
|
|
return first_table |
|
|
else: |
|
|
return output |
|
|
|
|
|
|
|
|
|
|
|
def full_analysis(user_query): |
|
|
pos = pos_function(user_query) |
|
|
ner = ner_function(user_query) |
|
|
analysis = analysis_function(user_query) |
|
|
return pos, ner, analysis |
|
|
|
|
|
iface = gr.Interface( |
|
|
fn=full_analysis, |
|
|
inputs=gr.Textbox(lines=5, label="Enter Customer Review"), |
|
|
outputs=[ |
|
|
gr.Textbox(label="POS Tags"), |
|
|
gr.Textbox(label="Named Entities"), |
|
|
gr.Textbox(label="Review Sentiment & Insights") |
|
|
], |
|
|
title="Customer Review Analyzer with Llama 3 (POS + NER + Sentiment)", |
|
|
description="This tool uses Meta-Llama-3-8B-Instruct (GGUF) to extract POS tags, named entities, and sentiment insights from customer reviews." |
|
|
) |
|
|
iface.launch() |
