In this project, we will develop a conversational AI assistant (chatbot) to handle customer inquiries and provide support for a hypothetical e-commerce company. The chatbot will be capable of understanding natural language queries, retrieving relevant information, and generating appropriate responses.
Before starting this project, ensure that you have the following prerequisites installed:
- Python 3.7 or higher
- PyTorch
- Transformers (Hugging Face library)
- RASA (Open-source conversational AI framework)
- spaCy (Natural Language Processing library)
- SQLite (or any other database engine)
First, we need to load the Stanford Conversational AI dataset, which contains conversations between customers and agents in various domains like banking, travel, and shopping.
import pandas as pd
# Load the dataset
data = pd.read_json('https://raw.githubusercontent.com/gunthercox/chatterbot-corpus/master/data/stanford_convertional_ai/stanford_convertional_ai.json')Next, we'll preprocess the data by separating the conversations into user utterances and agent responses.
from sklearn.model_selection import train_test_split
# Separate user utterances and agent responses
utterances = []
responses = []
for conversation in data['conversations']:
for utterance, response in zip(conversation['utterances'], conversation['responses']):
utterances.append(utterance)
responses.append(response)
# Split the data into train and test sets
X_train, X_test, y_train, y_test = train_test_split(utterances, responses, test_size=0.2, random_state=42)We'll use the RASA framework to build the Natural Language Understanding (NLU) model for intent classification and entity extraction.
from rasa.nlu.training_data import load_data
from rasa.nlu.config import RasaNLUModelConfig
from rasa.nlu.model import Trainer
# Create training data in RASA format
training_data = load_data("nlu_data.yml")
# Define the NLU pipeline
nlu_config = RasaNLUModelConfig({"pipeline": [
{"name": "SpacyNLP"},
{"name": "SpacyTokenizer"},
{"name": "SpacyFeaturizer"},
{"name": "RegexFeaturizer"},
{"name": "CRFEntityExtractor"},
{"name": "EntitySynonymMapper"},
{"name": "SklearnIntentClassifier"}
]})
# Train the NLU model
trainer = Trainer(nlu_config)
interpreter = trainer.train(training_data)Next, we'll define the dialogue management logic using RASA's domain file and story files.
from rasa.core.agent import Agent
from rasa.core.train import train_dialogue_model
# Load the domain and stories
domain = "domain.yml"
stories = "data/stories.md"
# Train the dialogue management model
agent = train_dialogue_model(domain, stories)We'll create an action server to handle custom actions and integrate with external systems (e.g., databases, APIs).
from rasa_sdk import Action
import sqlite3
class ActionCheckOrderStatus(Action):
def name(self):
return "action_check_order_status"
def run(self, dispatcher, tracker, domain):
# Get the order ID from the user's message
order_id = tracker.get_slot("order_id")
# Connect to the database and retrieve order status
conn = sqlite3.connect("orders.db")
c = conn.cursor()
c.execute("SELECT status FROM orders WHERE id=?", (order_id,))
status = c.fetchone()[0]
conn.close()
# Send the order status to the user
response = f"The status of your order {order_id} is: {status}"
dispatcher.utter_message(response)
return []Finally, we can run the chatbot by combining the NLU model, dialogue management model, and action server.
from rasa.core.agent import Agent
from rasa.core.interpreter import RasaNLUInterpreter
# Load the trained models
interpreter = RasaNLUInterpreter("models/nlu")
agent = Agent.load("models/dialogue", interpreter=interpreter)
# Run the chatbot
while True:
user_input = input("User: ")
responses = agent.handle_text(user_input)
for response in responses:
print("Chatbot:", response.get("text"))This example demonstrates how to build a conversational AI chatbot using the Stanford Conversational AI dataset, RASA framework, and Python. It covers the key steps of loading and preprocessing the dataset, training the NLU and dialogue management models, building custom actions, and running the chatbot.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent