Customer Sentiment Analysis

Customer sentiment analysis task for support interactions.

This module provides a predefined task for analyzing customer sentiment specifically in support contexts, including satisfaction levels and emotional states that affect customer experience and support strategy.

Example

Basic usage with BatchResponses:

from openai import OpenAI
from openaivec.responses import BatchResponses
from openaivec.task import customer_support

client = OpenAI()
analyzer = BatchResponses.of_task(
    client=client,
    model_name="gpt-4o-mini",
    task=customer_support.CUSTOMER_SENTIMENT
)

inquiries = [
    "I'm really disappointed with your service. This is the third time I've had this issue.",
    "Thank you so much for your help! You've been incredibly patient.",
    "I need to cancel my subscription. It's not working for me."
]
sentiments = analyzer.parse(inquiries)

for sentiment in sentiments:
    print(f"Sentiment: {sentiment.sentiment}")
    print(f"Satisfaction: {sentiment.satisfaction_level}")
    print(f"Churn Risk: {sentiment.churn_risk}")
    print(f"Emotional State: {sentiment.emotional_state}")

With pandas integration:

import pandas as pd
from openaivec import pandas_ext  # Required for .ai accessor
from openaivec.task import customer_support

df = pd.DataFrame({"inquiry": [
    "I'm really disappointed with your service. This is the third time I've had this issue.",
    "Thank you so much for your help! You've been incredibly patient.",
    "I need to cancel my subscription. It's not working for me."
]})
df["sentiment"] = df["inquiry"].ai.task(customer_support.CUSTOMER_SENTIMENT)

# Extract sentiment components
extracted_df = df.ai.extract("sentiment")
print(extracted_df[["inquiry", "sentiment_satisfaction_level", "sentiment_churn_risk", "sentiment_emotional_state"]])

Attributes:

Name	Type	Description
CUSTOMER_SENTIMENT	PreparedTask	A prepared task instance configured for customer sentiment analysis with temperature=0.0 and top_p=1.0 for deterministic output.

customer_sentiment(business_context='general customer support', temperature=0.0, top_p=1.0)

Create a configurable customer sentiment analysis task.

Parameters:

Name	Type	Description	Default
business_context	str	Business context for sentiment analysis.	'general customer support'
temperature	float	Sampling temperature (0.0-1.0).	0.0
top_p	float	Nucleus sampling parameter (0.0-1.0).	1.0

Returns:

Type	Description
PreparedTask	PreparedTask configured for customer sentiment analysis.

Source code in src/openaivec/task/customer_support/customer_sentiment.py

def customer_sentiment(
    business_context: str = "general customer support",
    temperature: float = 0.0,
    top_p: float = 1.0
) -> PreparedTask:
    """Create a configurable customer sentiment analysis task.

    Args:
        business_context: Business context for sentiment analysis.
        temperature: Sampling temperature (0.0-1.0).
        top_p: Nucleus sampling parameter (0.0-1.0).

    Returns:
        PreparedTask configured for customer sentiment analysis.
    """

    instructions = f"""Analyze customer sentiment in the context of support interactions, focusing on satisfaction, emotional state, and business implications.

Business Context: {business_context}

Sentiment Categories:
- positive: Customer is happy, satisfied, or grateful
- negative: Customer is unhappy, frustrated, or disappointed
- neutral: Customer is matter-of-fact, without strong emotions
- mixed: Customer expresses both positive and negative sentiments

Satisfaction Levels:
- very_satisfied: Extremely happy, praising service, expressing gratitude
- satisfied: Content, appreciative, positive feedback
- neutral: Neither satisfied nor dissatisfied, factual communication
- dissatisfied: Unhappy, expressing concerns, mild complaints
- very_dissatisfied: Extremely unhappy, angry, threatening to leave

Emotional States:
- happy: Cheerful, pleased, content
- frustrated: Annoyed, impatient, struggling with issues
- angry: Hostile, aggressive, demanding immediate action
- disappointed: Let down, expectations not met
- confused: Lost, needing clarification, overwhelmed
- grateful: Thankful, appreciative of help received
- worried: Anxious, concerned about outcomes

Churn Risk Assessment:
- low: Happy customers, positive experience
- medium: Neutral customers, some concerns but manageable
- high: Dissatisfied customers, multiple issues, expressing frustration
- critical: Extremely unhappy, threatening to cancel, demanding escalation

Relationship Status:
- new: First-time contact, tentative, learning
- loyal: Long-term customer, familiar with service
- at_risk: Showing signs of dissatisfaction, needs attention
- detractor: Actively unhappy, may spread negative feedback
- advocate: Extremely satisfied, promotes service to others

Response Approach:
- empathetic: Use compassionate language, acknowledge feelings
- professional: Maintain formal, solution-oriented communication
- solution_focused: Directly address problems, provide clear next steps
- escalation_required: Immediately involve management or specialists

Analyze tone indicators like:
- Positive: "thank you", "great", "helpful", "love", "excellent"
- Negative: "terrible", "disappointed", "frustrated", "awful", "horrible"
- Urgency: "urgent", "immediately", "ASAP", "critical"
- Threat: "cancel", "switch", "competitor", "lawyer", "report"

IMPORTANT: Provide analysis responses in the same language as the input text, except for the predefined categorical fields (sentiment, satisfaction_level, emotional_state, churn_risk, relationship_status, response_approach) which must use the exact English values specified above. For example, if the input is in Spanish, provide tone_indicators in Spanish, but use English values like "positive" for sentiment.

Provide comprehensive sentiment analysis with business context and recommended response strategy."""

    return PreparedTask(
        instructions=instructions,
        response_format=CustomerSentiment,
        temperature=temperature,
        top_p=top_p
    )