---
language: en  # The natural language of the dataset (e.g., en, fr, de, es, etc.)
license: apache-2.0 # The license of the dataset (e.g., apache-2.0, cc-by-sa-4.0, mit)
size_categories:
- 10K<n<100K # The size category of the dataset (e.g., n<1K, 1K<n<10K, 10K<n<100K)
source_datasets: ['original'] # The source of the dataset, can also be a Hugging Face dataset (e.g., original, extended, কোথ-2.1)
task_categories:
- text-classification # The task category of the dataset (e.g., text-classification, question-answering, summarization)
pretty_name: Airline Customer Service Dataset # A human-readable name for your dataset
tags:
- airline
- customer service
- text classification
- chatbot
- nlp
---

# Airline Customer Service Dataset

## Dataset Description

This dataset contains customer service interactions related to the airline industry. It can be used for training and evaluating natural language processing (NLP) models for tasks such as intent classification, sentiment analysis, and topic modeling.

**Homepage:** [link to your project's homepage or GitHub repository (if applicable)]

**Dataset Structure:**

The dataset is split into two subsets:

*   **train:** Used for training your models.
*   **test:** Used for evaluating your models.

**Data Fields:**

| Field Name        | Description                                                                         | Data Type             |
|-------------------|-------------------------------------------------------------------------------------|-----------------------|
| Utterance         | The text of the customer's utterance or message.                                   | string                |
| Predicted_Intent  | The predicted intent of the utterance (from a baseline model or previous labeling). | string (categorical) |
| Intent_Score      | The confidence score associated with the predicted intent.                        | float                 |
| Sentiment         | The sentiment expressed in the utterance (e.g., Positive, Negative).               | string (categorical) |
| RSICs             | The Reason for Seeking Information or Support (e.g., Low, Medium, High).            | string (categorical) |
| empathy_score    | A numerical score (1-5) representing the empathy level.                             | int                   |
| listening_score   | A numerical score (1-5) representing the listening/personalization level.         | int                   |
| fallback_type    | The type of fallback that occurred, if any.                                        | string (categorical) |
| Topic             | A numerical code representing the topic of the conversation.                      | int                   |
| Topic_Name        | A descriptive name for the topic (e.g., "General Inquiry", "Baggage Claim").        | string                |

**Data Splits:**

| Split | Number of Rows |
|-------|----------------|
| train | 12478          |
| test  | 3120           |

**Label Mapping (for categorical columns):**

You can find the mapping of labels to numerical IDs in the `label_maps` dictionary within the `data_loader.py` file. For example: