bareethul/nyc-landmark-descriptions

Dataset Card for NYC Landmark Descriptions

This dataset card documents the NYC Landmark Descriptions dataset. It contains 100+ manually written landmark descriptions, each ~200 characters long, labeled with vibe and a binary touristy tag. The augmented split expands to 1,000 samples.

Dataset Details

Dataset Description

• —Curated by: Bareethul Kader (Carnegie Mellon University)
• —Language(s): English
• —License: CC BY 4.0
• —Repository: bareethul/nyc-landmark-descriptions

Uses

Direct Use

• —Educational practice in NLP dataset creation and augmentation.
• —Classification task: predict the vibe of a landmark (Iconic, Hidden Gem, Peaceful, Touristy).
• —Binary task: predict whether the landmark is touristy or not.

Out-of-Scope Use

• —Not intended as a factual tourist guide.
• —Descriptions are partly fictional or paraphrased — not reliable for travel planning.

Dataset Structure

• —Original split: 100 manually authored landmark descriptions.
• —Augmented split: 1,000 text samples generated with Easy Data Augmentation (EDA) techniques (synonym replacement, random swap, insertion, deletion).

Features:

• —Landmark (string, e.g., Statue of Liberty)
• —Description (~200-character text)
• —Vibe (categorical: Iconic, Hidden Gem, Peaceful, Touristy)
• —is_touristy (binary target: 1 = yes, 0 = no)

Dataset Creation

Curation Rationale To study how short text descriptions of landmarks can be used for sentiment-style classification (vibe) and binary tagging (touristy or not).

Data Collection and Processing

• —All descriptions manually written or paraphrased by the dataset creator.
• —Each description is unique and ≥200 characters.
• —Augmentation performed using word-level EDA with NLTK/WordNet synonyms and random perturbations.

Source Data Producers

• —Original author: Bareethul Kader.
• —Data inspired by common NYC landmarks (fictionalized/paraphrased).

Annotations

• —Annotation Process:
• —Vibe assigned manually (Iconic, Hidden Gem, Peaceful, Touristy).
• —is_touristy assigned based on whether a landmark is typically crowded/visited.
• —Annotators: Dataset creator.

Personal and Sensitive Information

• —No personal or sensitive information included.
• —All text is fictional or paraphrased descriptions of public places.

Bias, Risks, and Limitations

• —Dataset is small (100 samples) and handcrafted — not representative of all NYC landmarks.
• —Subjective labels like vibe may vary across annotators.
• —Augmentation may occasionally create awkward or ungrammatical text.

Recommendations

Users should treat this dataset as an educational text classification resource, not as factual information about NYC landmarks. It is best used for NLP experiments, demos, and student projects.

Citation

BibTeX:

@dataset{bareethul_nyc_landmark_descriptions,
  author       = {Kader, Bareethul},
  title        = {NYC Landmark Descriptions},
  year         = {2025},
  publisher    = {Hugging Face Datasets},
  url          = {https://huggingface.co/datasets/bareethul/nyc-landmark-descriptions}
}