Dataset

restaurant-reviews

A dataset card entry for restaurant reviews that flags the need for additional information beyond what is currently provided.

Open on Hugging Face

Data preview

Subset

Loading preview...

Dataset card

The restaurant-reviews dataset, published on Hugging Face by user jannikseus, is a small text corpus of short consumer reviews annotated for token-level sequence labelling. It is distributed as parquet files under an unstated license and has been downloaded only a handful of times.

What this dataset contains

The dataset holds short, informal English review snippets (for example two- to five-word opinions about a restaurant) that have been tokenised and tagged for a sequence-labelling task. Splits include train, validation and test partitions totalling roughly 2,676 rows, placing it in the 1K–10K size band noted in its tags. The unstated license means redistribution terms are not specified in the documentation.

Structure and fields

Each row carries three columns: an id string identifying the review, a labels column holding one ClassLabel per token from the set {B, I, O} (a BIO tagging scheme typically used for named-entity or aspect tagging), and a sentence column containing the list of string tokens. The file format is parquet, and the dataset is compatible with the datasets, pandas, polars and mlcroissant libraries.

How analysts use it

Because the corpus is tiny, unlicensed and oriented toward NLP benchmarking rather than market coverage, it has limited direct value for an investment workflow. It could serve only as a teaching example for analysts exploring how consumer-review text might be processed for downstream sentiment or aspect extraction, and any production use would require a substantially larger and clearly licensed alternative.

Documentation & references 1