Dataset
airline-dataset
An apache-2.0 multilingual corpus tailored for fine-tuning models that must answer technical airline-domain queries drawn from aviation source documents.
Data preview
Dataset card
The Davidsv airline-dataset is a small, openly distributed Hugging Face collection of 167 rows pairing image files with textual queries. Its contents, as published, do not clearly correspond to passenger-flow or foot-traffic data despite being filed under that category.
What this dataset contains
The dataset is distributed in Parquet format and is tagged for use with the Hugging Face datasets, pandas, polars, and mlcroissant libraries, with a stated size category of fewer than 1,000 samples. According to the schema, each row holds a file_name, a page number, a query string, and an image, suggesting paired document-page imagery and associated text rather than structured foot-traffic observations. The masked sample rows preview queries written in French about accounting regulations and financial conclusions, which further indicates the underlying files are scanned financial or regulatory documents rather than travel-industry passenger counts.
Structure and fields
The single defined split, train, contains 167 rows. Four columns are exposed: file_name as a string Value, page as an int64 Value, query as a string Value, and image as an Image type hosted via the Hugging Face datasets-server asset endpoint. Because file_name values are masked in the published sample, individual document identifiers cannot be confirmed from the listing.
How analysts use it
Given the schema of page, query, and image fields, the collection is most plausibly suited to training or evaluating document-understanding models on paired page imagery and natural-language questions, not to deriving airport or hub foot-traffic signals. Without further publisher documentation, claims of investment-fund use cases for passenger-flow analysis would not be supported by the available facts.
Documentation & references 1
Related datasets
Browse moreView source information and access options.
Foot-traffic panels built from mobile-device location signals, measuring visits, dwell time and cross-shopping for US retailers, malls and real-estate assets, delivered on a daily schedule to investors and real-estate desks.
Indexes built from foot-traffic location data for Japanese business sites including production facilities and retail outlets.
Foot-traffic-based indices covering locations operated by Japanese real estate investment trust firms.
Location-driven metrics derived from foot-traffic patterns at Japanese real estate investment trust properties.
Location-based indices calculated from foot-traffic signals for Japanese corporate sites such as plants and stores.
Discussion
Questions or observations about this dataset? Join the conversation on X.
No comments yet. Start the conversation on X.