This dataset compiles the top 2500 datasets from Kaggle, encompassing a diverse range of topics and contributors. It provides insights into dataset creation, usability, popularity, and more, offering valuable information for researchers, analysts, and data enthusiasts.

Research Analysis: Researchers can utilize this dataset to analyze trends in dataset creation, popularity, and usability scores across various categories.

Contributor Insights: Kaggle contributors can explore the dataset to gain insights into factors influencing the success and engagement of their datasets, aiding in optimizing future submissions.

Machine Learning Training: Data scientists and machine learning enthusiasts can use this dataset to train models for predicting dataset popularity or usability based on features such as creator, category, and file types.

Market Analysis: Analysts can leverage the dataset to conduct market analysis, identifying emerging trends and popular topics within the data science community on Kaggle.

Educational Purposes: Educators and students can use this dataset to teach and learn about data analysis, visualization, and interpretation within the context of real-world datasets and community-driven platforms like Kaggle.

Column Definitions:

Dataset Name: Name of the dataset. Created By: Creator(s) of the dataset. Last Updated in number of days: Time elapsed since last update. Usability Score: Score indicating the ease of use. Number of File: Quantity of files included. Type of file: Format of files (e.g., CSV, JSON). Size: Size of the dataset. Total Votes: Number of votes received. Category: Categorization of the dataset's subject matter.

From wiki

Kaggle, a subsidiary of Google LLC, is an online community of data scientists and machine learning practitioners. Kaggle allows users to find and publish data sets, explore and build models in a web-based data-science environment, work with other data scientists and machine learning engineers, and enter competitions to solve data science challenges.

Kaggle got its start in 2010 by offering machine learning competitions and now also offers a public data platform, a cloud-based workbench for data science, and Artificial Intelligence education. Its key personnel were Anthony Goldbloom and Jeremy Howard. Nicholas Gruen was founding chair succeeded by Max Levchin. Equity was raised in 2011 valuing the company at $25 million. On 8 March 2017, Google announced that they were acquiring Kaggle.[1][2]

Source: Kaggle

Context

Kaggle is one of the largest communities of data scientists and machine learning practitioners in the world, and its platform hosts thousands of datasets covering a wide range of topics and industries. With so many options to choose from, it can be difficult to know where to start or what datasets are worth exploring. That's where this dataset comes in. By scraping information about the top 10,000 datasets on Kaggle, we have created a single source of truth for the most popular and useful datasets on the platform. This dataset is not just a list of names and numbers, but a valuable tool for data enthusiasts and professionals alike, providing insights into the latest trends and techniques in data science and machine learning

Column description - Dataset_name - Name of the dataset - Author_name - Name of the author - Author_id - Kaggle id of the author - No_of_files - Number of files the author has uploaded - size - Size of all the files - Type_of_file - Type of the files such as csv, json etc. - Upvotes - Total upvotes of the dataset - Medals - Medal of the dataset - Usability - Usability of the dataset - Date - Date in which the dataset is uploaded - Day - Day in which the dataset is uploaded - Time - Time in which the dataset is uploaded - Dataset_link - Kaggle link of the dataset

Acknowledgements The data has been scraped from the official Kaggle Website and is available under the Creative Common License.

Enjoy & Keep Learning !!!

These data contain the results of GC-MS, LC-MS and immunochemistry analyses of mask sample extracts. The data include tentatively identified compounds through library searches and compound abundance. This dataset is not publicly accessible because: EPA cannot release personally identifiable information regarding living individuals, according to the Privacy Act and the Freedom of Information Act (FOIA). This dataset contains information about human research subjects. Because there is potential to identify individual participants and disclose personal information, either alone or in combination with other datasets, individual level data are not appropriate to post for public access. Restricted access may be granted to authorized persons by contacting the party listed. It can be accessed through the following means: The data can not be accessed. Format: The dataset contains the identification of compounds found in the mask samples as well as the abundance of those compounds for individuals who participated in the trial. This dataset is associated with the following publication: Pleil, J., M. Wallace, J. McCord, M. Madden, J. Sobus, and G. Ferguson. How do cancer-sniffing dogs sort biological samples? Exploring case-control samples with non-targeted LC-Orbitrap, GC-MS, and immunochemistry methods. Journal of Breath Research. Institute of Physics Publishing, Bristol, UK, 14(1): 016006, (2019).

This dataset contains data collected during a study ("Towards High-Value Datasets determination for data-driven development: a systematic literature review") conducted by Anastasija Nikiforova (University of Tartu), Nina Rizun, Magdalena Ciesielska (Gdańsk University of Technology), Charalampos Alexopoulos (University of the Aegean) and Andrea Miletič (University of Zagreb) It being made public both to act as supplementary data for "Towards High-Value Datasets determination for data-driven development: a systematic literature review" paper (pre-print is available in Open Access here -> https://arxiv.org/abs/2305.10234) and in order for other researchers to use these data in their own work.

The protocol is intended for the Systematic Literature review on the topic of High-value Datasets with the aim to gather information on how the topic of High-value datasets (HVD) and their determination has been reflected in the literature over the years and what has been found by these studies to date, incl. the indicators used in them, involved stakeholders, data-related aspects, and frameworks. The data in this dataset were collected in the result of the SLR over Scopus, Web of Science, and Digital Government Research library (DGRL) in 2023.

Methodology

To understand how HVD determination has been reflected in the literature over the years and what has been found by these studies to date, all relevant literature covering this topic has been studied. To this end, the SLR was carried out to by searching digital libraries covered by Scopus, Web of Science (WoS), Digital Government Research library (DGRL).

These databases were queried for keywords ("open data" OR "open government data") AND ("high-value data*" OR "high value data*"), which were applied to the article title, keywords, and abstract to limit the number of papers to those, where these objects were primary research objects rather than mentioned in the body, e.g., as a future work. After deduplication, 11 articles were found unique and were further checked for relevance. As a result, a total of 9 articles were further examined. Each study was independently examined by at least two authors.

To attain the objective of our study, we developed the protocol, where the information on each selected study was collected in four categories: (1) descriptive information, (2) approach- and research design- related information, (3) quality-related information, (4) HVD determination-related information.

Test procedure Each study was independently examined by at least two authors, where after the in-depth examination of the full-text of the article, the structured protocol has been filled for each study. The structure of the survey is available in the supplementary file available (see Protocol_HVD_SLR.odt, Protocol_HVD_SLR.docx) The data collected for each study by two researchers were then synthesized in one final version by the third researcher.

Description of the data in this data set

Protocol_HVD_SLR provides the structure of the protocol Spreadsheets #1 provides the filled protocol for relevant studies. Spreadsheet#2 provides the list of results after the search over three indexing databases, i.e. before filtering out irrelevant studies

The information on each selected study was collected in four categories: (1) descriptive information, (2) approach- and research design- related information, (3) quality-related information, (4) HVD determination-related information

Descriptive information
1) Article number - a study number, corresponding to the study number assigned in an Excel worksheet 2) Complete reference - the complete source information to refer to the study 3) Year of publication - the year in which the study was published 4) Journal article / conference paper / book chapter - the type of the paper -{journal article, conference paper, book chapter} 5) DOI / Website- a link to the website where the study can be found 6) Number of citations - the number of citations of the article in Google Scholar, Scopus, Web of Science 7) Availability in OA - availability of an article in the Open Access 8) Keywords - keywords of the paper as indicated by the authors 9) Relevance for this study - what is the relevance level of the article for this study? {high / medium / low}

Approach- and research design-related information 10) Objective / RQ - the research objective / aim, established research questions 11) Research method (including unit of analysis) - the methods used to collect data, including the unit of analy-sis (country, organisation, specific unit that has been ana-lysed, e.g., the number of use-cases, scope of the SLR etc.) 12) Contributions - the contributions of the study 13) Method - whether the study uses a qualitative, quantitative, or mixed methods approach? 14) Availability of the underlying research data- whether there is a reference to the publicly available underly-ing research data e.g., transcriptions of interviews, collected data, or explanation why these data are not shared? 15) Period under investigation - period (or moment) in which the study was conducted 16) Use of theory / theoretical concepts / approaches - does the study mention any theory / theoretical concepts / approaches? If any theory is mentioned, how is theory used in the study?

Quality- and relevance- related information
17) Quality concerns - whether there are any quality concerns (e.g., limited infor-mation about the research methods used)? 18) Primary research object - is the HVD a primary research object in the study? (primary - the paper is focused around the HVD determination, sec-ondary - mentioned but not studied (e.g., as part of discus-sion, future work etc.))

HVD determination-related information
19) HVD definition and type of value - how is the HVD defined in the article and / or any other equivalent term? 20) HVD indicators - what are the indicators to identify HVD? How were they identified? (components & relationships, “input -> output") 21) A framework for HVD determination - is there a framework presented for HVD identification? What components does it consist of and what are the rela-tionships between these components? (detailed description) 22) Stakeholders and their roles - what stakeholders or actors does HVD determination in-volve? What are their roles? 23) Data - what data do HVD cover? 24) Level (if relevant) - what is the level of the HVD determination covered in the article? (e.g., city, regional, national, international)

Format of the file .xls, .csv (for the first spreadsheet only), .odt, .docx

Licenses or restrictions CC-BY

For more info, see README.txt

Time-coincident load, wind, and solar data including actual and probabilistic forecast datasets at 5-min resolution for ERCOT, MISO, NYISO, and SPP. Wind and solar profiles are supplied for existing sites as well as planned sites based on interconnection queue projects as of 2021. For ERCOT actuals are provided for 2017 and 2018 and forecasts for 2018, and for the remaining ISOs actuals are provided for 2018 and 2019 and forecasts for 2019. There datasets were produced by NREL as part of the ARPA-E PERFORM project, an ARPA-E funded program that aim to use time-coincident power and load seeks to develop innovative management systems that represent the relative delivery risk of each asset and balance the collective risk of all assets across the grid. For more information on the datasets and methods used to generate them see https://github.com/PERFORM-Forecasts/documentation.

This data set contains example data for exploration of the theory of regression based regionalization. The 90th percentile of annual maximum streamflow is provided as an example response variable for 293 streamgages in the conterminous United States. Several explanatory variables are drawn from the GAGES-II data base in order to demonstrate how multiple linear regression is applied. Example scripts demonstrate how to collect the original streamflow data provided and how to recreate the figures from the associated Techniques and Methods chapter.

Dataset

This dataset was created by Study Mart

Released under Apache 2.0

Contents

Access the Home Depot products dataset, a comprehensive collection of web-scraped data featuring home improvement products. Discover trending tools, hardware, appliances, décor, and gardening essentials to enhance your projects. From power tools and building materials to lighting, furniture, and outdoor living items, this dataset provides insights into top-rated products, best-selling brands, and emerging trends.

Download now to explore detailed product data for smarter decision-making in home improvement, DIY, and construction projects.

For a closer look at the product-level data we’ve extracted from Home Depot, including pricing, stock status, and detailed specifications, visit the Home Depot dataset page. You can explore sample records and submit a request for tailored extracts directly from there.

Three datasets are intended to be used for exploring machine learning applications in materials science. They are formatted in simple form and in particular for easy input into the MAterials Simulation Toolkit - Machine Learning (MAST-ML) package (see https://github.com/uw-cmg/MAST-ML).Each dataset is a materials property of interest and associated descriptors. For detailed information, please see the attached REAME text file.The first dataset for dilute solute diffusion can be used to predict an effective diffusion barrier for a solute element moving through another host element. The dataset has been calculated with DFT methods.The second dataset for perovskite stability gives energies of compostions of potential perovskite materials relative to the convex hull calculated with DFT. The perovskite dataset also includes columns with information about the A site, B site, and X site in the perovskite structure in order to perform more advanced grouping of the data.The third dataset is a metallic glasses dataset which has values of reduced glass transition temperature (Trg) for a variety of metallic alloys. An additional column is included for majority element for each alloy, which can be an interesting property to group on during tests.

MLFMF MLFMF (Machine Learning for Mathematical Formalization) is a collection of data sets for benchmarking recommendation systems used to support formalization of mathematics with proof assistants. These systems help humans identify which previous entries (theorems, constructions, datatypes, and postulates) are relevant in proving a new theorem or carrying out a new construction. The MLFMF data sets provide solid benchmarking support for further investigation of the numerous machine learning approaches to formalized mathematics. With more than 250,000 entries in total, this is currently the largest collection of formalized mathematical knowledge in machine learnable format. In addition to benchmarking the recommendation systems, the data sets can also be used for benchmarking node classification and link prediction algorithms. The four data sets Each data set is derived from a library of formalized mathematics written in proof assistants Agda or Lean. The collection includes

the largest Lean 4 library Mathlib, the three largest Agda libraries:

the standard library the library of univalent mathematics Agda-unimath, and the TypeTopology library. Each data set represents the corresponding library in two ways: as a heterogeneous network, and as a list of syntax trees of all the entries in the library. The network contains the (modular) structure of the library and the references between entries, while the syntax trees give complete and easily parsed information about each entry. The Lean library data set was obtained by converting .olean files into s-expressions (see the lean2sexp tool). The Agda data sets were obtained with an s-expression extension of the official Agda repository (use either master-sexp or release-2.6.3-sexp branch). For more details, see our arXiv copy of the paper. Directory structure First, the mlfmf.zip archive needs to be unzipped. It contains a separate directory for every library (for example, the standard library of Agda can be found in the stdlib directory) and some auxiliary files. Every library directory contains

the network file from which the heterogeneous network can be loaded, a zip of the entries directory that contains (many) files with abstract syntax trees. Each of those files describes a single entry of the library. In addition to the auxiliary files which are used for loading the data (and described below), the zipped sources of lean2sexp and Agda s-expression extension are present. Loading the data In addition to the data files, there is also a simple python script main.py for loading the data. To run it, you will have to install the packages listed in the file requirements.txt: tqdm and networkx. The easiest way to do so is calling pip install -r requirements.txt. When running main.py for the first time, the script will unzip the entry files into the directory named entries. After that, the script loads the syntax trees of the entries (see the Entry class) and the network (as networkx.MultiDiGraph object). Note. The entry files have extension .dag (directed acyclic graph), since Lean uses node sharing, which breaks the tree structure (a shared node has more than one parent node). More information For more information about the data collection process, detailed data (and data format) description, and baseline experiments that were already performed with these data, see our arXiv copy of the paper. For the code that was used to perform the experiments and data format description, visit our github repository https://github.com/ul-fmf/mlfmf-data. Funding Since not all the funders are available in the Zenodo's database, we list them here:

This material is based upon work supported by the Air Force Office of Scientific Research under award number FA9550-21-1-0024. The authors also acknowledge the financial support of the Slovenian Research Agency via the research core funding No. P2-0103 and No. P1-0294.

The protein-protein interface comparison software PiMine was developed to provide fast comparisons against databases of known protein-protein complex structures. Its application domains range from the prediction of interfaces and potential interaction partners to the identification of potential small molecule modulators of protein-protein interactions.[1]

The protein-protein evaluation datasets are a collection of five datasets that were used for the parameter optimization (ParamOptSet), enrichment assessment (Dimer597 set, Keskin set, PiMineSet), and runtime analyses (RunTimeSet) of protein-protein interface comparison tools. The evaluation datasets contain pairs of interfaces of protein chains that either share sequential and structural similarities or are even sequentially and structurally unrelated. They enable comparative benchmark studies for tools designed to identify interface similarities.

Data Set description:

The ParamOptSet was designed based on a study on improving the benchmark datasets for the evaluation of protein-protein docking tools [2]. It was used to optimize and fine-tune the geometric search parameters of PiMine.

The Dimer597 [3] and Keskin [4] sets were developed earlier. We used them to evaluate PiMine’s performance in identifying structurally and sequentially related interface pairs as well as interface pairs with prominent similarity whose constituting chains are sequentially unrelated.

The PiMine set [1] was constructed to assess different quality criteria for reliable interface comparison. It consists of similar pairs of protein-protein complexes of which two chains are sequentially and structurally highly related while the other two chains are unrelated and show different folds. It enables the assessment of the performance when the interfaces of apparently unrelated chains are available only. Furthermore, we could obtain reliable interface-interface alignments based on the similar chains which can be used for alignment performance assessments.

Finally, the RunTimeSet [1] comprises protein-protein complexes from the PDB that were predicted to be biologically relevant. It enables the comparison of typical run times of comparison methods and represents also an interesting dataset to screen for interface similarities.

References:

[1] Graef, J.; Ehrt, C.; Reim, T.; Rarey, M. Database-driven identification of structurally similar protein-protein interfaces (submitted)
[2] Barradas-Bautista, D.; Almajed, A.; Oliva, R.; Kalnis, P.; Cavallo, L. Improving classification of correct and incorrect protein-protein docking models by augmenting the training set. Bioinform. Adv. 2023, 3, vbad012.
[3] Gao, M.; Skolnick, J. iAlign: a method for the structural comparison of protein–protein interfaces. Bioinformatics 2010, 26, 2259-2265.
[4] Keskin, O.; Tsai, C.-J.; Wolfson, H.; Nussinov, R. A new, structurally nonredundant, diverse data set of protein–protein interfaces and its implications. Protein Sci. 2004, 13, 1043-1055.

Nationally representative, longitudinal information from an evaluation where children were randomly assigned to Head Start or community services as usual;direct assessments and observations of children as well as parent and staff interviews were conducted

It is a widely accepted fact that evolving software systems change and grow. However, it is less well-understood how change is distributed over time, specifically in object oriented software systems. The patterns and techniques used to measure growth permit developers to identify specific releases where significant change took place as well as to inform them of the longer term trend in the distribution profile. This knowledge assists developers in recording systemic and substantial changes to a release, as well as to provide useful information as input into a potential release retrospective. However, these analysis methods can only be applied after a mature release of the code has been developed. But in order to manage the evolution of complex software systems effectively, it is important to identify change-prone classes as early as possible. Specifically, developers need to know where they can expect change, the likelihood of a change, and the magnitude of these modifications in order to take proactive steps and mitigate any potential risks arising from these changes. Previous research into change-prone classes has identified some common aspects, with different studies suggesting that complex and large classes tend to undergo more changes and classes that changed recently are likely to undergo modifications in the near future. Though the guidance provided is helpful, developers need more specific guidance in order for it to be applicable in practice. Furthermore, the information needs to be available at a level that can help in developing tools that highlight and monitor evolution prone parts of a system as well as support effort estimation activities. The specific research questions that we address in this chapter are: (1) What is the likelihood that a class will change from a given version to the next? (a) Does this probability change over time? (b) Is this likelihood project specific, or general? (2) How is modification frequency distributed for classes that change? (3) What is the distribution of the magnitude of change? Are most modifications minor adjustments, or substantive modifications? (4) Does structural complexity make a class susceptible to change? (5) Does popularity make a class more change-prone? We make recommendations that can help developers to proactively monitor and manage change. These are derived from a statistical analysis of change in approximately 55000 unique classes across all projects under investigation. The analysis methods that we applied took into consideration the highly skewed nature of the metric data distributions. The raw metric data (4 .txt files and 4 .log files in a .zip file measuring ~2MB in total) is provided as a comma separated values (CSV) file, and the first line of the CSV file contains the header. A detailed output of the statistical analysis undertaken is provided as log files generated directly from Stata (statistical analysis software).

This EPA SH entry contains the link to the public archive for all datasets collected by EPA and collaborators for the Lake Michigan Ozone Study conducted in 2017. The archive resides at NASA Langley Research Center. EPA-ORD data sets can be found by clicking on the individual site tabs on the main LMOS archive webpage and EPA-ORD will be listed at the PI.

Dataset Card for Magazine dataset

  Dataset Summary

A large-scale magazine layout dataset with fine-grained layout annotations and keyword labeling.

  Supported Tasks and Leaderboards

[More Information Needed]

  Languages

[More Information Needed]

  Dataset Structure





  Data Instances

import datasets as ds

dataset = ds.load_dataset("creative-graphic-design/Magazine")

  Data Fields

[More Information Needed]

  Data Splits… See the full description on the dataset page: https://huggingface.co/datasets/creative-graphic-design/Magazine.

This Public Health Portfolio (Directly Funded Research - Programme and Training Awards) dataset contains NIHR directly funded research awards where the funding is allocated to an award holder or host organisation to carry out a specific piece of research or complete a training award. The NIHR also invests significantly in centres of excellence, collaborations, services and facilities to support research in England. Collectively these form NIHR infrastructure support. NIHR infrastructure supported projects are available in the Public Health Portfolio (Infrastructure Support) dataset which you can find here.NIHR directly funded research awards (Programmes and Training Awards) that were funded between January 2006 and the present extraction date are eligible for inclusion in this dataset. An agreed inclusion/exclusion criteria is used to categorise awards as public health awards (see below). Following inclusion in the dataset, public health awards are second level coded to one of the four Public Health Outcomes Framework domains. These domains are: (1) wider determinants (2) health improvement (3) health protection (4) healthcare and premature mortality.More information on the Public Health Outcomes Framework domains can be found here.This dataset is updated quarterly to include new NIHR awards categorised as public health awards. Please note that for those Public Health Research Programme projects showing an Award Budget of £0.00, the project is undertaken by an on-call team for example, PHIRST, Public Health Review Team, or Knowledge Mobilisation Team, as part of an ongoing programme of work.Inclusion CriteriaThe NIHR Public Health Overview project team worked with colleagues across NIHR public health research to define the inclusion criteria for NIHR public health research. NIHR directly funded research awards are categorised as public health if they are determined to be ‘investigations of interventions in, or studies of, populations that are anticipated to have an effect on health or on health inequity at a population level.’ This definition of public health is intentionally broad to capture the wide range of NIHR public health research across prevention, health improvement, health protection, and healthcare services (both within and outside of NHS settings). This dataset does not reflect the NIHR’s total investment in public health research. The intention is to showcase a subset of the wider NIHR public health portfolio. This dataset includes NIHR directly funded research awards categorised as public health awards. This dataset does not include public health awards or projects funded by any of the three NIHR Research Schools or NIHR Health Protection Research Units.DisclaimersUsers of this dataset should acknowledge the broad definition of public health that has been used to develop the inclusion criteria for this dataset. Please note that this dataset is currently subject to a limited data quality review. We are working to improve our data collection methodologies. Please also note that some awards may also appear in other NIHR curated datasets. Further InformationFurther information on the individual awards shown in the dataset can be found on the NIHR’s Funding & Awards website here. Further information on individual NIHR Research Programme’s decision making processes for funding health and social care research can be found here.Further information on NIHR’s investment in public health research can be found as follows:The NIHR is one of the main funders of public health research in the UK. Public health research falls within the remit of a range of NIHR Directly Funded Research (Programmes and Training Awards), and NIHR Infrastructure Support. NIHR School for Public Health here.NIHR Public Health Policy Research Unit here. NIHR Health Protection Research Units here.NIHR Public Health Research Programme Health Determinants Research Collaborations (HDRC) here.NIHR Public Health Research Programme Public Health Intervention Responsive Studies Teams (PHIRST) here.

We use journal articles published by Scientometrics in 2016-2020 as the data source. Through the analysis of the data set usage records of scientometrics research, the frequency ranking of the usage of datasets on respectively social network, patents, academic literature, and encyclopedia are listed, so as to provide a reference for the selection of data sets for scientometrics research.

This Public Schools feature dataset is composed of all Public elementary and secondary education facilities in the United States as defined by the Common Core of Data (CCD, https://nces.ed.gov/ccd/ ), National Center for Education Statistics (NCES, https://nces.ed.gov ), US Department of Education for the 2017-2018 school year. This includes all Kindergarten through 12th grade schools as tracked by the Common Core of Data. Included in this dataset are military schools in US territories and referenced in the city field with an APO or FPO address. DOD schools represented in the NCES data that are outside of the United States or US territories have been omitted. This feature class contains all MEDS/MEDS+ as approved by NGA. Complete field and attribute information is available in the ”Entities and Attributes” metadata section. Geographical coverage is depicted in the thumbnail above and detailed in the Place Keyword section of the metadata. This release includes the addition of 3065 new records, modifications to the spatial location and/or attribution of 99,287 records, and removal of 2996 records not present in the NCES CCD data.

This replication package contains datasets and scripts related to the paper: "*How do Hugging Face Models Document Datasets, Bias, and Licenses? An Empirical Study*"

## Root directory

- `statistics.r`: R script used to compute the correlation between usage and downloads, and the RQ1/RQ2 inter-rater agreements

- `modelsInfo.zip`: zip file containing all the downloaded model cards (in JSON format)

- `script`: directory containing all the scripts used to collect and process data. For further details, see README file inside the script directory.

## Dataset

- `Dataset/Dataset_HF-models-list.csv`: list of HF models analyzed

- `Dataset/Dataset_github-prj-list.txt`: list of GitHub projects using the *transformers* library

- `Dataset/Dataset_github-Prj_model-Used.csv`: contains usage pairs: project, model

- `Dataset/Dataset_prj-num-models-reused.csv`: number of models used by each GitHub project

- `Dataset/Dataset_model-download_num-prj_correlation.csv` contains, for each model used by GitHub projects: the name, the task, the number of reusing projects, and the number of downloads

## RQ1

- `RQ1/RQ1_dataset-list.txt`: list of HF datasets

- `RQ1/RQ1_datasetSample.csv`: sample set of models used for the manual analysis of datasets

- `RQ1/RQ1_analyzeDatasetTags.py`: Python script to analyze model tags for the presence of datasets. it requires to unzip the `modelsInfo.zip` in a directory with the same name (`modelsInfo`) at the root of the replication package folder. Produces the output to stdout. To redirect in a file fo be analyzed by the `RQ2/countDataset.py` script

- `RQ1/RQ1_countDataset.py`: given the output of `RQ2/analyzeDatasetTags.py` (passed as argument) produces, for each model, a list of Booleans indicating whether (i) the model only declares HF datasets, (ii) the model only declares external datasets, (iii) the model declares both, and (iv) the model is part of the sample for the manual analysis

- `RQ1/RQ1_datasetTags.csv`: output of `RQ2/analyzeDatasetTags.py`

- `RQ1/RQ1_dataset_usage_count.csv`: output of `RQ2/countDataset.py`

## RQ2

- `RQ2/tableBias.pdf`: table detailing the number of occurrences of different types of bias by model Task

- `RQ2/RQ2_bias_classification_sheet.csv`: results of the manual labeling

- `RQ2/RQ2_isBiased.csv`: file to compute the inter-rater agreement of whether or not a model documents Bias

- `RQ2/RQ2_biasAgrLabels.csv`: file to compute the inter-rater agreement related to bias categories

- `RQ2/RQ2_final_bias_categories_with_levels.csv`: for each model in the sample, this file lists (i) the bias leaf category, (ii) the first-level category, and (iii) the intermediate category

## RQ3

- `RQ3/RQ3_LicenseValidation.csv`: manual validation of a sample of licenses

- `RQ3/RQ3_{NETWORK-RESTRICTIVE|RESTRICTIVE|WEAK-RESTRICTIVE|PERMISSIVE}-license-list.txt`: lists of licenses with different permissiveness

- `RQ3/RQ3_prjs_license.csv`: for each project linked to models, among other fields it indicates the license tag and name

- `RQ3/RQ3_models_license.csv`: for each model, indicates among other pieces of info, whether the model has a license, and if yes what kind of license

- `RQ3/RQ3_model-prj-license_contingency_table.csv`: usage contingency table between projects' licenses (columns) and models' licenses (rows)

- `RQ3/RQ3_models_prjs_licenses_with_type.csv`: pairs project-model, with their respective licenses and permissiveness level

## scripts

Contains the scripts used to mine Hugging Face and GitHub. Details are in the enclosed README