Author: Gayatri Babel, Shreyans Sethi, Ayush Sehgal
Organization: UC Berkeley Info 159
The questions in this section are primarily intended to encourage dataset creators to clearly articulate their reasons for creating the dataset and to promote transparency about funding interests.
-
For what purpose was the dataset created? Was there a specific task in mind? Was there a specific gap that needed to be filled? Please provide a description.
The purpose of this dataset was to analyze condescension in New York Times article headlines and descriptions. The aim was to investigate the level of condescension, if any, of news articles from the New York Times.
-
Who created this dataset (e.g. which team, research group) and on behalf of which entity (e.g. company, institution, organization)?
*This dataset was created by a team of students (Ayush Sehgal, Shreyans Sethi, Gayatri Babel) for a project in the course INFO 159: Natural Language Processing at UC Berkeley. *
-
What support was needed to make this dataset? (e.g. who funded the creation of the dataset? If there is an associated grant, provide the name of the grantor and the grant name and number, or if it was supported by a company or government agency, give those details.)
*The dataset was created with the support of the course staff of INFO 159. *
-
Any other comments?
Your Answer Here
Dataset creators should read through the questions in this section prior to any data collection and then provide answers once collection is complete. Most of these questions are intended to provide dataset consumers with the information they need to make informed decisions about using the dataset for specific tasks. The answers to some of these questions reveal information about compliance with the EU’s General Data Protection Regulation (GDPR) or comparable regulations in other jurisdictions.
-
What do the instances that comprise the dataset represent (e.g. documents, photos, people, countries)? Are there multiple types of instances (e.g. movies, users, and ratings; people and interactions between them; nodes and edges)? Please provide a description.
The instances that comprise the dataset are portions of online documents, specifically the headline and article description of online news articles from the New York Times online archive. These are accompanied by instances of ratings that dictate the level of condescension found for each article description and headline.
-
How many instances are there in total (of each type, if appropriate)?
There are 1000 instances of each instance.
-
Does the dataset contain all possible instances or is it a sample (not necessarily random) of instances from a larger set? If the dataset is a sample, then what is the larger set? Is the sample representative of the larger set (e.g. geographic coverage)? If so, please describe how this representativeness was validated/verified. If it is not representative of the larger set, please describe why not (e.g. to cover a more diverse range of instances, because instances were withheld or unavailable).
This dataset was only a random sample of news articles from New York Times’ online web archive of all past articles. In order to ensure a representative sample was collected, 100 news articles from each year from the past decade. The 100 articles were randomly selected to avoid bias and ensure equal coverage of every geographic region; however, it is important to note that geographic coverage is also dependent on the outlet itself and the breaking news at teh time.
-
What data does each instance consist of? "Raw" data (e.g. unprocessed text or images) or features? In either case, please provide a description.
The raw data of each instance is composed of unprocessed text (headline + article description) along with a condescension rating given by members of the team.
-
Is there a label or target associated with each instance? If so, please provide a description.
A condescension rating between 1 - 5 was associated with each instance. These labels were given based on guidelines composed by the team that created the dataset.
-
Is any information missing from individual instances? If so, please provide a description, explaining why this information is missing (e.g. because it was unavailable). This does not include intentionally removed information, but might include, e.g. redacted text.
There was no information missing from individual instances.
-
Are relationships between individual instances made explicit (e.g. users' movie ratings, social network links)? If so, please describe how these relationships are made explicit.
Although the instances were all from the same source (New York Times website), the instances themselves are individual.
-
Are there recommended data splits (e.g. training, development/validation, testing)? If so, please provide a description of these splits, explaining the rationale behind them.
When splitting the dataset, it is important to be mindful of the split of the labels. The majority of the dataset is composed of instances labeled with 1, with very few instances labeled with ratings between 2 - 5. In order to ensure adequate model training and testing, it is important to ensure that there is a strong presence of minority labeled instances (instances labeled 2 - 5) in the training, validation, and testing sets.
-
Are there any errors, sources of noise, or redundancies in the dataset? If so, please provide a description.
There is a majority of instances labeled as 1s. To counteract this, the dataset must be downsampled or undersampled or more instances with minority labels should be added to the dataset.
-
Is the dataset self-contained, or does it link to or otherwise rely on external resources (e.g. websites, tweets, other datasets)? If it links to or relies on external resources, a) are there guarantees that they will exist, and remain constant, over time; b) are there official archival versions of the complete dataset (i.e., including the external resources as they existed at the time the dataset was created); c) are there any restrictions (e.g. licenses, fees) associated with any of the external resources that might apply to a future user? Please provide descriptions of all external resources and any restrictions associated with them, as well as links or other access points, as appropriate.
The article headlines and descriptions were primarily sourced from New York Times web archive and thus, is linked to the New York Times web archive. The archive includes all the New York Times articles dating from the 1800s and it will likely exist in the future. There are some licensing restrictions as the entire article cannot be used, but the headlines and article descriptions are publicly available.
-
Does the dataset contain data that might be considered confidential (e.g. data that is protected by legal privilege or by doctor-patient confidentiality, data that includes the content of individuals' non-public communications)? If so, please provide a description.
The dataset is not considered confidential.
-
Does the dataset contain data that, if viewed directly, might be offensive, insulting, threatening, or might otherwise cause anxiety? If so, please describe why.
No.
-
Does the dataset relate to people? If not, you may skip the remaining questions in this section.
The dataset relates to news about the world and individuals, thus it is related to people.
-
Does the dataset identify any subpopulations (e.g. by age, gender)? If so, please describe how these subpopulations are identified and provide a description of their respective distributions within the dataset.
Some instances in the dataset may identify particular regions or demographics based on the news topic. These subpopulations are often identified within the headline or article description. It was difficult to ensure equal distributions of these subpopulations and regions as this is highly dependent on the news that was reported on at the time. Additionally, it was difficult to gather a random sample while ensuring an equal distribution.
-
Is it possible to identify individuals (i.e., one or more natural persons), either directly or indirectly (i.e., in combination with other data) from the dataset? If so, please describe how.
If the article was about a particular individual, then one could identify the specific natural person within the headline or article as it would mention the specific individual(s).
-
Does the dataset contain data that might be considered sensitive in any way (e.g. data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history)? If so, please provide a description.
No.
-
Any other comments?
No.
As with the previous section, dataset creators should read through these questions prior to any data collection to flag potential issues and then provide answers once collection is complete. In addition to the goals of the prior section, the answers to questions here may provide information that allow others to reconstruct the dataset without access to it.
-
How was the data associated with each instance acquired? Was the data directly observable (e.g. raw text, movie ratings), reported by subjects (e.g. survey responses), or indirectly inferred/derived from other data (e.g. part-of-speech tags, model-based guesses for age or language)? If data was reported by subjects or indirectly inferred/derived from other data, was the data validated/verified? If so, please describe how.
All the data we used was readily available as an entry in the table on the New York Times Archives website (linked here: https://www.nytimes.com/search/?srchst=nyt). This data is scraped by the internal New York Times team with the headlines, genre, and a short description, along with the author of every article. This data off the shelf has data from 1981 to present, what we did was focus on data in the last 10 years up until the current year. We also only focused on articles under the ‘World’ section with a type specified as ‘Article’. There is no verification or inference needed here as this dataset just pulls from existing articles.
-
What mechanisms or procedures were used to collect the data (e.g. hardware apparatus or sensor, manual human curation, software program, software API)? How were these mechanisms or procedures validated?
The New York Times has digitized their catalog of articles into a searchable database at the link in the above question. This involved a mix of having to scan articles by hand and of storing existing articles in a fresh database available for querying. Because we only used articles from the past 10 years, we did not use any of the hand-scanned articles.
-
If the dataset is a sample from a larger set, what was the sampling strategy (e.g. deterministic, probabilistic with specific sampling probabilities)?
The dataset is the entire collection of articles, it is not a sample. However, to choose the 1000 to bring into our dataset we did that randomly, where each article had an equal probability of being included. However, we made sure to pick a roughly equal number of articles from each of the 10 years, so we had 100 articles per year between 2011-2022.
-
Who was involved in the data collection process (e.g. students, crowdworkers, contractors) and how were they compensated (e.g. how much were crowdworkers paid)?
The data was collected by simply scraping the archives page and transferring that onto a spreadsheet.
-
Over what timeframe was the data collected? Does this timeframe match the creation timeframe of the data associated with the instances (e.g. recent crawl of old news articles)? If not, please describe the timeframe in which the data associated with the instances was created. Finally, list when the dataset was first published.
The archives have been divided into 1851-1981 and 1981-present, we used the latter. This archive is automatically and periodically updated so it even has articles published on the current day.
-
Were any ethical review processes conducted (e.g. by an institutional review board)? If so, please provide a description of these review processes, including the outcomes, as well as a link or other access point to any supporting documentation.
Since these are New York Times articles, the review process took place when the articles were being written and before they got approval for publication.
-
Does the dataset relate to people? If not, you may skip the remainder of the questions in this section.
N/A
-
Did you collect the data from the individuals in question directly, or obtain it via third parties or other sources (e.g. websites)?
N/A
-
Were the individuals in question notified about the data collection? If so, please describe (or show with screenshots or other information) how notice was provided, and provide a link or other access point to, or otherwise reproduce, the exact language of the notification itself.
N/A
-
Did the individuals in question consent to the collection and use of their data? If so, please describe (or show with screenshots or other information) how consent was requested and provided, and provide a link or other access point to, or otherwise reproduce, the exact language to which the individuals consented.
N/A
-
If consent was obtained, were the consenting individuals provided with a mechanism to revoke their consent in the future or for certain uses? If so, please provide a description, as well as a link or other access point to the mechanism (if appropriate).
N/A
-
Has an analysis of the potential impact of the dataset and its use on data subjects (e.g. a data protection impact analysis) been conducted? If so, please provide a description of this analysis, including the outcomes, as well as a link or other access point to any supporting documentation.
N/A
-
Any other comments?
We specifically chose the headlines and descriptions because we wanted to see the impact the data you will see about an article at first glance has on condescension levels in that article. So the archive was perfect because it has this concise information in a structured list.
Dataset creators should read through these questions prior to any pre-processing, cleaning, or labeling and then provide answers once these tasks are complete. The questions in this section are intended to provide dataset consumers with the information they need to determine whether the “raw” data has been processed in ways that are compatible with their chosen tasks. For example, text that has been converted into a “bag-of-words” is not suitable for tasks involving word order.
-
Was any preprocessing/cleaning/labeling of the data done (e.g. discretization or bucketing, tokenization, part-of-speech tagging, SIFT feature extraction, removal of instances, processing of missing values)? If so, please provide a description. If not, you may skip the remainder of the questions in this section.
Labeling of the data was performed on a scale from 1 to 5 measuring condescension. A rank of 1 implies a factual and not condescending article and 5 is the opposite, indicating a very condescending article. We did perform Named Entity Recognition, POS tagging, and applied a BERT encoder to the data but not as a preprocessing step, but in our final model instead.
-
Was the "raw" data saved in addition to the preprocessed/cleaned/labeled data (e.g. to support unanticipated future uses)? If so, please provide a link or other access point to the "raw" data.
The raw data was saved as the X values in our data with their corresponding Y values being the ranking (1 to 5).
-
Is the software used to preprocess/clean/label the instances available? If so, please provide a link or other access point.
So far, the model mentioned above has not been published on Github or any other place so there is no available link, but can be provided on special request.
-
Any other comments?
N/A
These questions are intended to encourage dataset creators to reflect on the tasks for which the dataset should and should not be used. By explicitly highlighting these tasks, dataset creators can help dataset consumers to make informed decisions, thereby avoiding potential risks or harms.
- Has the dataset been used for any tasks already? If so, please provide a description.
So far, the dataset has been used to train and test a logistic regression model that aims to predict the condescension of the articles. 600 of the data points were used for training with the others being used for hyperparameter tuning and testing. However, this model has not been deployed anywhere or released publicly since it was part of an internal assignment. There are no other tasks this dataset is being used for.
- Is there a repository that links to any or all papers or systems that use the dataset? If so, please provide a link or other access point.
So far, the model mentioned above has not been published on Github or any other place so there is no available link.
- What (other) tasks could the dataset be used for?
Since it is a collection of New York Times articles under the “World” category, it could be used for a sentiment analysis purpose in which we could investigate how many articles have a positive or negative connotation. It could also be used to look more into the style of the language used in a typical article headline - Is it passive, assertive, aggressive, etc.
- Is there anything about the composition of the dataset or the way it was collected and preprocessed/cleaned/labeled that might impact future uses? For example, is there anything that a future user might need to know to avoid uses that could result in unfair treatment of individuals or groups (e.g. stereotyping, quality of service issues) or other undesirable harms (e.g. financial harms, legal risks) If so, please provide a description. Is there anything a future user could do to mitigate these undesirable harms?
The articles were collected from 2012 to 2022 (inclusive) with 90-100 articles being collected from each year. However, during a given year, the articles chosen were not spread out evenly through the year (i.e. Approximately 2 articles per week) but rather taken from 3 general periods - Start of the year, sometime in the summer, End of the year (30-35 articles from each period). As such, the actual news mentioned within the articles would bias towards the events that took place around these given time periods and the countries/individuals involved in those events. As such, certain countries may be overrepresented/underrepresented in the dataset simply as a result of whether they had important events taking place during those times and therefore, the dataset should not be used to check which countries receive more attention.
In terms of legal risks, the New York Times requires that its article headlines and descriptors can be used but they must be used with the links to the original articles included so any future user must remember this.
- Are there tasks for which the dataset should not be used? If so, please provide a description.
As mentioned above, the dataset should not be used for investigations into who the New York Times typically writes about (and correspondingly, which countries/people get less attention) as the data was not collected in a time-consistent manner so it will be biased by the events taking place at the time. The dataset should also not be used for investigating any topic that is not centered on World News as this was the only type of NYT articles that were collected for the dataset.
- Any other comments?
N/A
Dataset creators should provide answers to these questions prior to distributing the dataset either internally within the entity on behalf of which the dataset was created or externally to third parties.
- Will the dataset be distributed to third parties outside of the entity (e.g. company, institution, organization) on behalf of which the dataset was created? If so, please provide a description.
N/A - This will not be distributed to anyone but rather used for an internal assignment completed by students at the University of California, Berkeley. The only individuals with access to it are the students themselves and the course staff for the Natural Language Processing course.
- How will the dataset will be distributed (e.g. tarball on website, API, GitHub)? Does the dataset have a digital object identifier (DOI)?
N/A
- When will the dataset be distributed?
N/A
- Will the dataset be distributed under a copyright or other intellectual property (IP) license, and/or under applicable terms of use (ToU)? If so, please describe this license and/or ToU, and provide a link or other access point to, or otherwise reproduce, any relevant licensing terms or ToU, as well as any fees associated with these restrictions.
N/A
- Have any third parties imposed IP-based or other restrictions on the data associated with the instances? If so, please describe these restrictions, and provide a link or other access point to, or otherwise reproduce, any relevant licensing terms, as well as any fees associated with these restrictions.
N/A
- Do any export controls or other regulatory restrictions apply to the dataset or to individual instances? If so, please describe these restrictions, and provide a link or other access point to, or otherwise reproduce, any supporting documentation.
N/A
- Any other comments?
N/A
As with the previous section, dataset creators should provide answers to these questions prior to distributing the dataset. These questions are intended to encourage dataset creators to plan for dataset maintenance and communicate this plan with dataset consumers.
- Who is supporting/hosting/maintaining the dataset?
The dataset is currently not being hosted at a publicly available repository and, as such, is not being supported or maintained for future updates. Individuals can recreate the dataset by visiting the New York Times Archive (https://archive.nytimes.com/www.nytimes.com/ref/membercenter/nytarchive.html) and choose articles within their desired time range and topics since this archive will keep being updated as the NYT publishes more articles.
- How can the owner/curator/manager of the dataset be contacted (e.g. email address)?
The internal database of 1000 data points is not being managed but for inquiries, the 3 students who worked on a project with this data can be contacted at: gbabel@berkeley.edu, ayushs25@berkeley.edu and shreyanssethi@berkeley.edu. The NYT can also be contacted directly for more detailed questions about the production of the original data points at this phone number: +1 855-698-1157
- Is there an erratum? If so, please provide a link or other access point.
N/A
- Will the dataset be updated (e.g. to correct labeling errors, add new instances, delete instances)? If so, please describe how often, by whom, and how updates will be communicated to users (e.g. mailing list, GitHub)?
As of right now, there are no plans to update the dataset but if the dataset is eventually released publicly, it will be published at Github where it will be updated yearly to include articles from that year and users can see announcements about changes there.
- If the dataset relates to people, are there applicable limits on the retention of the data associated with the instances (e.g. were individuals in question told that their data would be retained for a fixed period of time and then deleted)? If so, please describe these limits and explain how they will be enforced.
N/A
- Will older versions of the dataset continue to be supported/hosted/maintained? If so, please describe how. If not, please describe how its obsolescence will be communicated to users.
If the dataset is eventually published on Github, it will be updated on a yearly basis to add more articles and their scores. However, older data points will not be removed so users will still have access to those points. They can also view older versions of the dataset on Github if they wish not to have the new articles included in their investigations.
- If others want to extend/augment/build on/contribute to the dataset, is there a mechanism for them to do so? If so, please provide a description. Will these contributions be validated/verified? If so, please describe how. If not, why not? Is there a process for communicating/distributing these contributions to other users? If so, please provide a description.
Since the dataset is right not not available to the public, there is no easy mechanism for contributing to it. However, if it is eventually published to Github, users will be able to fork off the main branch and add new data points. These changes will then be reviewed by the 3 original creators of the dataset to see if they fulfill basic requirements (Datapoints come from the NYT, include the URL of the article, include the headline and the description, are in the ‘World’ category, don’t include opinion articles or video articles, are from a chosen year, etc.) and if they are successful, this other branch will be merged with the original to be available to all users.
- Any other comments?
N/A