< BACK TO ALL BLOGS
Glossary(Arrange in alphabetical order)
June 9, 2023
API
An API is a way for different programs to talk to each other and share information like
two people talking. The Moderation API can detect inappropriate content, such as hate speech,
spam, pornographic material, or graphic violence, and highlight or remove it.
Automated & AI-powered Moderation
Content moderation that designed to analyze user-generated content such as text, images
or videos and determine whether they comply with platform policies or not. This is the main
work we at NEXTDATA do successfully for many of our customers.
Automation Rate
A measure of how much of a job can be automated. Besedo has helped companies like Kaidee
and Change.org achieve really high numbers of automation.
Average Reviewing Time (ART)
The average time it takes a piece of content to be reviewed. Latency kills, but faster is not
always more accurate.
Balancing Free Speech and Content Restrictions
The tension between allowing free expression and maintaining a safe and respectful environment.
Platforms must strike a balance between allowing users to express themselves freely while also
enforcing content policies to prevent harmful or inappropriate content from being shared. NEXTDATA
provides a one-stop solution for child safety protection, safeguarding the physical and mental health
of minors
Child Safety
Child safety can be achieved in many ways, such as by providing a secure and nurturing
environment, teaching children basic safety skills, monitoring their use of technology and online
activities, reporting any suspected abuse or neglect to appropriate authorities, and providing access
to support and resources for both children and parents or caregivers.
Code of Conduct
A set of ethical guidelines that govern the behavior of users on a platform. The code of
conduct usually includes policies on respectful behavior, non-discrimination, and other ethical
considerations.
Community Guidelines
Guidelines that outline the rules and expectations for platform users. These include policies
on content, behavior, and conduct.
Content Policies
The Content policies outline what types of content are allowed or prohibited on a platform.
What can users write, and what type of images and videos can they post? This can include guidelines
on hate speech, harassment, explicit content, and other inappropriate content.
Copyright Infringement
The unauthorized use of copyrighted material in a way that violates one of the copyright owner’s
exclusive rights, such as the right to reproduce or perform the copyrighted work or to do derivative
works. Examples of copyright infringement include copying a song from the internet without permission,
downloading pirated movies, or it could be using images on an online marketplace without permission.
Copyright infringement is illegal and is subject to criminal and civil penalties.
False Positive
An alert that incorrectly indicates that malicious activity is occurring.
Filters
Filters play a crucial role in content moderation as they can automatically identify and remove
inappropriate content, such as hate speech or explicit images, before reaching the audience on a
platform.
Fraud Prevention
Generative AI models are typically trained on large datasets and use various techniques such as
deep learning, reinforcement learning, and evolutionary algorithms to learn patterns and generate new
data. This can involve training models to generate new images or videos by learning patterns and
features from existing data, or generating new pieces of text or music by learning from patterns and
trends in existing examples.Some of the key applications of generative AI include generating realistic
images and videos for virtual and augmented reality applications, creating natural language processing
models for chatbots and other conversational interfaces, and developing creative tools for artists and
designers.
Hate Speech And Harassment
Offensive, threatening, or discriminatory speech. Targeted attacks on individuals or groups
based on race, gender, religion, or other characteristics.
Human Exploitation
Some common forms of human exploitation include human trafficking, forced labor, child
labor, sex trafficking, and debt bondage. These practices are often associated with organized crime
groups, but can occur in a variety of contexts, including in domestic settings, businesses, and
industries.
Image Recognition
Technology that can identify and classify images. In content moderation, this is used to
identify and remove inappropriate or explicit images. It can be nudity, text in images, underage
people, and a lot more.
Inappropriate Content
Simply content that violates a platform’s community guidelines or terms of service. This
can include hate speech, harassment, and explicit content that violate platform policies. What this
entail is different from platform to platform.
Machine Learning
A type of artificial intelligence that allows the software to learn and improve over time
without being explicitly programmed. This can be used in automated moderation tools to improve
accuracy and efficiency.
Misinformation And Fake News
False information that is spread intentionally or unintentionally. Including conspiracy
theories, hoaxes, and other forms of misinformation.
Natural Language Processing (NLP)
Technology that can analyze and understand human language. In content moderation, NLP identifies
and removes inappropriate language and hate speech. But it’s so much more than that. Natural language
processing is also a way for a machine to learn the difference between online banter and actual
threats. It’s a way for the machine to learn about sarcasm and all those things we humans take for
granted. Including the recently popular ChatGPT.
Platform-generated Content
Platform-generated content is designed to promote the platform's brand, provide information to
users, and enhance the user's experience by offering personalized recommendations or suggestions. This
type of content is typically controlled by the platform itself, so it can be tailored to fit their
specific goals and values, and to align with the interests of their target audience.
Precision Rate
Precision rate is a measure of the accuracy of positive predictions made by a model. It tells
us how many of the selected risks are correctly identified (i.e. “true positives”)?(It can be
understood as:Risk content that can be correctly identified among all identified risks)
Recall
Recall is a measure of the completeness or sensitivity of a classification model. It tells us
How many risks are selected?(It can be understood as:All suspected risks that can be found)
Spam And Scams
Unsolicited messages or attempts to deceive users for financial gain. Oftentimes this includes
pornographic diversion, phishing, spam website/email and other information, and other forms of unwanted
communication.
Take-down
Action to remove content or a user form platform.
Trust & Safety
Refers to measures to ensure a safe and trustworthy environment for users, including policies,
reporting tools, and risk identification systems, to build user trust and protect against harmful or
abusive content or behavior.
User-generated Content (UGC)
Content that is created by users of a platform or website. Examples include any text, images,
and videos uploaded by users.