Case Studies

How world-renowned publishers, platforms, and researchers are using Perspective API

Though toxicity is pervasive on the internet, each publisher or platform is presented with a unique challenge in reducing it. Developers and publishers can use the API to give feedback to commenters, help moderators review comments, allow readers to more easily find interesting or productive comments, and more.

Show all
Publishers
Platforms
Inspirational Use

Many readers of The New York Times will tell you they value the conversations in the comment section almost as much as the journalism itself. Sometimes the comments themselves become part of the story.

Local newspapers help people learn about and act on local issues. The forums of local news outlets have taken the place of town squares, and offer communities the opportunity for genuine connection and local participation.

Developers create bots to moderate content in real-time

Historically, publishers have had two ways to moderate content—hire human moderators to review all comments before being posted, or ask users to flag bad content on the site. Neither option allowed for real-time moderation.

Coral by Vox Media provides publishers with world-class conversation tools and strategies. More than 200 sites in 21 countries trust Coral to power their communities, including the Washington Post, the Wall Street Journal, The Financial Times, the LA Times, USA Today, Der Spiegel, and Wired.

With the explosive growth of online gaming sites comes more risk of toxicity and online harassment. FACEIT, one of Europe’s largest gaming platforms, wanted to find new ways for the community to engage free of harassment. Perspective API provided FACEIT with the opportunity to make gamers more mindful of their behavior at scale.

Moderators can combine APIs to moderate audio

Audio content has traditionally been challenging for platforms to moderate. To help moderators manage audio content at scale, the Google Cloud team combined several machine learning-based APIs including Google’s Natural Language API and Perspective API, into a pipeline which can analyze audio content and provide a corresponding tool to view the results.

Disqus is the largest comment hosting service, with more than 50 million monthly comments across 200 countries, 2 billion monthly unique visitors, and 17 billion monthly page views. Their mission is to help publishers succeed in an open and diverse internet and they are trusted by millions of publishers, websites and blogs around the world, including ABC News, Entertainment Weekly, and Rotten Tomatoes.

OpenWeb is a leading audience engagement platform that hosts 100 million active users per month. Their clients include publishers like AOL, HuffPost, TechCrunch, Fox News, Hearst, Refinery29, Sky Sports, MarketWatch, Salon, Newsweek and many others. Their mission is to elevate conversations online: they want to inspire an open exchange of ideas across the web — with less hostility and toxicity. Their goal is to encourage healthy debate and make room for diverse opinions, without suppressing free speech.

The New York Times

Ensuring the comment section stays open so communities can thrive

Despite their best efforts, The New York Times only had the capacity to moderate comments on about 10% of articles, and had to disable them on the rest. They turned to Perspective API for a solution that would help them support comments across a larger number of articles. The New York Times and the Perspective API team worked together to develop Moderator, a tool that can be used to find patterns, spot abusive language or online harassment, and score comments based on toxicity. The New York Times can use this information to give real-time feedback to commenters and help human moderators sort comments faster. Through the use of Perspective API, the Times was able to triple the number of articles offering comments.

3x the number of articles offering comments

Southeast Missourian

Improving conversations and building community via online comments

The Southeast Missourian was ready to turn off the comment section because it was too expensive and time consuming to moderate. Instead, they integrated with Perspective API to help manage the conversations happening on their site. Perspective API gives a toxicity score to each piece of content. Using this score as a guide, Southeast Missourian moderators can make the final decision to remove or not allow a comment to be posted. Due to this experiment, comments that were highly likely to be toxic were reduced by 96%. Now, this digital town hall can foster a sense of community without fear of harassment.

96% decrease in toxic comments

Developers create bots to moderate content in real-time

Allowing content platforms to handle bad user input in real time

Using Perspective, developers can create an AI-powered moderation bot to help solve this problem. When a chat is sent, bots can utilize Perspective API to analyze and score the text for different attributes including “Insult”, “Spam”, and “Flirtation.” From there, the bots can react with a dynamic emoji. For example, if a user writes a toxic message, the bots could flag it with a dynamite emoji. When a user gets too many emojis, they could be kicked out of the conversation. Check out this video to learn more.

Coral

Helping commenters improve their behavior online

In 2018, the Perspective API team conducted a study in partnership with Wikipedia which suggested that a significant portion of toxic comments came from people who did not have a history of posting offensive content. This led moderators to wonder if they could help these users make healthier contributions by giving them a moment to consider the impact of their language through a technical intervention.

Perspective API has been used by a number of organizations, including Coral, to explore this idea. Coral provides feedback to commenters with the help of Perspective API, which allows Coral to pre-screen comments for toxicity, tone, or language, and give commenters an opportunity to change their wording before they post. Coral teamed up with McClatchy to do a controlled study to measure its impact and found that 36% of users changed their comment in response to feedback.

36% of users changed their comment in response to feedback

FACEIT

Tackling toxicity on gaming sites through machine learning

Perspective API provides a toxicity score, which the team uses alongside two additional metrics, to flag incidents of potential toxicity or harassment and decide whether a warning or ban should be automatically issued to the user involved. Gamers then have the ability to appeal this decision for review by human moderators. With this strategy, FACEIT is able to use Perspective API to reduce toxicity in real time, while still enabling moderators to make corrections when needed. The team also uses Perspective API to pinpoint the exact messages that triggered the incidents and provide feedback to the gamers. FACEIT has had a 20% reduction in toxic messages since working with Perspective API.

20% reduction in toxic messaging

Moderators can combine APIs to moderate audio

Quickly categorize and evaluate audio content

For each new audio file uploaded, Perspective API analyzes both blocks of text and individual words to generate a prediction on the toxicity of each chunk. The final product is represented in a user interface which displays the audio transcripts and probability that it contains toxic content. Users can filter the results based on the generated toxicity levels, and organizations can choose a certain limit above which they manually review all files. More information can be found here on the Google Cloud blog.

El País

Making the comments section less toxic

El País hoped its comments section would facilitate positive community engagement. Instead, trolls and toxic comments took over. When El País was introduced to Perspective API, they saw an opportunity to keep the comments section open. They partnered with the Perspective API team to analyze their vast trove of public comments and train a model to spot toxicity in Spanish. After integrating with Perspective API, El País is now able to identify when someone tries to post a toxic comment and show the writer a message in real time suggesting they make changes or rewrite the comment. Since El País started using Perspective API to complement their human moderation system, the number of comments has increased 19% and the average toxicity of comments decreased 7%.

19% increase in comments

7% decrease in toxicity

Disqus

Reducing toxicity in comment sections across the world.

Disqus was one of the first platforms to adopt Perspective API for their detection and management of toxic comments, through their Toxicity Mod Filter which uses natural language processing and machine learning including Perspective API to detect and tag comments in the Moderation Panel that have a high likelihood of being toxic. Moderators can use this information to prioritize their efforts and lower the negative impact of toxic content on their communities, instead of relying on restricted words and user flagging in order to identify comments for moderation.

Open Web

Moderating comments in real-time to improve conversations online.

OpenWeb’s moderation platform has been using Jigsaw’s Perspective API for a little over three years but they recently launched a “Real-Time Feedback” feature with the API at its core. The feature moderates comments in real-time, using Perspective API to determine whether comments violate community guidelines. The user can either choose to edit their comment and repost, or they can post it anyway, accepting the outcome.

This new feature was launched after an OpenWeb study, conducted as part of a broader Jigsaw research study, found that out of all commenters who edited their language, 44.7% of commenters replaced or removed offensive words, and 7.6% elected to rewrite their comment entirely.

44.7% of commenters replaced or removed offensive words

The New York Times

Ensuring the comment section stays open so communities can thrive

3x the number of articles offering comments

Southeast Missourian

Improving conversations and building community via online comments

96% decrease in toxic comments

El País

Making the comments section less toxic

19% increase in comments

7% decrease in toxicity

Disqus

Reducing toxicity in comment sections across the world.

Coral

Helping commenters improve their behavior online

36% of users changed their comment in response to feedback

FACEIT

Tackling toxicity on gaming sites through machine learning

20% reduction in toxic messaging

Open Web

Moderating comments in real-time to improve conversations online.

44.7% of commenters replaced or removed offensive words

Developers create bots to moderate content in real-time

Moderators can combine APIs to moderate audio

Developers create bots to moderate content in real-time

Allowing content platforms to handle bad user input in real time

Moderators can combine APIs to moderate audio

Quickly categorize and evaluate audio content

“As Reddit scales, the integrity of our platform and ensuring healthy discourse among users and communities remains a priority. Perspective has been a valuable tool as we continue to strengthen the safety measures and tooling that we have in place.”

—Chris Slowe, Chief Technology Officer at Reddit

Case Studies

How world-renowned publishers, platforms, and researchers are using Perspective API

Developers create bots to moderate content in real-time

Moderators can combine APIs to moderate audio

Developers create bots to moderate content in real-time

Moderators can combine APIs to moderate audio

Read more about our partners