By Afzaal Ahmad Zeeshan, Alibaba Cloud Community Blog author.
For the continuous business growth and customer satisfaction, content is the key factor which plays the most critical yet crucial role. There come a thousand strategies for content creation and monitoring after almost every couple of days, still, the possibilities of adapting and implementing them seem endless. Similarly, another integral component in this realm is content moderation. There should be a mechanism to manage the content privacy and regulations as the threats and ill-intend attempts to destroy either the transparency or the security of the content is never stagnant.
On a broader level, critical businesses such as news reporting and social media advertisements have more corrosive effects in the case of content infringement. Consequently, social media platforms must put strong checks over their advertisement campaigns thoroughly before elections or any other political event. To manage such situations, social platforms must come up with robust and advance systems to filter submitted content and requests, an example is a political advertisement ID system to take down any content which can create trouble and distress in the current situations. On the lower level, take any live session or speech cast either to talk about human rights or religious matters, there can be a thousand of threats and risks associated with the online presence if we exempt content moderation regulations.
So, to be aware of these distressing and detrimental impacts, first, we will discuss the different types of content moderations and then we will have a theoretical grasp and technical insights of Alibaba Content Moderation service which is an exemplary solution to implement almost all the discussed types of moderations in a handiest and cost-effective way.
There are three major types which are must to consider and worry about by any online platform, community or channel content manager while submitting or processing any request coming from end-user.
Pre moderation technique is the most common one the online forums and communities which submit and queue user request until a moderator investigates it and allow to be publicly available. Pre moderation is a commonly used moderation technique, but the availability of moderator is an overhead for the small platforms because if the submitted request kept off from the site for a little longer, the purpose of the request is killed eventually. The content can be of any type, image, text, video, and the purpose of moderation can also vary.
As an alternative, we can consider post-moderation. Post moderation is a viable option when we want to display the submitted content in the real-time without any delay. This way, content gets public the moment it is submitted, but a copy of it is replicated in the system queue to be reviewed later as per the availability of the moderator. Post moderation can be a lot risky if moderator checks will take more than an advisable time span¡ªtypical online forums take this approach, as it lets users submit the content as soon as possible, and let the moderators or sometimes other users flag the content for any unwanted content.
In this type of moderation strategy, the main party to accept or reject the content are the community members themselves. When content is submitted, the front tier moderation checks are performed and then this content becomes available to all users. This way, all the community members participate actively (maintaining reputation and badges is one of the main tractions behind this participation) and report or red flag any content they do not deem to be left published anymore. In my own experience, we participate in this type of moderation every day on the communities, such as CodeProject.
The distributed approach is based more on the members reactions rather than community standards which is risky and can cause biasness in the community. Moreover, this if something inappropriate and humiliating is kept published on the site due to partial or no responses then the entire impression and respect of the group can be disturbed¡ªthis can also lead the entire community to change their motives and policies based on how the majority votes.
Automated moderation uses some tools and scripts to filter out words and images for the banned content. It is a powerful and widely used option to skip publishing the unwanted and undesirable content online. Automated Moderations work based on the policies that are defined by the organization, and they can allow content, or reject the content solely based on the policies without any human bias or interaction. Once the moderation is set in place, it uses machine learning and artificial intelligence to learn how the content publishing strategies are changing and how to adapt.
As per the platform selection, different tools and applications can be used to filter out the words and images before publishing, similarly, the user and IP can also be tracked and saved for the future reference for a time with some messages sent to the request submitter¡ªbased on how the response was to be sent.
Nonetheless, all the discussed ways of moderating content are highly prone to risk and threats. They require human input and time, still seems to be highly unlike to produce desired results timely. We require smart and highly rational service which can take this overhead away and produce custom and highly diversified results as per the business and community demand. This is where automated moderations come in action, which made use of highly capable and expeditious machine learning analysis and patterns at the backend to filter out text bases content and other images and video-based submissions in a cost-efficient way reliably and automatically. Alibaba Cloud offers such a service as Alibaba Cloud Content Moderation Service, which uses policies to filter out unwanted content, especially media content that can contain off-topic or enraging content¡ªranging from political, racy, pornography to terrorist activities.
Alibaba Cloud Content Moderation service follows a subscription-based model. It implements deep learning methodologies which are verified and tested on more than a million-size data sets stored in the Alibaba Cloud's data centers and from the users that actively participate in the product improvement. Alibaba content moderation helps to monitor adult content such as terrorism, violence, illegal information exchange etc., also, it leverages its users from spam advertisement and irritating experience from the online world.
As no human intervention is required, the backend algorithm is highly accurate and fast, it generates results with more than 95 percent accuracy under a second¡ªa couple of milliseconds only. Using advanced machine learning strategies, it accesses billion of images and text records on the daily basis to be a most suitable candidate for such services from small-scale businesses to giant clients and continues to learn from the input data.
To meet customized requirements based on user preferences, the content moderation API comes up with potent systematic methods that detects adult, violent and any inappropriate content from text, images and videos which you do not want to allow getting visible on your site. It also provides features such as tagging, moderation, fixing or replacing banned words from your submissions to facilitate user requests as per the individual requirements.
Images and video are the major source of transferring information to ensure content authenticity and outreach. Images and videos are more engaging, on the daily basis, more than a billion images and video get online region wise. Almost more than 50% of them is spam and subjected to violate content policies described by higher regulations. To detect the ill-suited element from this content Alibaba content moderation API scans this media thoroughly and take the enlisted actions on the go.
Undoubtedly, Alibaba Cloud is the helping hand in the content security realm. It provides a range of diversified options to ensure the protection of content policies at its highest possibility. The custom image library functions offer you to blacklist and whitelist IPs for the future ease and record.
Oh, and Alibaba Cloud themselves have a went through a couple of internationally recognized licenses and certificates to ensure your data's privacy.
Content moderation service supports a wide range of widely used programming languages; Java, PHP (v5.3 and above) and Python (v2.7 and above). However, in case, if these languages cannot meet the user requirements then you are provided with an option called Call Methods to use any third-party provided API, you can call methods from their API and use them on the base line of content moderation service to get advantage of other functionality¡ªwe discussed such services in the other article of mine where we explored the Alibaba Cloud Open API Explorer, do check that out.
You can read this part of the documentation to learn how to make the calls from Java, Python or PHP SDKs.
To process HTTP requests, information such as domain name, URL query parameters, header format and type of returned code is descriptively available. The given API to call the API functions is designed sophisticatedly to give handy user experience. Both the asynchronous and synchronous interfaces are used for image filtration. It scans images within a few seconds with the longest response time of 6 seconds in case of synchronous response.
HTTP header contains all the nitty-gritty information required to pre-process the coming request. These fields include accepting response type in JSON, specified content type, date and authorization signature, etc.
HTTP query parameters include client info; serialized from client table in JSON, SDK version, user-type and other details such as operator device code and IP, etc. One can create its own HTTP request to call the content moderation API, for this encryption signature and returned value should be dealt with considerably.
It instantly detects logo, patterns and targeted text from images and videos such as OCR to detect text and. Filters adult and aggressive content, spam and savage material either in real-time such as live streaming or video broadcasting or from stored submissions.
Actively finds out inappropriate requests within a few seconds from a wide range of image and video submissions. Give optional visibility of the violent and distressing content based on the user settings and preference.
600 transactions per minute and scanning of about 50 images per second is the default setting, you can request to customize this limit as per your need.
Content moderation supports two types of processing; synchronous and asynchronous. Processing images synchronously supports one image at a time while in the asynchronous model you can process multiple items using any scheduling algorithm such as round-robin.
Note that the billing cycle of Content Moderation API is 24 hours. Alibaba Cloud will send you to bill the next day according to the service volume of today. Then you would check the bill from your Alibaba Cloud account. The payment time is 8 to 10 hours after the billing cycle ends.
Alibaba Clouder - November 19, 2018
Alibaba Cloud Community - December 23, 2021
Alibaba Clouder - January 4, 2021
Alibaba Cloud MaxCompute - March 24, 2021
ClouderLouder - November 25, 2020
Alibaba Clouder - September 2, 2019
An end-to-end platform that provides various machine learning algorithms to meet your data mining and analysis requirements.Learn More
This technology can be used to predict the spread of COVID-19 and help decision makers evaluate the impact of various prevention and control measures on the development of the epidemic.Learn More
A unified, efficient, and secure platform that provides cloud-based O&M, access control, and operation audit.Learn More
Accelerate static and dynamic web content in a fast, reliable, and safe way using Secure DCDN (Dynamic Route for CDN)Learn More
More Posts by afzaalvirgoboy