The platform of Keepcon’s latest generation helps automate the identification process, filtering and moderating User Generated Content that does not meet your company’s standards.
The employment of our platform allows the moderation service to better achieve the following:
1) To progressively automate the moderation process while eliminating human moderation, therefore accelerating the over-all process and reducing costs.
2) To guarantee the quality and effectiveness of the manual moderation. At the same time, we take full advantage of the information generated by the automatic process, which has been condensed in the moderator GUI across a board of indicators. The board suggests detailed revisions and warns you of the level of risk in each content revised.
Our moderation workflow allows us to model more adequate circuits for the type of control and content you need moderated.
From our extranet, our clients are able to obtain reports in real time in order to control the continuation/advancing of the moderation process. Also, our clients can adapt business rules and regulations across respective formulas that feed the automatic processors. But, fundamentally, they can adapt the thresholds of automatic moderation to determine which content should be manually revised and which should not.
The platform relies on a model of pre-processing or semi-automatic moderation, based on the moderation engines that feed from defined rules by our clients and our base of knowledge.
The distinct engines of artificial intelligence recreate the rules and learn only those with the purpose of determining content that can be moderated automatically (SPAM, pornography) as well as with the purpose of enriching the content and adding useful information at the time of manual revision.
The entire pre-processing model is flexible and can be altered. Our development team is available to implement new rules or to add to those that exist in order to improve the moderation quality. But, above all, our pre-processing model is created to progressively achieve the automation process and the consequential cost decrease, while accelerating the overall moderation process.
Plugins of Automatic Moderation:
Dictionary of Languages: to automatically determine the level of orthographic accuracy and the written text of the content.
List of Prohibited Words: Prohibited words are fed to a database, in order to be detected in the analyzed content. The list is constantly adjusted and added to by our clients as well as the moderation team.
Regular Expression Engines: determines patterns of character chains identified inside the analyzed content, such as: contact information, email address, telephone number, URL, etc.
Automatic moderation of images and videos (image mining/image filter): automatically determines if the image has drawing, graphics from sites (publicity), pornography and technical drawings. A score is given to the image or video that allows the content to be automatically moderated or categorized for the human moderation board of indicators.
History of User / Blacklist and Whitelist: We save the moderation history of each user, his or her IP and/or username to validate the past behavior and to better understand the profile of each content creator. In this way, we maintain a black list to automatically filter the contents according to its author. We also have a white list that positively categorized the content from certain authors
Control of Spam and Phishing: Automatic determination of the probability that the content has spam or a phishing intent.
AdHoc User Rules: Our development team is available to develop particular rules that each client wishes to implement.
After the contents are pre-processed, our Automatic Moderation model, depending on thresholds, decides which content does not need to be manually moderated and which should be inserted to the manual moderation workflow. It is this module decider that evaluates which flow the contents should follow, and our clients would be able to adjust the depending thresholds of each plugin with the intention of automating the decision. For example, certain thresholds can be adjusted depending on the number of prohibited words found. The platform will then automatically moderate the content or send it to be manually moderated depending on the adjustment of the different agent thresholds.
I+D:
Automatic Opinion Mining: we are researching and developing this discipline with the aim to progressively automate the inference process of free text to implicate the opinions and feelings of the contents’ authors.
Web Semantic: we are researching on how to improve our automatic moderation process with the semantic input that allows the moderation plugins to be more sensitive and intelligent. Also, the tags that we add to the content contain semantic information bound with the process of automatic Opinion Mining.
