Introducing Curation Rank & Auto Curation
We recently introduced a new feature that accelerates the curation of content through machine learning. We’re excited to show that to you and explain the thought process behind it.
The curation or refinement of news and information is increasingly important and necessary to deal with the overwhelming amounts of content related to the work that we each do. Curation helps us focus on the information we care about and save time in the process. The importance of this will only increase with time.
How and where curation happens is evolving to address the reality that information flows from countless sources in various formats. Traditional manual approaches can not keep up. New methods and enabling technology are needed.
There is also a need to look for help from subject matter experts within organizations that may not currently curate others’ information. One of my favorite examples of this is when product managers curate topics related to their markets. They are already following this information and understand the context. But in most cases, these insights only trickled out to others inconsistently and selfdom in a timely way. Recognizing the opportunity, the product manager began curating Attensa Topics related to their markets. This improved their ability to keep up to date for their own purposes. More importantly, it enabled the resulting topics to be available to others throughout the organization so that the knowledge could be leveraged.
Attensa has three layers of curation, a machine layer, a human+machine layer, and a human+machine learning layer.
The Machine Layer
The Machine layer is the foundation of Attensa Topics. Each Topic has a collection of trusted sources that are refined using text filtering and semantic analysis to match a particular topic or interest. As new information comes through the Topic sources, the content is automatically updated with any relevant new items. There are tens of thousands of Attensa topics using this approach. Once set up, they are easy to modify but require very little maintenance.
The Human+Machine Layer
Human+machine curation enables a curator to review and modify the results before the Topic’s followers are updated. Often this process is performed by information professionals supporting the information needs of various groups inside the organization. We have designed methods for Human+machine curation that are very efficient, and benchmarks have shown gains in curators’ productivity over 50%. The process itself utilizes a visibility feature available in all topics. Individual items can be toggled between show and hide. Hidden items are not visible to the Topics followers. This approach is so efficient because the initial machine layer gathers content from various sources and refines it. The curator needs only to review those results and select the items they want others to see.
The Augmentation Layer (new)
The new human+machine learning curation approach allows a curator to train Attensa to recognize the content they desire for a given topic and uses machine learning to implement that strategy. Once trained, each item in the Topic has a curation rank based on how well it fits the training model. This allows curators to quickly focus on the most relevant items. If desired, the entire process can be automated. A curator can set the desired rank, and Attensa will automatically curate set the visibility of an item based on that ranking. As with each of the curation approaches, a human curator can always modify the results. The full archive of content is archived in the Topic. The video shows this feature in action.
This new curation feature enables curators to work even faster and more accurately - particularly the non-traditional curators who are often adding curation to an already busy day. The three curation layers provide flexibility to adapt the curation process to the Topic, the nature of the audience, and the desired workflow.
If you have any questions or would like more information about curation using Attensa, please contact email@example.com.