A few days ago, we released a new AI-based feature, Content Classification, available within Radarly workspaces and Insight Pages. Previously we had a way to classify content within the platform. However, we completely reinvented the feature to make it stronger and more powerful. We're pretty proud of the result. Because of this, we wanted to share more about it and why it’s so beneficial for our Linkfluence clients.
Let’s dive into it.
Our new content classification classifies content between a predefined set of categories. It is available in Radarly and the Insight Pages through different widgets like the Topic Wheel (which is a Sunburst Visualization), Barcharts and Topic Clouds.
Why is this new categorization better than the previous one?
We now use an ontology of categories inspired by the IAB Content Taxonomy which is the standard of marketing and online advertising that all major providers in the sector currently use. The feature is more adapted to social media than our current categorization - previously only using an ontology that came from the press. This is much better to address our clients' current business use cases.
Our new Content Classification classifies content between a predefined set of 370 categories belonging to a 3-level ontology. Before, we classified content in only a hundred categories at a 2-level granularity. As a result, we are now able to go deeper and be more granular when classifying content.
Another great part of the feature is that documents can now be assigned to several categories. Our current process to categorize content obligated us to detect an entity (brand, organization, …) to categorize a document. This is no longer the case. Content classification also allows us to classify short posts like tweets and Instagram posts rather than before, we were not able to classify this kind of documents.
The feature covers 95% of English content and pretty soon it will classify content in French, German, Spanish, Portuguese, Sweden and Italian. All those improvements allow use to get more than 10x documents classified.
As you can probably gather, thanks to this awesome feature, we will be able to bring our clients a much more accurate and deeper content classification.
Why is this good news for you?
An easier exploration mode
If you use Radarly to explore what is being said around your brand or specific topics, this feature will help a lot. Here are some high lights of how the feature can help:
- Brand monitoring: content classification will help you understand more clearly and precisely what topics are associated with your brand (or your competitors’ ones).
- Topic monitoring: if you need to gain a deeper understanding of a specific topic, you’ll be able to look at this topic with more depth, and understand what are the related topics, which brands are trending, be able to follow the evolution, and so on.
- Content Inspiration: another interesting use of this feature is to explore the topics of interests for a specific audience. If you set up a specific corpus of people that match your personas, you’ll be able to understand what their topics of interests are and understand what the related topics are in order to give you inspiration for your next pieces of content.
- Trend monitoring: Using content classification, you’ll be able to listen to a category directly in Radarly, without biasing the results with your query.
Getting more accurate boolean queries
In addition to making your life easier to the specific use cases we mention, this feature will also be a huge help to create more accurate boolean queries.
Today, when you want to explore a specific trend, you have to build a very complex query to allow you to capture all the posts around a specific trend without forgetting a brand that leads the conversation or a similar expression to the one you want to cover to make sure your results will be accurate.
If you try today, you’ll see that it’s hard to build the right query as you could:
- Build a too narrow of a query: you might forgot some important items on your query
- Get too noisy of results: if you build too large of a query
With content classification your life will be much easier as you’ll just have to filter from our list of 400 categories.
This way, you’ll be sure to get more accurate results without the need to get your hands on complex boolean queries and to retrieve all the interesting posts about the topic you want to explore.
Moreover, the contextualization could be hard to be translated into queries as, sometimes, you need to exclude a lot of terms or include other ones so that the machine can understand what you mean. With the content classification you just need to specify the category you want to follow to make sure the system will retrieve the right posts for you. For instance, if you are working for Jimmy Choo, and you want to follow what’s being said about your brand, no need to exclude a lot of keywords anymore to make sure the posts retrieved will talk about Fashion.
We hope you’ll be as excited as we are about this new feature and that you’ll be able to get insightful results playing with it! Feel free to contact your account manager if you need more information about it.