Every so often we hear from people who are curious about the technology driving MarketMuse. Where do all those content suggestions and enhancements originate? Is it a secret team of content fairies spreading their pixie dust? Do we use an API? Or is it something more?
If you picked door number three, you’re right!
MarketMuse is powered by an advanced collection of technologies we refer to broadly as artificial intelligence (AI). Some of those technologies involve machine learning, deep learning, knowledge representation, natural language processing, and natural language generation.
How MarketMuse Works
MarketMuse has patented systems and methods for semantic keyword analysis. While data science geeks may enjoy perusing the patent, “How MarketMuse Identifies Topics That Make a Page More Comprehensive” explains it in layman’s terms.
Basically, we use an ensemble of algorithms (that fall within the domain of artificial intelligence) to create a robust topic model for any given subject. That model (ontology) is a representation of a set of objects and their corresponding relationships.
Relevance and importance aren’t determined solely on the basis of frequency. A concept can be mentioned infrequently, yet be extremely relevant and critical to the conversation. That’s the problem with simple algorithms like TF IDF. They cannot capture those complex semantic relationships in the same manner as a topic model.
We don’t just analyze a page of content. We analyze the content of your entire site and its topical coverage in order to determine what content is most likely to succeed.
Soon, we’ll be introducing MarketMuse First Draft. Our natural language generation (NLG) technology creates long-form content without the use of templates. Unlike existing NLG models, that can only produce 300 – 500 words before serious quality degradation, MarketMuse First Draft does not suffer from this restriction.
So when we say we use AI, we really mean it.
How We Continue to Improve Our AI
We train our models on
thousands, make that hundreds of thousands, of pages of text. Our data science team continually strives to enhance the algorithms in order to improve its accuracy and make our suggestions that much more helpful.
Building our platform from the ground up using the latest methodologies in machine learning enables us to offer unique and valuable solutions for content strategists and marketers. We’re constantly looking to turn the latest research surrounding artificial intelligence into practical and useful products.
Why We Take This Approach to Content
Some content optimization tools rely on third-party APIs (application programming interface) for their artificial intelligence implementation. But there’s a problem with that approach. It seriously restricts what these tools can offer and prevents them from improving the core functionality.
MarketMuse took a different approach when we first started in 2013. There simply were no APIs available that could achieve what we envisioned. Starting from scratch has given us more flexibility to provide some very innovative technologies.
Content strategy is hard. It’s not easy to determine the best path to establishing subject authority. Creating quality content at scale is no simple task either. Our goal is to help you determine what you should write about, how you should write it, what topics to use, understand what your competition is doing, how you can be better, and what content is most likely to succeed.