Keywords are dead – long live keywords!

How we should think about music data, catalogue management and search when everything is a prompt-based AI

Cyanite helps music companies turn their catalogues into their own personal Spotify – providing music libraries with the usability, transparency and functionality users have become accustomed to. 

As the use of prompt-based AI-driven search algorithms becomes more widespread, many AI companies are talking up the end of keywords. Whether that’s in regular searches, or using keywords to search for music, a lot of people are pointing to the advantages that using prompts has for users, including greater freedom of language, and individual contextual understanding. 

At Cyanite we were the first music company to launch a search algorithm that could understand natural language when our users were searching for music, and with this we saw a shift in how the music industry thinks about AI.

But are keywords actually obsolete? Should we be abandoning keywords and allowing AI to execute searches instead?

We don’t think so. In fact, at Cyanite we think there’s a strong argument that keywords empower content teams and streamline global marketing operations by enabling two key mechanisms for navigating music catalogues: generalisation and a unified understanding.

In an environment where catalogues frequently change hands and marketing strategies become less focused on individual national campaigns and more focused on a unified global approach, keywords are more important than ever.

Advanced search algorithms can become black boxes

There are lots of different ways to search for music. Music search algorithms like similarity search or prompt-based searches are designed to simplify the process of finding the right music. These algorithms work by analysing thousands of different attributes of a track and then suggest music based on a complex ‘recipe’ that looks at those attributes to try and find a good match. 

While this can be powerful for individual music searches where a user might have fuzzy definitions of what they want, this can actually diminish the ‘common understanding’ of music.

One person’s ‘sad’ might be another person’s ‘bittersweet’, while ‘upbeat’ and ‘high energy’ could mean the same thing to one person, but two very different things to another. When teams are working together on the release of one track, or are working with an entire catalogue, it’s important that they have a shared understanding and basket of terms that can be used to define and describe the music.

Modern sophisticated algorithms often incorporate a personalisation element, adapting to the user’s individual search behavior. So if someone searches and then selects a track from the results of that search, the algorithm might learn from that and apply that understanding to future searches. 

This means that two people could conduct the same search and get the same results but click on different tracks; if they then conduct another identical search, each person could be shown a different selection of tracks, based on the algorithm having noted their previous preference. This increases the risk of one person’s understanding becoming distanced from that of other people.

The end result of things is that it will become increasingly difficult to reconcile the individual language people use to describe music with broader musical understanding, which will worsen internal team communications.

The importance of keywords and descriptive frameworks

Categorising things helps us humans make sense of the world. Think about nature: we’ve developed a hyper-specific categorisation of organisms from species to genus, to family, to order, to class. We know that a ‘tree’ and a ‘grass’ are not the same things – but share a common root. We can talk about ‘fish’, ‘flat fish’ and ‘Pleuronectes platessa’, better known as ‘Plaice’. 

Keywords help us abstract the world and discover patterns to derive heuristics – a mental framework that gives us greater understanding. This is an exclusive ability of mankind. Music is no exception, albeit the keywords can be less hierarchical.

When we are trying to define music using keywords there are two key categories we can apply: objective keywords and subjective keywords. Objective keywords are fixed, well-defined qualities such as the types of instrument used and whether the music is instrumental or contains vocals, the key, the time signature. Alongside these defined qualities we can apply ‘fuzzier’ qualities using subjective keywords – the genre, mood, function, and so on.

Cyanite.ai – AI For Music Tagging and Similarity Search
AI solutions for the people who pioneer how we search and discover music. Cyanite offers AI music recommendation, tagging and music search.

Objective keywords are useful and very few people would dispute that. The purpose and value of subjective keywords can feel less clear though. Different interpretations – between people, between countries, between generations – and a changing creative landscape, coupled with blurry definitions can cause disconnects and lengthy discussions. 

However, subjective keywords more closely capture the actual essence of musical content and so in many ways are the much more important class of keywords for describing music.

Conversations about music often revolve around which genre a song is, the moods it evokes in people, or activities that fit the song. If you want to listen to happy music – personal preference aside – does it matter if the happy music is played on guitars or synths? Music curation would be pretty much impossible without subjective keywords. How many times have you listened to a playlist called something like “Songs in A minor that feature the trumpet”?

The critical role keywords play in catalogue management + acquisitions

 Almost every day, we hear about another catalogue acquisition. While the big, attention-grabbing deals might only happen once or twice a quarter, it’s important to note that smaller catalogue deals happen much more frequently; and each one of these changes the overall catalogue dynamics of the acquirer.

It goes without saying that the financial top line remains the number one metric used for calculating the multiple that someone is willing to pay to acquire a catalogue. However, with thousands of rights changing hands, having accurate keyword tags for the acquired rights is important to be able to assess the specific value that a target catalogue might represent to your business. You might want to pay a higher price for songs complementary to your existing catalogue rather than buying more of the same thing.

Being able to pick up on trends and assess the health of your catalogue and how future-proofed you are is vital to any good catalogue strategy. Keywords create clarity and transparency for music catalogue, and without keywords you are flying blind.

Keywords harmonise global marketing efforts

 As social media has become a dominant trend, artists have become near omnipresent on socials and we are exposed to their brand almost 24/7. It’s not humanly possible for an artist to personalise their messaging for each and every market. Increasingly, artist brands – and especially those of big artists – are managed at a global level. That means it’s ever more important for teams to be able to coordinate globally and align internal comms efforts across markets.  

Choosing the right keywords to describe music (and sticking with them!) helps to streamline those conversations and help drive the right approaches for an artist’s marketing strategy. The right keywords are a valuable shorthand that provide clarity, help avoid lengthy discussions, lower the risk of disconnects and misunderstandings, and make for faster execution: less time wasted, less potential for error. Who can argue with that?

Personal insights from Cynanite’s development

 At Cyanite, we were pioneers in launching ‘free text’ search for music to our customers around the world – prompt-based searches that let users describe in natural language what they wanted, rather than having to use keyword driven search. And at first, we wondered whether this was going to be the future of music search. Our innovative, prompt-based, natural language searches enabled more intuitive searches, and parameters that felt less restrictive. 

However, the more time we spent in extensive collaboration with our clients, the more clearly an important insight became: we discovered that most users still rely heavily on auto-tagging tools that can apply keywords to music, and they were using these alongside search algorithms. These tools complement each other perfectly. Keywords provide a structured way to start a search, while advanced algorithms can help to refine the results based on more nuanced attributes.

We also learned that starting a music search with an empty field without any filter options can be overwhelming. Users often feel lost without the guiding framework that keywords provide. Imagine walking into a library that has no sections or labels – just shelf after shelf of books. The sheer volume of information is paralysing. 

Keywords: not going away any time soon

The death of keywords is a popular narrative in the age of AI, with natural language and prompting being talked up as more ‘human’ ways of searching. In reality, keywords provide us with the structure and shared language that we need in the music industry to be able to navigate, understand and work with vast music catalogues. They add an additional layer of value on advanced search algorithms and AI-powered prompt-based search, and help create a more balanced and more effective search strategy. And most importantly keywords help us move faster and with less risk of misunderstanding or error.

As AI continues to develop, embracing a hybrid approach ensures that we harness both the full value that  keywords can offer alongside the full potential of AI, meaning that catalogue management and music search becomes better, more powerful, and more efficient.

Keywords are dead – long live keywords!

Cyanite.ai – AI For Music Tagging and Similarity Search
AI solutions for the people who pioneer how we search and discover music. Cyanite offers AI music recommendation, tagging and music search.

From their headquarters in Mannheim, Germany, Cyanite’s team develops AI-powered music analysis and recommendation software that enables effective keywording and efficient music research based on it. This enables music, entertainment and advertising companies to quickly and cost-effectively deliver the right songs for their customers’ search queries based on keywords, reference titles or free text. 

Via API or no-code solution, Cyanite supports some of the most renowned and innovative players in the music, entertainment and advertising industries. These include production music libraries APM Music, Pond5 and Epidemic Sound, music publishers BMG, Nettwerk Music Group, and Schubert Music, and sound branding agencies amp sound branding, Antfood, and Human Worldwide. 

In 2023 Cyanite was awarded the VIA 2023 Award in the Best New Music Business category by Germany’s Association of Independent Musicians and Music Companies. 

Over 100,000 musicians and producers worldwide have already signed up for Cyanite’s free web app to have their music analyzed and tagged by AI.