Genius Accuses Google of Lifting Its Lyrics Transcriptions in Knowledge Panels wsj.com

Robert McMillan, Wall Street Journal (paywalled, but Engadget has a good summary):

“Over the last two years, we’ve shown Google irrefutable evidence again and again that they are displaying lyrics copied from Genius,” said Ben Gross, Genius’s chief strategy officer, in an email message. The company said it used a watermarking system in its lyrics that embedded patterns in the formatting of apostrophes. Genius said it found more than 100 examples of songs on Google that came from its site.

Starting around 2016, Genius said, the company made a subtle change to some of the songs on its website, alternating the lyrics’ apostrophes between straight and curly single-quote marks in exactly the same sequence for every song.

When the two types of apostrophes were converted to the dots and dashes used in Morse code, they spelled out the words “Red Handed.”

Google previously scraped material wholesale from Celebrity Net Worth, Yelp, TripAdvisor, and Amazon. The Knowledge Panels and Featured Snippets features are unreliable, and they power a huge scale of Business Insider-style aggregation that discourages reading the source. Google denies any wrongdoing here and have passed the buck to unidentified partners, but it’s disgraceful that they either cannot stop scraping from smaller businesses, or they cannot be bothered to verify that materials from partners are reliable and original.

Incidentally, the case of Genius is interesting because song lyrics are not exactly their intellectual property — they are the property of the songwriters, and are published under license. However, there is no centralized lyrics database to which artists can publish; therefore, the lyrics that are shown on Genius are a best guess original transcription, some of which are verified by artists. The lack of a canonical lyrics database is also why lyrics shown in, say, Apple Music may be different than lyrics shown on a lyrics website or in the booklet in the physical copy.