Apple has been granted a patent (number 11573998 B2) for a “lyric search service” that shows the company wants Apple Music to be better able to locate and display song lyrics.
About the patent
As the tech giant notes in the patent, search algorithms have been developed that are uniquely tailored for querying large datasets based on text-based data contained within a data repository. Text-based search fails when the dataset is represented in another format. For example, audio files are typically searched based on metadata associated with the audio file (e.g., filename, artist, song title, etc.), but the search fails to compare a query string to audio waveforms within the audio file.
In some cases, specialized services have been developed that can compare a query based on an audio snippet with a repository of audio files, searching for a match of the audio snippet to the audio waveforms within the audio files. However, Apple says that these services cannot match a text-based query string with the content contained within the audio waveform in the audio files. Consequently, a user can’t easily search for a song based on the lyrics of the song.
Apple says that the best that a user can typically do is perform a text-based search on the Internet and hope that the query matches a transcription of the lyrics in a song that is published on a website. Once the user has the information about the song, they can then perform another search to locate an audio file of the song within a data repository maintained by a music sharing service.
Apple wants to equip Apple Music with even better lyric finding capabilities.
Summary of the patent
Here’s Apple’s abstract of the patent: “This application relates to a client-server architecture that enables search queries to be applied to transcription information for multimedia files. A server device implements a service configured to query a search platform to retrieve results associated with a plurality of multimedia files stored in a content database. The results are ordered according to a plurality of heuristic values calculated based on a text relevance analysis.
“The service is configured to modify the heuristic values to adjust an order of the results, and generate a response to a search request that includes a representation of at least a portion of the transcription information of the multimedia files referenced by the results. The heuristic values are modified based on at least one of a popularity score for a corresponding multimedia file, a weight associated with a particular field, or a relevance score based on feedback signals.”
