
You get to see current lyrics statistics, if your search does not return any results (e.g. 'djidjibiji...'). It is good to know that people sing more frequently about love than hate. This incidently corroborates the observation of Carnegie Melon's Prof. Lorrie Cranor who studied password distributions. Love permeates people's privacy and people's publicness.
I own almost an entire string section consisting of three classical acoustic guitars, one electric guitar and a tambura. I enjoy playing them all, and I do not admit three classical guitars is a little overboard!
In that sense, I am a hobbyist musician. Accordingly, I like listening to a lot of different types of music and songs.
Sometimes, I cannot understand/hear the words in a song; or, I know the words, but I can't remember the song's name and author; or, I know the author and song, but don't know the album, or... well you know what I mean.
Often I found myself on one of those many lyrics sites complete with ads, popups, redirects, comments sections, album covers, MIDI files... I thought these sites were annoying. So, I decided to make my own :)
Project Mirlix is the result of that effort.
Mirlix currently searches over 800,000 songs and updates its lists once per year (next update is scheduled for Nov. 2014).
In its first iterations, Mirlix relied on a very fast key-value index of the documents (lyrics, albums, artists, etc.). The search results were ranked based on a simple tf-idf algorithm.
In its current iteration, Mirlix is integrated with the Sphinx Search API. I felt this provides more flexibility and extensibility to the project. Although slightly more complicated, the ranking algorithm is still a form of tf-idf.
Mirlix provides a karaoke feature, too, although, still in development. Find a song, click on the karaoke link and edit your video. NOTE: currently there is no transcoding enabled! You cannot save or export the video you see. This of course will change in the next months. Stay tuned!
There are a few new features to be added to Mirlix as well: multi-language lyrics translation, collaborative social collage images illustrating lyrics texts, and more!
Mirlix does not generate any revenue. It is commercial- and ad-free. Hence, ideally it should also cost me nothing. Therefore, it runs on a tiny, free Amazon EC2 instance. This is not an issue since the code is rather efficient and the Mirlix engine and interface can serve fast many, many requests simultaneously.