I am trying to use Wolfram alpha to tag important phrases like movie names, actor names, director names, etc. from a text input. For example, for the input query: "Play some rock by Metallica" (http://www.wolframalpha.com/input/?i=play+some+rock+by+metallica), it interprets 'rock' as a 'Mineral'.
Is there a way to set a parameter that tells wolfram alpha that the query is "music related" and that it should only confine its search in that domain?
Are you just using the Wolfram|Alpha website?
If you're just using the Wolfram|Alpha website, then the short answer is that there isn't a way to do what you're describing. Wolfram|Alpha can do many things, but it can't play rock music by certain bands. It won't recognize what the query means. Take a look at the example page for music to see what kinds of queries Wolfram|Alpha can return information for: http://www.wolframalpha.com/examples/Music.html
If you're writing a program which needs to undesrtand that "Rock" in this instance is a a musical genre rather than as a natural feature , then what you're basically describing is called Named Entity Recognition.
Thanks for your response, Sean. I am using Wolfram Alpha Pro Web API (that responds with an XML response).
Also, I am not asking it to "Play" any multimedia content, I am just trying to tag certain phrases in the input text as song name, band name, etc. Yes, you are absolutely correct that this is called NER problem in conventional literature.
For example, if I give an inpul "Play lithium", I want wolfram alpha to confine its NER knowledge to "Music" domain and suggest that "Lithium" is a song, and not a chemical element. Are there any filters that I could apply while running my queries to confine its domain to music?