Generating audio annotations for search and retrieval

Abstract

Embodiments of a computer system to determine one or more annotation items associated with an audio file are described. During operation, the computer system provides an interactive environment in which multiple users listen to the audio file within a time interval. Next, the computer system receives one or more annotation items associated with the audio file from the multiple users. Then, the computer system displays the received one or more annotation items from the multiple users in the interactive environment, thereby enabling the multiple users to provide feedback to a given user in the multiple users.

1 - 20 . (canceled) 
     
     
         21 . A computer-implemented method for providing an audio recommendation, comprising:
 receiving a search query that includes information corresponding to audio content;   generating a search expression based on the search query, wherein generating the search expression involves determining a subset of annotation items that characterize the audio content, and wherein the generating is based on a supervised-learning model that specifies statistical relationships between audio files and the annotation items, which characterizes the audio files using the annotation items;   using the computer, calculating match scores between the search expression and the audio files based on the supervised-learning model, wherein a given match score indicates a statistical relationship between the search expression and a given audio file in the audio files;   selecting a subset of the audio files based on the match scores; and   providing the audio recommendation based on the selected subset.   
     
     
         22 . The method of  claim 21 , wherein the information includes a semantic description. 
     
     
         23 . The method of  claim 21 , wherein the search expression includes a query multinomial for the subset of the annotation items; and
 wherein a contribution to the query multinomial for a given annotation item in the subset of the annotation items is inversely proportional to a number of terms in the search query.   
     
     
         24 . The method of  claim 23 , wherein calculating the match scores involves calculating distances between the query multinomial and the audio files. 
     
     
         25 . The method of  claim 21 , wherein the search query includes the audio content. 
     
     
         26 . The method of  claim 25 , wherein determining the subset of annotation items involves annotating the audio content. 
     
     
         27 . The method of  claim 21 , wherein the audio recommendation includes a song. 
     
     
         28 . The method of  claim 21 , wherein the search query includes personal information of a user;
 wherein the personal information includes one or more of: age, gender, geographic location, nationality, cultural background, education, profession, income and musical tastes; and   wherein the subset of the audio files is selected based on the personal information.   
     
     
         29 . The method of  claim 28 , wherein the supervised-learning model further specifies the statistical relationships between the audio files and additional personal information of at least a group of users. 
     
     
         30 . The method of  claim 29 , wherein the personal information of the user matches at least some of the additional personal information of at least the group of users. 
     
     
         31 . A computer program product for use in conjunction with a computer system, the computer program product comprising a non-transitory computer-readable storage medium and a computer-program mechanism embedded therein to provide an audio recommendation, the computer-program mechanism including:
 instructions for receiving a search query that includes information corresponding to audio content;   instructions for generating a search expression based on the search query, wherein generating the search expression involves determining a subset of annotation items that characterize the audio content, and wherein the generating is based on a supervised-learning model that specifies statistical relationships between audio files and the annotation items, which characterizes the audio files using the annotation items;   instructions for calculating match scores between the search expression and the audio files based on the supervised-learning model, wherein a given match score indicates a statistical relationship between the search expression and a given audio file in the audio files;   instructions for selecting a subset of the audio files based on the match scores; and   instructions for providing the audio recommendation based on the selected subset.   
     
     
         32 . The computer program product of  claim 31 , wherein the information includes a semantic description. 
     
     
         33 . The computer program product of  claim 31 , wherein the search expression includes a query multinomial for the subset of the annotation items; and
 wherein a contribution to the query multinomial for a given annotation item in the subset of the annotation items is inversely proportional to a number of terms in the search query.   
     
     
         34 . The computer program product of  claim 33 , wherein calculating the match scores involves calculating distances between the query multinomial and the audio files. 
     
     
         35 . The computer program product of  claim 31 , wherein the search query includes the audio content. 
     
     
         36 . The computer program product of  claim 35 , wherein determining the subset of annotation items involves annotating the audio content. 
     
     
         37 . The computer program product of  claim 31 , wherein the audio recommendation includes a song. 
     
     
         38 . The computer program product of  claim 31 , wherein the search query includes personal information of a user;
 wherein the personal information includes one or more of: age, gender, geographic location, nationality, cultural background, education, profession, income and musical tastes; and   wherein the subset of the audio files is selected based on the personal information.   
     
     
         39 . The computer program product of  claim 38 , wherein the supervised-learning model further specifies the statistical relationships between the audio files and additional personal information of at least a group of users; and
 wherein the personal information of the user matches at least some of the additional personal information of at least the group of users.   
     
     
         40 . A computer system, comprising:
 a processor;   memory; and   a program module, wherein the program module is stored in the memory and configured to be executed by the processor to provide an audio recommendation, the program module including:
 instructions for receiving a search query that includes information corresponding to audio content; 
 instructions for generating a search expression based on the search query, wherein generating the search expression involves determining a subset of annotation items that characterize the audio content, and wherein the generating is based on a supervised-learning model that specifies statistical relationships between audio files and the annotation items, which characterizes the audio files using the annotation items; 
 instructions for calculating match scores between the search expression and the audio files based on the supervised-learning model, wherein a given match score indicates a statistical relationship between the search expression and a given audio file in the audio files; 
 instructions for selecting a subset of the audio files based on the match scores; and 
 instructions for providing the audio recommendation based on the selected subset.