A Preference Extraction Method Using Installed Applications and Their Descriptions

Abstract

As personalization technologies are widely used, preference extraction is becoming important. In this work, we propose a preference extraction method on the basis of applications that are installed on a user's smart device. In this method, keywords are extracted from descriptions of the installed applications on an application market. Then, each keyword is scored by three scoring criteria: degree of characteristic of the keyword in the description, scarcity and usage time of the application. As a result, the user's preference is expressed in the form of a list of pairs of a word and its importance score. From the experimental result, we confirmed that the installed applications are useful as an information source for user preference extraction. We also examined the effectiveness of each of the scoring criteria. From the result, we found that the criterion that represents scarcity of the application is effective for users installing a large number of applications. On the other hand, the criterion that represents usage time of the application tends to be effective for users using their device for a long time.

Predicting Next-use Mobile Apps Using App Semantic Representations

“…We collected an app usage dataset using our developed Android app called "context monster" [25]. Our developed app observes app usages, i.e., app launchings, and records application IDs as well as timestamps of app usage to be transmitted to a server computer.…”