Personalized Query Auto Completion Using Social Networks? Login

摘要

Query auto completion in search engines aim to suggest the most accurate queries matching the few characters entered by the user. Present search trends and query logs are the commonly used criteria. This can be widely improved by incorporating user specific details from social network profiles. Most of the Social Networks allow access to the user info via apps that use their API. For example, Facebook offers user info by default which includes their name, gender, location and other details. This can be further extended to even more details regarding user activity on the social network. Such info can lead to a much more personalized and user- centric auto completion experience. We demonstrate the user details that can be of great value for QAC and the ways of getting these details from the most popular social networks out there. Later we provide a working model consisting of AOL query log, trending topics and Facebook likes. 1. INTRODUCTION Query auto completion is one of the most visible features in Web Search today. It is offered by all major search engines and in almost all their search boxes. Query auto completion helps the user formulate her query, while she is typing it. Its main purpose is to predict the user’s intended query and thereby save her keystrokes. With the advent of instant as- you-type search results (a la the Google Instant), the importance of correct query prediction is even more acute, because it determines the speed at which the user sees the suitable results for her intended search and the amount of irrelevant results that are displayed to her along the way. The basic principle that underlies most query auto completion systems is the wisdom of the crowds. The search engine suggests to the user the completions that have been most popular among users in the past (we call this algorithm MostPopularCompletion). For example, for the prefix am, Bing suggests amazon and american express as the top completions, because these have been the most popular queries starting with am. As the user is typing more characters, the space of possible completions narrows down, and thus the prediction probability increases. Clearly, during the first few keystrokes the user is typing, the search engine has little information about her real intent, and thus the suggested completions are likely to wrongly predict her query. Social networks have become a vital and indispensable part of the life of the present generation. Most of the youngsters are addicted to sites like Facebook, Instagram etc. With the increasing use of social networks, their importance and data value is also increasing and so is the amount of user details available in them. User profiles are more and more like mirrors or rather windows to the user’s personality, likes, interests and preferences. Thus a lot can be concluded about a user by viewing there social network profile. This can be of great help for applications that aim to personalize the user experience. Most of the social networks enable developers to access user info via an Application Program Interface (API). An API is a method of accessing the user details on the social network via an “app” of the social network itself. Such APIs mostly work by giving access of user info to an App via a Login button. Once the user logins on the site using their login credentials, she is given an option to deny or allow access to her details to the app. These buttons have an added advantage of easy account creation to the user. So it a win-win situation for both the user and the developer. Once we have access to the user details and activity on the social network, we can extract valuable info including user’s age, gender, location etc. These bits of information go a long way in creating a personalized user experience. These Even more details like the user likes, interests,.