Study Of Algorithms To Rescognise The Strength Of Users’ Video Query Intention

With the explosive growth of data, there are more and more difficulties while choosing the useful information. The search engine plays an important role for users to obtain information and with the popularity of smart devices the mobile search is becoming more and more important. The smart devices have limit space to show information and as a result we need to return to users the most useful information he/she wants, so we need to identify the users’ query intent clearly. However, actually the users prefer to offer short string, which has 3-4 terms commonly, to the search engine and the string may have more than one meanings which makes it difficult to identify the users’intent clearly. In this paper, we analysis and solve the problem of recognising the strength of video intention based on abundant data sources and user interactions in search engine. The system is applied to general search engine and video retrieval system. We can identify the strength of video intent by analyzing the users’ query string so that we can show the results to users in friendly way.In this paper, firstly, we expand the query string with the results of search engine as well as the users’ clicks to obtain a longer string and at the same time we propose a new text feature selection method based on entropy and word frequency according to the characteristics of high contact ratio between these categories. Next, we design and extract 5 groups of features and different ways to combine these features. The features we designed are:text features and statistics of video hosts and the types of results returned by search engine and semantic features based on deep langrage model and statistics of session. Next, inspired by word2vec, a deep langrage model, we propose a new way to represent the hosts with a vector, called Host2vec, with which we use the deep langrage model to solve the problem of identify the video intention of query string offered by users. Last, we try to analysis the relationships between video intention of query strings and time.