Cross-media retrieval through collaborative use of image/audio signals and text data

我们实现了一种向用户推荐所需内容的有效方法。该方法接受图像、视频或音乐片段作为查询，并且基于搜索理论，可以处理对每个图像、视频或音乐片段唯一的模糊用户请求。具体来说，为了实现这种方法，我们将搜索结果缩小到图像、视频和音频三种媒体类型，并实现跨媒体搜索，即使用户无法使用关键字或同一类型的媒体表达相同的查询，也可以获得所需的内容。

We implemented an effective method for recommending the desired contents to a user. The method accepts an image, video or music clip as a query, and is based on search theory that can deal with ambiguous user requests that are unique to each image, video or music clip. Specifically, in order to implement this method, we narrow down the search results to the three media types of image, video and audio, and implement cross-media search, allowing the user to acquire the desired contents even in cases where they cannot express the same query using keywords or media of the identical type.