Structured natural language descriptions for semantic content retrieval of visual data

用于视觉数据的语义内容检索的结构化自然语言描述。视觉数据的丰富性使为其内容编制索引以进行数据库检索的任务变得复杂。基于语义内容的检索比基于纹理或颜色的检索更自然，但最常见的方法-关键字搜索-缺乏精确度，而较新的方法--自由文本描述的自动解析--缺乏准确性。我们提出了一个元数据和内容描述系统，其中内容描述必须是由基本语法术语组成的短语。描述组件可以是来自诸如同义词表和数据库等外部资源的元素，为通过语义内容对视觉数据进行有意义的检索提供了丰富的上层建筑。

Structured natural language descriptions for semantic content retrieval of visual data. The richness of visual data complicates the task of indexing their content for database retrieval. Retrieval based on semantic content is more natural than retrieval based on textures or colours but the most common approach - keyword search - lacks precision, while a newer approach - automatic parsing of free-text descriptions - lacks accuracy. We propose a system of metadata and content descriptions, where content descriptions must be phrases composed of basic grammatical terms. Description components may be elements from external resources such as thesauri and databases, providing a rich superstructure for meaningful retrieval of visual data by semantic content.