Abstract: Generating code is disclosed. A specification of one or more translation patterns is received. The one or more translation patterns are used to generate at least a portion of code associated with a translator. Using the one or more translation patterns to generate at least a portion of code associated with the translator results in the translator being configured to create a target object model. Creating the target object model includes populating one or more elements of the target object model in a processing order at least in part associated with an order of elements in the one or more translation patterns.

Abstract: Content encoding rendering is disclosed. An indication of a desired interpretation of a starting content having a first encoding is received. The starting content is processed to generate a representation data comprising a second encoding of the starting content, wherein the second encoding is determined automatically and at least in part by the desired interpretation. The representation data is rendered using a view associated with the desired interpretation.

Abstract: Determining a semantic relationship is disclosed. Source content is received. Cluster analysis is performed at least in part by using at least a portion of the source content. At least a portion of a result of the cluster analysis is used to determine the semantic relationship between two or more content elements comprising the source content.

Abstract: Determining a style attribute is disclosed. An indication of a new object to be added to a content document associated with a template data is received. The style attribute for the new object is determined based at least in part on the template data.