Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for identifying languages that are relevant to resources. In one aspect, a method includes selecting in a data processing apparatus a first resource; accessing click data that identifies, for each of a plurality of requests for the first resource, a respective search engine user interface from which the request was received; identifying a search engine user interface language for each of the plurality of requests based on the click data; determining a respective language relevance score for the first resource for each identified search engine user interface language; and selecting one or more languages as being relevant to the first resource based on the language relevance scores.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for identifying languages that are relevant to resources. In one aspect, a method includes selecting in a data processing apparatus a first resource; accessing click data that identifies, for each of a plurality of requests for the first resource, a respective search engine user interface from which the request was received; identifying a search engine user interface language for each of the plurality of requests based on the click data; determining a respective language relevance score for the first resource for each identified search engine user interface language; and selecting one or more languages as being relevant to the first resource based on the language relevance scores.

Abstract: Methods, systems, and apparatus, including computer program products, for identifying languages that are relevant to resource. In an aspect, language features are identified for incoming resource links to a resource and outgoing resource links from the resource. The language features or use by a language classification model to generate language relevance scores. The language relevance scores for each of the incoming resource links and outgoing resource links are used to generate a corresponding relevance measure for each of a plurality of languages. Each relevance measure is a measure of the relevance of the language to the resource.