A cross-lingual search engine

Principal Investigator: Oren Etzioni

Image search engines, such as Google, are used to locate images on the web by querying words. The images are tagged by the words in their original language, which limits the ability of a searcher to retrieve them using another language. Although images are universal, an English searcher will fail to find images tagged in Chinese, and a Spanish searcher will fail to find images tagged in English. Cross-lingual homonyms cause problems as well. For example, the Hungarian word for tooth is ‘fog’, which makes it challenging for Hungarian searchers to find images of teeth on the Web. Therefore, there is an unmet need for a cross-lingual image search engine to search for images that may have been tagged in various languages.

Dr. Etzioni and his colleagues introduce a novel cross-lingual search engine. Using this software, users will be able to translate and disambiguate their queries before sending them to Google. This software displays words (or phrases) in a plurality of different languages with a translation graph after a lexical translation. It utilizes several machine-readable dictionaries composed into a translation graph that records relationships between words across different languages. A node in the translation graph represents a word or phrase in a particular language, and an edge denotes a wordsense shared between words (and/or phrases) in a pair of languages. For queries in languages that do not have high web presence, this new search engine increases the total number of results 57-fold and it supports queries in over 50 languages.

Applications 

• Online image, video or audio search 

• Online social tagging 

• Online games Advantages 

• Larger reachable database 

• More accurate results 

• Time saving


For more info, contact: Laura Dorsey

  • swap_vertical_circlelibrary_booksReferences (1)
    1. Kobi Reiter, Stephen Soderland, Oren Etzioni , Cross-Lingual Image Search on the Web, Proc. of the 20th International Joint Conference on Artificial Intelligence
  • swap_vertical_circlecloud_downloadDownloads (0)
    Files marked with an asterix (*) can only be downloaded by users that have the appropriate product licence. The license must be active and you must be logged into your account.