Transliteration For Query Expansion

Abstract

Methods, systems, and apparatus, including computer program products, for identifying candidate synonyms of transliterated terms for query expansion. In one aspect, a method includes identifying multiple transliterated terms in a target language. For each transliterated term of the multiple transliterated terms in the target language, the transliterated term is mapped to one or more terms in a source language. For a first transliterated term of the multiple transliterated terms in the target language, one or more second transliterated terms of the multiple transliterated terms in the target language are identified as candidate synonyms of the first transliterated term, where each of the one or more second transliterated terms is mapped to at least one term in the source language that is also mapped from the first transliterated term.

What is claimed is: 
     
         1 . A computer-implemented method comprising:
 identifying, using one or more computers, a plurality of transliterated terms in a target language, each of the transliterated terms representing a conversion in the target language of one or more respective source terms in a source language;   for each transliterated term of the plurality of transliterated terms in the target language, mapping the transliterated term to one or more of the source terms in the source language;   determining that a first transliterated term of the transliterated terms is mapped to one or more terms of the source terms in the source language and that one or more second transliterated terms of the transliterated terms are also mapped to the one or more terms; and   identifying, based on the one or more terms being mapped to the first transliterated term and being mapped to the one or more second transliterated terms, the one or more second transliterated terms as candidate synonyms of the first transliterated term.   
     
     
         2 . The method of  claim 1 , where mapping the transliterated term to one or more source terms in the source language further comprises:
 transliterating the transliterated term in the target language to the one or more source terms in the source language.   
     
     
         3 . The method of  claim 2 , where each of the one or more second transliterated terms identified as candidate synonyms of the first transliterated term has a confidence value with respect to the first transliterated term that satisfies a specified threshold. 
     
     
         4 . The method of  claim 3 , where the confidence value of a second transliterated term of the one or more second transliterated terms is a function of a number of the one or more terms in the source language that are mapped from both the first transliterated term and the second transliterated term. 
     
     
         5 . The method of  claim 3 , where transliterating the transliterated term in the target language to the one or more source terms in the source language further comprises:
 generating a transliteration score for each transliteration of the transliterated term in the target language to a respective source term of the one or more source terms.   
     
     
         6 . The method of  claim 5 , where the confidence value of a second transliterated term of the one or more second transliterated terms is a function of one or more of a probability of occurrence of the second transliterated term in web resources, the transliteration scores for the transliteration of the second transliterated term to the one or more same terms that are also mapped from the first transliterated term, and the transliteration scores for the transliteration of the first transliterated term to the one or more same terms. 
     
     
         7 . The method of  claim 1 , where identifying the plurality of transliterated terms in the target language further comprises:
 identifying terms containing only characters of the target language.   
     
     
         8 . The method of  claim 7 , further comprising:
 computing a statistic for each identified term containing only characters of the target language;   comparing the statistic for each said identified term to a specified threshold; and   including a particular said identified term in the plurality of transliterated terms in the target language if the statistic for the particular identified term satisfies the specified threshold.   
     
     
         9 . The method of  claim 1 , further comprising:
 for the first transliterated term of the plurality of transliterated terms in the target language, identifying the one or more same terms in the source language that are mapped from the first transliterated term and from the one or more second transliterated terms as a candidate synonym of the first transliterated term.   
     
     
         10 . The method of  claim 1 , further comprising:
 receiving a query including the first transliterated term;   expanding the query with one or more of the candidate synonyms of the first transliterated term;   providing the expanded query to a search engine; and   receiving search results for the expanded query.   
     
     
         11 . The method of  claim 1 , further comprising:
 receiving a query including the first transliterated term; and   providing one or more expanded queries for selection by a user, each expanded query including the query and one or more of the candidate synonyms of the first transliterated term.   
     
     
         12 . The method of  claim 1 , further comprising:
 receiving a query including the first transliterated term;   providing the query to a search engine, where the search engine identifies as a possible search result for the query a web resource that includes at least one of the candidate synonyms of the first transliterated term but does not include any term in the query; and   modifying a score associated with the web resource, the score for use in ranking possible search results for the query.   
     
     
         13 . The method of  claim 1 , further comprising:
 receiving a query including the first transliterated term;   providing the query to a search engine, where the search engine identifies as a possible search result for the query a web resource that includes at least one of the terms in the source language that is mapped from the first transliterated term and from the one or more second transliterated terms but does not include any term in the query; and   modifying an information retrieval score associated with the web resource, the information retrieval score for use in ranking possible search results for the query.   
     
     
         14 . A system comprising:
 one or more computers configured to perform operations including:   identifying, using one or more computers, a plurality of transliterated terms in a target language, each of the transliterated terms representing a conversion in the target language of one or more respective source terms in a source language;   for each transliterated term of the plurality of transliterated terms in the target language, mapping the transliterated term to one or more of the source terms in the source language;   determining that a first transliterated term of the transliterated terms is mapped to one or more terms of the source terms in the source language and that one or more second transliterated terms of the transliterated terms are also mapped to the one or more terms; and   identifying, based on the one or more terms being mapped to the first transliterated term and being mapped to the one or more second transliterated terms, the one or more second transliterated terms as candidate synonyms of the first transliterated term.   
     
     
         15 . The system of  claim 14 , where mapping the transliterated term to one or more source terms in the source language further comprises:
 transliterating the transliterated term in the target language to the one or more source terms in the source language.   
     
     
         16 . The system of  claim 15 , where each of the one or more second transliterated terms identified as candidate synonyms of the first transliterated term has a confidence value with respect to the first transliterated term that satisfies a specified threshold. 
     
     
         17 . The system of  claim 16 , where the confidence value of a second transliterated term of the one or more second transliterated terms is a function of a number of the one or more terms in the source language that are mapped from both the first transliterated term and the second transliterated term. 
     
     
         18 . The system of  claim 16 , where transliterating the transliterated term in the target language to the one or more source terms in the source language further comprises:
 generating a transliteration score for each transliteration of the transliterated term in the target language to a respective source term of the one or more source terms.   
     
     
         19 . The system of  claim 18 , where the confidence value of a second transliterated term of the one or more second transliterated terms is a function of one or more of a probability of occurrence of the second transliterated term in web resources, the transliteration scores for the transliteration of the second transliterated term to the one or more same terms that are also mapped from the first transliterated term, and the transliteration scores for the transliteration of the first transliterated term to the one or more same terms. 
     
     
         20 . The system of  claim 14 , where identifying the plurality of transliterated terms in the target language further comprises:
 identifying terms containing only characters of the target language.   
     
     
         21 . The system of  claim 20 , where the one or more computers are further configured to perform operations including:
 computing a statistic for each identified term containing only characters of the target language;   comparing the statistic for each said identified term to a specified threshold; and   including a particular said identified term in the plurality of transliterated terms in the target language if the statistic for the particular identified term satisfies the specified threshold.   
     
     
         22 . The system of  claim 14 , where the one or more computers are further configured to perform operations including:
 for the first transliterated term of the plurality of transliterated terms in the target language, identifying the one or more same terms in the source language that are mapped from the first transliterated term and from the one or more second transliterated terms as a candidate synonym of the first transliterated term.