Estimating confidence for query revision models

A system comprising:

• one or more computers;
• a computer-readable medium coupled to the one or more computers having instructions stored thereon which, when executed by the one or more computers, cause the one or more computers to perform operations comprising:
• obtaining a predictive model that is trained to predict a likelihood of a user remaining on a clicked-through page associated with a result of a given revised query that is obtained by a query revisor for a given original query, wherein the predictive model is trained based at least in part on (i) features of original queries that have been submitted to a search engine, and, (ii) for each original query, features of revised queries which were generated by the query revisor for the original query;
• obtaining (i) a particular original query, and (ii) one or more particular revised queries that the query revisor has generated for the particular original query;
• providing, to the predictive model, (i) one or more features of the particular original query, and (ii) one or more features from each of one or more of the particular revised queries that the query revisor has generated for the particular original query;
• and
• obtaining, from the predictive model, a respective predicted likelihood for each of the one or more of the particular revised queries.

A non-transitory computer-readable medium encoded with a computer program comprising instructions that, when executed, operate to cause a computer to perform operations comprising:

• obtaining a predictive model that is trained to predict a likelihood of a user remaining on a clicked-through page associated with a result of a given revised query that is obtained by a query revisor for a given original query, wherein the predictive model is trained based at least in part on (i) features of original queries that have been submitted to a search engine, and, (ii) for each original query, features of revised queries which were generated by the query revisor for the original query;
• obtaining (i) a particular original query, and (ii) one or more particular revised queries that the query revisor has generated for the particular original query;
• providing, to the predictive model, (i) one or more features of the particular original query, and (ii) one or more features from each of one or more of the particular revised queries that the query revisor has generated for the particular original query;
• and
• obtaining, from the predictive model, a respective predicted likelihood for each of the one or more of the particular revised queries.