Predicting site quality

1. A method comprising:

obtaining baseline site quality scores for a plurality of previously scored sites;

generating a phrase model for a plurality of sites including the plurality of previously scored sites, wherein the phrase model defines, for each of a plurality of phrases, a mapping from each of one or more phrase specific relative frequency measures to a corresponding phrase specific average site quality score, wherein:

each respective phrase-specific relative frequency measure associated with a phrase represents a measure of how many pages in a site contain the phrase relative to how many pages are on the respective site; and

the average site quality score for a particular phrase specific relative frequency measure is a measure of a central tendency of the baseline site quality scores of those previously scored sites that have the phrase in the pages of the site with a relative frequency that corresponds to the particular phrase specific relative frequency measure;

for a new site, the new site not being one of the plurality of previously scored sites, obtaining a respective relative frequency measure for each of a plurality of phrases found on the new site, wherein the new site comprises a plurality of pages, and wherein, for each phrase of the plurality of phrases, the relative frequency measure is a measure of how many of the pages on the new site contain the phrase relative to a count of how many pages are on the new site;

determining an aggregate site quality score for the new site from the phrase model using the relative frequency measures of the plurality of phrases found on the new site, comprising:

determining, for each of the plurality of phrases found on the new site, a corresponding phrase-specific average site quality score from the phrase model;

determining, as the aggregate site quality score for the new site, a measure of central tendency of the phrase-specific average site quality scores corresponding to the plurality of phrases found on the new site; and

determining a predicted site quality score for the new site from the aggregate site quality score;

generating respective ranking scores for each of a plurality of resources that satisfy a search query, each of the plurality of resources being in a respective site, including generating the respective ranking scores using the baseline site quality scores of the site for each resource in a previously-scored site and generating the respective ranking scores using the respective predicted site quality score for each resource in a site having a predicted site quality score; and

using the respective ranking scores to rank the plurality of resources that satisfy the search query.

2. The method of claim 1 , wherein:

each phrase is an n-gram of tokens, the n-gram being a 2-gram, 3-gram, 4-gram or 5-gram.

3. The method of claim 1 , wherein the phrase model defines a mapping from each of one or more phrase-specific relative frequency measures to a corresponding vector of phrase-specific average site quality scores.

4. A non-transitory storage medium encoded with instructions that, when executed by one or more computers, cause the one or more computers to perform operations comprising:

obtaining baseline site quality scores for a plurality of previously scored sites;

generating a phrase model for a plurality of sites including the plurality of previously scored sites, wherein the phrase model defines, for each of a plurality of phrases, a mapping from each of one or more phrase specific relative frequency measures to a corresponding phrase specific average site quality score, wherein:

each respective phrase-specific relative frequency measure associated with a phrase represents a measure of how many pages in a site contain the phrase relative to how many pages are on the respective site; and

the average site quality score for a particular phrase specific relative frequency measure is a measure of a central tendency of the baseline site quality scores of those previously-scored sites that have the phrase in the pages of the site with a relative frequency that corresponds to the particular phrase specific relative frequency measure;

for a new site, the new site not being one of the plurality of previously scored sites, obtaining a respective relative frequency measure for each of a plurality of phrases found on the new site, wherein the new site comprises a plurality of pages, and wherein, for each phrase of the plurality of phrases, the relative frequency measure is a measure of how many of the pages on the new site contain the phrase relative to a count of how many pages are on the new site;

determining an aggregate site quality score for the new site from the phrase model using the relative frequency measures of the plurality of phrases found on the new site, comprising:

determining, for each of the plurality of phrases found on the new site, a corresponding phrase-specific average site quality score from the phrase model;

determining, as the aggregate site quality score for the new site, a measure of central tendency of the phrase-specific average site quality scores corresponding to the plurality of phrases found on the new site; and

determining a predicted site quality score for the new site from the aggregate site quality score;

generating respective ranking scores for each of a plurality of resources that satisfy a search query, each of the plurality of resources being in a respective site, including generating the respective ranking scores using the baseline site quality scores of the site for each resource in a previously-scored site and generating the respective ranking scores using the respective predicted site quality score for each resource in a site having a predicted site quality score; and

using the respective ranking scores to rank the plurality of resources that satisfy the search query.

5. The non-transitory storage medium of claim 4 , wherein:

each phrase is an n-gram of tokens, the n-gram being a 2-gram, 3-gram, 4-gram or 5-gram.

6. The non-transitory storage medium of claim 4 , wherein the phrase model defines a mapping from each of one or more phrase-specific relative frequency measures to a corresponding vector of phrase-specific average site quality scores.

7. A system comprising one or more computers and one or more storage devices storing instructions that are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:

obtaining baseline site quality scores for a plurality of previously scored sites;

generating a phrase model for a plurality of sites including the plurality of previously scored sites, wherein the phrase model defines, for each of a plurality of phrases, a mapping from each of one or more phrase specific relative frequency measures to a corresponding phrase specific average site quality score, wherein:

each respective phrase-specific relative frequency measure associated with a phrase represents a measure of how many pages in a site contain the phrase relative to how many pages are on the respective site; and

the average site quality score for a particular phrase specific relative frequency measure is a measure of a central tendency of the baseline site quality scores of those previously-scored sites that have the phrase in the pages of the site with a relative frequency that corresponds to the particular phrase specific relative frequency measure;

for a new site, the new site not being one of the plurality of previously scored sites, obtaining a respective relative frequency measure for each of a plurality of phrases found on the new site, wherein the new site comprises a plurality of pages, and wherein, for each phrase of the plurality of phrases, the relative frequency measure is a measure of how many of the pages on the new site contain the phrase relative to a count of how many pages are on the new site;

determining an aggregate site quality score for the new site from the phrase model using the relative frequency measures of the plurality of phrases found on the new site, comprising:

determining, for each of the plurality of phrases found on the new site, a corresponding phrase-specific average site quality score from the phrase model;

determining, as the aggregate site quality score for the new site, a measure of central tendency of the phrase-specific average site quality scores corresponding to the plurality of phrases found on the new site; and

determining a predicted site quality score for the new site from the aggregate site quality score;

generating respective ranking scores for each of a plurality of resources that satisfy a search query, each of the plurality of resources being in a respective site, including generating the respective ranking scores using the baseline site quality scores of the site for each resource in a previously-scored site and generating the respective ranking scores using the respective predicted site quality score for each resource in a site having a predicted site quality score; and

using the respective ranking scores to rank the plurality of resources that satisfy the search query.

8. The system of claim 7 , wherein:

each phrase is an n-gram of tokens, the n-gram being a 2-gram, 3-gram, 4-gram or 5-gram.

9. The system of claim 7 , wherein the phrase model defines a mapping from each of one or more phrase-specific relative frequency measures to a corresponding vector of phrase-specific average site quality scores.