Utilizing user co-search behavior to identify search queries seeking inappropriate content

A computer-implemented method comprising:

• obtaining a first set of seed terms related to a particular type of content;
• obtaining a second set of seed terms related to a subset of the particular type of content;
• identifying a set of search queries that include one or more terms from the first set of seed terms and one or more terms from the second set of seed terms;
• classifying one or more queries of the identified set of search queries as reference queries, the reference queries being associated with the subset of the particular type of content;
• identifying, for each of the reference queries that are associated with the subset of the particular type of content, a set of co-occurring queries that co-occur with any reference query in a collection of search queries received by a search engine;
• determining, for each of the co-occurring queries in the set of the co-occurring queries, (i) a number of occurrences of the co-occurring query in the collection of the search queries received by the search engine, and (ii) a number of co-occurrences of the co-occurring query with any reference query in the collection of the search queries received by the search engine;
• determining, for each of the co-occurring queries, a co-occurrence value based on (i) the number of occurrences of the co-occurring query in the collection of the search queries received by the search engine, and (ii) the number of co-occurrences of the co-occurring query with any reference query in the collection of the search queries received by the search engine;
• classifying, for each of the co-occurring queries, the co-occurring query as (i) a candidate query, or (ii) not a candidate query, based at least on the co-occurrence value;
• determining, for each candidate query, whether the candidate query satisfies one or more criteria associated with selecting queries that are likely associated with the subset of the particular type of content;
• classifying each candidate query as (i) likely associated with the subset of the particular type of content, or (ii) unlikely to be associated with the subset of the particular type of content, based at least on whether the candidate query is determined to satisfy the one or more criteria;
• and
• providing one or more candidate queries that are classified as likely associated with the particular subset of the particular type of content as training data for training a classifier.

A system comprising:

• one or more computers and one or more storage devices storing instructions that are operable and when executed by one or more computers, cause the one or more computers to perform actions comprising:
• obtaining a first set of seed terms related to a particular type of content;
• obtaining a second set of seed terms related to a subset of the particular type of content;
• identifying a set of search queries that include one or more terms from the first set of seed terms and one or more terms from the second set of seed terms;
• classifying one or more queries of the identified set of search queries as reference queries, the reference queries being associated with the subset of the particular type of content;
• identifying, for each of the reference queries that are associated with the subset of the particular type of content, a set of co-occurring queries that co-occur with any reference query in a collection of search queries received by a search engine;
• determining, for each of the co-occurring queries in the set of the co-occurring queries, (i) a number of occurrences of the co-occurring query in the collection of the search queries received by the search engine, and (ii) a number of co-occurrences of the co-occurring query with any reference query in the collection of the search queries received by the search engine;
• determining, for each of the co-occurring queries, a co-occurrence value based on (i) the number of occurrences of the co-occurring query in the collection of the search queries received by the search engine, and (ii) the number of co-occurrences of the co-occurring query with any reference query in the collection of the search queries received by the search engine;
• classifying, for each of the co-occurring queries, the co-occurring query as (i) a candidate query, or (ii) not a candidate query, based at least on the co-occurrence value;
• determining, for each candidate query, whether the candidate query satisfies one or more criteria associated with selecting queries that are likely associated with the subset of the particular type of content;
• classifying each candidate query as (i) likely associated with the subset of the particular type of content, or (ii) unlikely to be associated with the subset of the particular type of content, based at least on whether the candidate query is determined to satisfy the one or more criteria;
• and
• providing one or more candidate queries that are classified as likely associated with the particular subset of the particular type of content as training data for training a classifier.