User query mining for advertising matching

A computing device comprising:

• a memory;
• one or more processors operatively coupled to the memory;
• a session segmentation module maintained in the memory and executed on the one or more processors configured to segment search session logs of a user into search sessions and to identify a first keyword set and a second keyword set from a search session;
• and
• a similarity calculation module maintained in the memory and executed on the one or more processors configured to:
• determine a minimum frequency as a smaller of a frequency at which the first keyword set occurs and a frequency at which the second keyword set occurs;
• combine the first keyword set and the second keyword set to form a keyword set pair;
• calculate a similarity value for the first and second keyword sets based at least in part on a comparison made between the minimum frequency and a frequency at which the keyword set pair occurs in the search session;
• compare the similarity value with a similarity threshold;
• and
• display one or more semantically relevant keyword sets based on the comparison;
• and
• a threshold calculation module maintained in the memory and executed on the one or more processors configured to:
• generate a training set which includes a predetermined number of queries obtained from one or more training search sessions;
• identify a predetermined number of suggested queries for each query in the training set based on a relevance of the suggested queries to each query in the training set;
• categorize the suggested queries into one or more groups based on the relevance;
• determine a level of relevance for each of the one or more groups based on the relevance of the suggested queries categorized into each of the one or more groups;
• and
• determine the similarity threshold based in part on the level of relevance of at least one of the one or more groups.

A computer-readable data storage medium having a set of computer readable instructions that, when executed by a processor, perform acts comprising:

• identifying search session logs of a user;
• segmenting the search session logs into search sessions;
• analyzing each of the search sessions to determine a first keyword set and a second keyword set;
• combining the first keyword set and the second keyword set to form a pair of keyword sets;
• determining a minimum frequency as a smaller of a frequency at which the first keyword set occurs and a frequency at which the second keyword set occurs;
• comparing a value proportional to the minimum frequency with a frequency at which the pair of keyword sets occur;
• computing semantic relevance between the first keyword set and the second keyword set based at least in part on the comparing, wherein the computing further comprises:
• measuring a similarity value between the first and second keyword sets based at least in part on the minimum frequency and the frequency at which the pair of keyword sets occur;
• and
• evaluating the similarity value against a similarity threshold, wherein the similarity threshold is computed by:
• generating a training set wherein the training set includes a first predetermined number of queries obtained from a training search session;
• recognizing a second predetermined number of suggested queries for each query in the training set based on a relevance of the suggested queries to the query, wherein the relevance is determined either by a mutual information analysis or a cosine similarity analysis;
• and
• determining a similarity threshold for each group of suggested queries, wherein the suggested queries are grouped into one or more groups based on the relevance;
• and
• displaying one or more semantically relevant keyword sets based on the computation.