Method of and system for enriching search queries for ranking search results

A computer-implemented method for ranking search results in response to a current query using a first machine learning algorithm (MLA), the method being executed by a server hosting a second MLA, the server being connected to a database, the server being connected to an electronic device via a communication network, the second MLA having been trained to determine query similarity based on textual content thereof, the method comprising:

• receiving, by the server from the electronic device, an indication of the current query, the current query not having been previously submitted to the server;
• generating, by the server, a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• retrieving, by the second MLA from the database, a plurality of past queries, each past query of the plurality of past queries having been previously submitted to the server;
• computing, by the second MLA, a respective similarity parameter between the current query and a respective one of the plurality of past queries, the respective similarity parameter being indicative of a degree of similarity between the current query and the respective one of the plurality of past queries, the computing for a given one of the plurality of past queries being based on:
• at least one textual feature of the new query, and
• at least one textual feature of the given one of the plurality of past queries;
• selecting a subset of the plurality of the past queries based on the respective similarity parameter being above a pre-determined similarity threshold,
• ranking the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account the subset of the plurality of past queries as a ranking feature;
• and
• transmitting, by the server, a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A computer-implemented method for ranking search results in response to a current query using a first machine learning algorithm (MLA), the method executable by a server, the server being connected to a database, the server being connected to an electronic device via a communication network, the method comprising:

• receiving, by the server from the electronic device, an indication of the current query;
• generating, by the server, a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• determining, by the server, a presence of past user interactions with current documents in the set of current documents, the past user interactions with the current documents having been performed in response to a past query similar to the current query;
• retrieving, by the server from the database, a set of a plurality of past queries, each given past query of the plurality of past queries being associated with a respective set of past documents,
• the respective set of past documents having been presented as search results in response to the given past query of at least some others of the plurality of past queries,
• each past document of the respective set of past documents being associated with past user interactions with the respective past document presented in response to the given past query;
• the retrieving being executed:
• in response to the determining resulting in a positive outcome:
• generating, by the server, a current query vector associated with the current query based on:
• current documents in the set of current documents and the past user interactions associated with the current documents in the set of current documents;
• generating, by the server, a respective query vector associated with a each of the plurality of past queries, the generating being based on:
• the past documents and the past user interactions with the past documents in the respective set of past documents;
• selecting the plurality of past queries based on the similarity of the current vector sand the respective query vector;
• in response to the determining resulting in a negative outcome:
• generating, by the server, the current query vector associated with the current query based on at least one textual feature of the current query;
• generating, by the server, a respective query vector associated with each given past query of the plurality of past queries, the generating being based on:
• at least one textual feature of the given past query;
• selecting the plurality of past queries based on the similarity of the current query vector sand the respective query vector;
• ranking the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account at least one past query of the set of the plurality of past queries;
• and
• transmitting, by the server, a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A computer-implemented method for ranking search results in response to a current query using a first machine learning algorithm (MLA), the method executable by a server, the server being connected to a database, the server being connected to an electronic device via a communication network, the method comprising:

• receiving, by the server from the electronic device, an indication of the current query;
• generating, by the server, a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• retrieving, by the server from the database, based on the indication of the current query, a plurality of past queries;
• computing, by the server, a respective similarity parameter between the current query and a respective one of the plurality of past queries, the respective similarity parameter being indicative of a degree of similarity between the current query and the respective one of the plurality of past queries;
• ranking the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account inclusion of search terms of at least one past query of the plurality of past queries in a given one of the set of current documents so that inclusion of search terms promotes rank of the given current document;
• and
• transmitting, by the server, a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A system for ranking search results in response to a current query using a first machine learning algorithm (MLA), the system being connected to a database, the system being connected to an electronic device via a communication network, the system comprising:

• a processor;
• a non-transitory computer-readable medium comprising instructions;
• the processor, upon executing the instructions, being configured to:
• receive from the electronic device, an indication of the current query;
• generate a set of search results, the set of search results containing a set of current documents relevant to the current query;
• retrieve, from the database, a plurality of past queries, each given past query of the plurality of past queries being associated with a respective set of past documents,
• the respective set of past documents having been presented as search results in response to the given past query of at least some others of the plurality of past queries;
• compute a respective similarity parameter between the current query and a respective one of the plurality of past queries, the respective similarity parameter being indicative of a degree of similarity between the current query and the respective one of the plurality of past queries, the computing for a given one of the plurality of past queries being based on:
• current documents in the set of current documents, and
• past documents in the respective set of past documents associated with the given one of the plurality of queries;
• select a subset of the plurality of the past queries based on the respective similarity parameter being above a predetermined similarity threshold;
• rank the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account the subset of the plurality of past queries as a ranking feature;
• and
• transmit a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A system for ranking search results in response to a current query using a first machine learning algorithm (MLA), the system hosting a second MLA, the system being connected to a database, the system being connected to an electronic device via a communication network, the second MLA having been trained to determine query similarity based on textual content thereof, the system comprising:

• a processor;
• a non-transitory computer-readable medium comprising instructions;
• the processor, upon executing the instructions, being configured to:
• receive from the electronic device, an indication of the current query, the current query not having been previously submitted to the system;
• generate a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• retrieve by the second MLA from the database, a plurality of past queries, each past query of the plurality of past queries having been previously submitted to the system;
• compute, by the second MLA, a respective similarity parameter between the current query and a respective one of the plurality of past queries, the respective similarity parameter being indicative of a degree of similarity between the current query and the respective one of the plurality of past queries, the computing for a given one of the plurality of past queries being based on:
• at least one textual feature of the new query, and
• at least one textual feature of the given one of the plurality of past queries;
• select a subset of the plurality of the past queries based on the respective similarity parameter being above a pre-determined similarity threshold,
• rank the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account the subset of the plurality of past queries as a ranking feature;
• and
• transmit a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A system for ranking search results in response to a current query using a first machine learning algorithm (MLA), the system being connected to a database, the system being connected to an electronic device via a communication network, the system comprising:

• a processor;
• a non-transitory computer-readable medium comprising instructions;
• the processor, upon executing the instructions, being configured to:
• receive, from the electronic device, an indication of the current query;
• generate a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• determine a presence of past user interactions with current documents in the set of current documents, the past user interactions with the current documents having been performed in response to a past query similar to the current query;
• retrieve, from the database, a set of a plurality of past queries, each given past query of the plurality of past queries being associated with a respective set of past documents,
• the respective set of past documents having been presented as search results in response to the given past query of at least some others of the plurality of past queries,
• each past document of the respective set of past documents being associated with past user interactions with the respective past document presented in response to the given past query;
• the retrieving being executed:
• in response to the determining resulting in a positive outcome:
• generate a current query vector associated with the current query based on:
• current documents in the set of current documents and the past user interactions associated with the current documents in the set of current documents;
• generate a respective query vector associated with a each of the plurality of past queries, the generating being based on:
• the past documents and the past user interactions with the past documents in the respective set of past documents;
• select the plurality of past queries based on the similarity of the current vector sand the respective query vector;
• in response to the determining resulting in a negative outcome:
• generate the current query vector associated with the current query based on at least one textual feature of the current query;
• generate a respective query vector associated with each given past query of the plurality of past queries, the generating being based on:
• at least one textual feature of the given past query;
• select the plurality of past queries based on the similarity of the current query vector sand the respective query vector;
• rank the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account at least one past query of the set of the plurality of past queries;
• and
• transmit a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.

A system for ranking search results in response to a current query using a first machine learning algorithm (MLA), the system being connected to a database, the system being connected to an electronic device via a communication network, the system comprising:

• a processor;
• a non-transitory computer-readable medium comprising instructions;
• the processor, upon executing the instructions, being configured to:
• receive, from the electronic device, an indication of the current query;
• generate a set of search results, the set of search results containing a set of current documents relevant to the current search query;
• retrieve, from the database, based on the indication of the current query, a plurality of past queries;
• compute a respective similarity parameter between the current query and a respective one of the plurality of past queries, the respective similarity parameter being indicative of a degree of similarity between the current query and the respective one of the plurality of past queries;
• rank the set of current documents to obtain a ranked set of documents, the ranking being done by the first MLA, the first MLA taking into account inclusion of search terms of at least one past query of the plurality of past queries in a given one of the set of current documents so that inclusion of search terms promotes rank of the given current document;
• and
• transmit a search engine results page (SERP) for display on the electronic device, the SERP including the ranked set of documents.