Disambiguating a search query based on a difference between composite domain-confidence factors
AssigneeMicrosoft Corporation
InventorsThomas F. Bergstraesser, Bert Casper, Robert David Berg
Life of this patent
Estimating the term…
Abstract
A query may be ambiguous when the domain-confidence factors indicate a high likelihood that the search query may belong to two or more domains. A domain is a subject matter category such as shopping, sports, entertainment, music, or politics. A domain-confidence factor indicates the likelihood that a search query is related to a particular domain. Upon determining that a search query is ambiguous, embodiments of the present invention may present an interface that allows the user to place the search query into one or more domains. Updated search results, or other objects, may be returned based on the selection of a domain by the user.
Claims19
One or more computing devices, each coupled to one or more hardware processors, the one or more computing devices having computer-executable instructions embodied thereon for performing a method of determining whether a search query is ambiguous, the method comprising:
- receiving, at a server, the search query sent from a computing device;
- generating initial search results that are responsive to the search query;
- identifying a plurality of domains, wherein each domain of the plurality of domains is a subject-matter category,
- generating a composite domain-confidence factor for each domain of the plurality of domains, wherein the composite domain-confidence factor for each domain of the plurality of domains is derived from multiple domain-confidence factors calculated by one or more domain classifiers, wherein the composite domain-confidence factor indicates a probability that the search query is associated with an individual domain;
- determining that the search query is ambiguous because a difference between a highest composite domain-confidence factor and a second-highest composite domain-confidence factor generated for the search query is less than a threshold value, wherein the highest composite domain-confidence factor is associated with a first domain and the second-highest composite domain-confidence factor is associated with a second domain;
- and
- in response to said determining that the search query is ambiguous, displaying a user interface that includes the initial search results for the search query and allows a user of the computing device to filter the initial search results according to the first domain or the second domain.
A method of soliciting domain information from a user that submits an ambiguous search query, the method comprising:
- receiving, at a server, a search query from a computing device associated with the user;
- generating one or more domain-confidence factors for each domain within a domain ontology by submitting the search query to a plurality of domain classifiers, each of which generate a domain-confidence factor that expresses the probability that the search query is associated with an individual domain, wherein the plurality of domain classifiers includes at least two domain classifiers that are associated with the individual domain, and wherein a domain is a subject-matter category;
- calculating a composite domain-confidence factor for said each domain within the domain ontology based on multiple domain-confidence factors generated by the plurality of domain classifiers for said each domain;
- determining a highest composite domain-confidence factor calculated for the search query, wherein the highest composite domain-confidence factor is associated with a first domain;
- determining a second-highest composite domain-confidence factor calculated for the search query, wherein the second-highest composite domain-confidence factor is associated with a second domain;
- determining that the search query is ambiguous because a difference between the highest composite domain-confidence factor and the second-highest composite domain-confidence factor is less than a threshold amount;
- and
- upon said determining that the search query is ambiguous, displaying a user interface that allows the user to select an intended domain for the search query, wherein the user interface allows the user to select one or more of the first domain and the second domain.
One or more computing devices, each coupled to one or more hardware processors, the one or more computing devices having computer-executable instructions embodied thereon for performing a method of disambiguating a search query, the method comprising:
- communicating a user interface to a separate computing device that displays the user interface to a user, wherein the user interface includes a first portion that receives the search query from the user;
- receiving the search query through the user interface;
- generating one or more domain-confidence factors for each domain within a domain ontology by submitting the search query to a plurality of domain classifiers, each of which generate a domain-confidence factor that expresses the probability that the search query is associated with an individual domain, wherein the plurality of domain classifiers includes at least two domain classifiers that are associated with the individual domain, and wherein a domain is a subject-matter category;
- calculating a composite domain-confidence factor for said each domain within the domain ontology based on multiple domain-confidence factors generated by the plurality of domain classifiers for said each domain;
- determining a highest composite domain-confidence factor calculated for the search query, wherein the highest composite domain-confidence factor is associated with a first domain;
- determining a second-highest composite domain-confidence factor calculated for the search query, wherein the second-highest composite domain-confidence factor is associated with a second domain;
- determining that the search query is ambiguous because a difference between the highest composite domain-confidence factor and the second-highest composite domain-confidence factor is less than a threshold amount, wherein said determining that the search query is ambiguous is further based on a closeness of the domains in the domain ontology;
- and
- communicating, to the separate computing device, a second portion of the user interface that displays two or more domains into which the search query has a highest probability of being categorized, wherein the two or more domains are user selectable.
Dependencies are inferred from the claim text.