Continuous augmentation method for ranking components in information retrieval
According to one embodiment, a method, computer system, and computer program product for continuously ranking components in a live information is provided. The present embodiment may include receiving search feedback derived from interactions between users and information retrieval systems; assigning weights to each of the ranking components; adjusting the assigned weights based on search feedback; modifying the current set of ranking components based on the search feedback by adding new ranking components and deleting old ranking components; transmitting a query from the users to the current set of ranking components; aggregating ranking results from the transmitted query into a single ranking based on the weights; and transmitting the single ranking to the users.
1. A computer system for continuously ranking components in a live information retrieval system, the computer system comprising:
one or more processors, one or more computer-readable memories, one or more computer-readable tangible storage medium, and program instructions stored on at least one of the one or more tangible storage medium for execution by at least one of the one or more processors via at least one of the one or more memories, wherein the computer system is capable of performing a method comprising:
implementing, within a ranking program, one or more ranking components to form a current set of ranking components;
receiving a plurality of search feedback derived from one or more interactions between one or more users and one or more information retrieval systems;
assigning one or more weights to each of the one or more ranking components;
adjusting the one or more assigned weights based on a plurality of search feedback;
modifying the current set of ranking components based on the plurality of search feedback by adding one or more new ranking components and deleting one or more old ranking components;
transmitting a query from the one or more users to the current set of ranking components;
aggregating one or more ranking results from the transmitted query into a single ranking based on the one or more weights; and
transmitting the single ranking to the one or more users.
2. The computer system of claim 1 , wherein at least one of the current set of ranking components is a randomized ranking component.
3. The method of claim 1 , wherein the one or more old ranking components are deleted when the one or more weights associated with the one or more old ranking components fall below a threshold.
4. The computer system of claim 1 , wherein the one or more added ranking components are selected when the plurality of search feedback comprises a plurality of non-use-case-specific data or no data, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component that relies on base information retrieval, a ranking component trained on a plurality of pre-labelled data chosen based on a document match, a randomized ranking component, and a ranking component trained on a plurality of expert-provided data for one or more related general use cases.
5. The computer system of claim 1 , wherein the one or more added ranking components are selected when the search feedback comprises one or more query logs or one or more click logs, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component trained on a plurality of pre-labelled data chosen based on a feature vector similarity, a ranking component trained using one or more queries from one or more query logs, and a ranking component trained using one or more queries and one or more clicks from one or more logs.
6. The computer system of claim 1 , wherein the one or more added ranking components are selected when the search feedback comprises a plurality of explicit feedback, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component trained using one or more queries and one or more clicks with a bias correction, and a ranking component trained on a plurality of explicit feedback on a plurality of use-case-specific data.
7. The computer system of claim 1 , wherein the one or more ranking results are aggregated using an ensembling method that uses a weighted average of one or more normalized scores from the current set of ranking components.
8. A computer program product for continuously ranking components in a live information retrieval system, the computer program product comprising:
one or more computer-readable tangible storage medium and program instructions stored on at least one of the one or more tangible storage medium, the program instructions executable by a processor, the program instructions instructing the processor to carry out a method comprising:
providing one or more ranking components wherein at least one ranking component is randomized;
implementing, within a ranking program, one or more ranking components to form a current set of ranking components;
receiving a plurality of search feedback derived from one or more interactions between one or more users and one or more information retrieval systems;
assigning one or more weights to each of the one or more ranking components;
adjusting the one or more assigned weights based on a plurality of search feedback;
modifying the current set of ranking components based on the plurality of search feedback by adding one or more new ranking components and deleting one or more old ranking components;
transmitting a query from the one or more users to the current set of ranking components;
aggregating one or more ranking results from the transmitted query into a single ranking based on the one or more weights; and
transmitting the single ranking to the one or more users.
9. The computer program product of claim 8 , wherein at least one of the current set of ranking components is a randomized ranking component.
10. The computer program product of claim 8 , wherein the one or more old ranking components are deleted when the one or more weights associated with the one or more old ranking components fall below a threshold.
11. The computer program product of claim 8 , wherein the one or more added ranking components are selected when the plurality of search feedback comprises a plurality of non-use-case-specific data or no data, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component that relies on base information retrieval, a ranking component trained on a plurality of pre-labelled data chosen based on a document match, a randomized ranking component, and a ranking component trained on a plurality of expert-provided data for one or more related general use cases.
12. The computer program product of claim 8 , wherein the one or more added ranking components are selected when the search feedback comprises one or more query logs or one or more click logs, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component trained on a plurality of pre-labelled data chosen based on a feature vector similarity, a ranking component trained using one or more queries from one or more query logs, and a ranking component trained using one or more queries and one or more clicks from one or more logs.
13. The computer program product of claim 8 , wherein the one or more added ranking components are selected when the search feedback comprises a plurality of explicit feedback, and wherein the one or more added ranking components are selected from a group consisting of: a ranking component trained using one or more queries and one or more clicks with a bias correction, and a ranking component trained on a plurality of explicit feedback on a plurality of use-case-specific data.