Providing query explanations for automated sourcing
Techniques for providing explanations of candidate search queries are described. The queries can be created using query intent clustering in an automated sourcing tool. In an example embodiment, disclosed is a system that obtains one or more current candidate member profiles used as a basis for a search on member profiles in an online system. Additionally, the system extracts one or more attributes from the one or more current candidate member profiles. Moreover, the system identifies query intent clusters based on the extracted one or more attributes. Furthermore, the system generates a search query based on the identified query intent clusters. Next, an explanation of the search query can be displayed on a display device of the system. In some embodiments, the online system hosts a social networking service that includes the member profiles, and the identified query intent clusters include skills clusters.
1. A computer system, comprising:
one or more processors; and
a non-transitory computer readable storage medium storing instructions that when executed by the one or more processors cause the computer system to perform operations comprising:
obtaining training data;
dividing the training data into a positive class and a negative class, wherein the positive class includes data associated with items of a particular category and the negative class includes data associated with items not of the particular category;
sampling the training data by oversampling the positive class with respect negative class;
using the sampled training data to train a machine learned model using a machine learning algorithms, the training assigning weights to one or more query-based features;
obtaining one or more current candidate member profiles used as a basis for a search on member profiles in an online system having a plurality of members;
extracting one or more attributes from the one or more current candidate member profiles;
identifying a plurality of query intent clusters based on the extracted one or more attributes;
selecting one of the plurality of identified query intent clusters based on one or more criteria evaluated for the plurality of identified query intent clusters;
generating a search query based on the selected one of the one or more identified query intent clusters;
causing a search to be performed based on the search query, returning a recommended member profile;
receiving one or more refinements of the search query; and
retraining the machine learned model based on the one or more refinements, altering the weights assigned to the one or more query-based features.
2. The system of claim 1 , wherein the online system hosts a social networking service including the member profiles, and wherein the identified query intent clusters include skills clusters.
3. The system of claim 1 , wherein the obtaining comprises:
obtaining one or more profiles of current candidates having a given title; and
obtaining one or more profiles of current candidates having titles synonymous with the given title.
4. The system of claim 3 , wherein the given title is selected by a user, and wherein the obtaining further comprises retrieving the titles synonymous with the given title from a data store on the storage device that maps titles to their synonyms.
5. The system of claim 1 , wherein the selecting includes using the trained machine learned model to determine one arm from a set of arms, wherein each arm of the trained machine learned model corresponds to a different identified query intent cluster, the trained machine learned model returning a reward from an estimated distribution of each arm, wherein the reward is feedback received from a user via for a ranked set of results from a prior search query generated based on an arm.
6. The system of claim 1 , wherein the extracting comprises:
extracting skills from the one or more current candidate member profiles; and
mapping the extracted skills to skill identifiers.
7. The system of claim 1 , wherein the identifying comprises:
creating a skills taxonomy for the skill identifiers; and
representing each skill identifier in the skills taxonomy as a categorical variable that is either absent or present in a member profile.
8. A computer-implemented method, comprising:
obtaining training data;
dividing the training data into a positive class and a negative class, wherein the positive class includes data associated with items of a particular category and the negative class includes data associated with items not of the particular category;
sampling the training data by oversampling the positive class with respect to the negative class;
use the sampled training data to train a machine learned model using a machine learning algorithm, the training assigning weights to one or more query-based features;
obtaining one or more current candidate member profiles used as a basis for a search on member profiles in an online system having a plurality of members;
extracting one or more attributes from the one or more current candidate member profiles;
identifying a plurality of query intent clusters based on the extracted one or more attributes;
selecting, using the machine learned model, one of the plurality of identified query intent clusters based on one or more criteria evaluated for the plurality of identified query intent clusters;
generating a search query based on the selected one of the one or more identified query intent clusters;
causing a search to be performed based on the search query, returning a recommended member profile;
receiving one or more refinements of the search query; and
retraining the machine learned model based on the one or more refinements, altering the weights assigned to the one or more query-based features.
9. The method of claim 8 , wherein the online system hosts a social networking service including the member profiles, and wherein the identified query intent clusters include skills clusters.
10. The method of claim 8 , wherein the obtaining comprises:
obtaining one or more profiles of current candidates having a given title; and
obtaining one or more profiles of current candidates having titles synonymous with the given title.
11. The method of claim 10 , wherein the given title is selected by a user, and wherein the obtaining further comprises retrieving the titles synonymous with the given title from a database table that maps titles to their synonyms.
12. The method of claim 1 , wherein the selecting includes using a trained machine learned model to determine one arm from a set of aims, wherein each aim of the trained machine learned model corresponds to a different identified query intent cluster, the trained machine learned model returning a reward from an estimated distribution of each arm, wherein the reward is feedback received from a user via for a ranked set of results from a prior search query generated based on an arm.
13. The method of claim 8 , further comprising:
performing the search on the member profiles in the social networking service using the generated search query, returning one or more result member profiles as candidates.
14. The method of claim 8 , wherein the extracting comprises:
extracting skills from the one or more current candidate member profiles; and
mapping the extracted skills to skill identifiers.
15. The method of claim 14 , wherein the identifying comprises performing query intent clustering for the skill identifiers.
16. The method of claim 15 , wherein the identifying comprises creating a skills taxonomy for the skill identifiers; and
representing each skill identifier in the skills taxonomy as a categorical variable that is either absent or present in a member profile.
17. A non-transitory machine-readable storage medium comprising instructions that, when executed by one or more processors of a machine, cause the machine to perform operations including:
obtaining training data;
dividing the training data into a positive class and a negative class, wherein the positive class includes data associated with items of a particular category and the negative class includes data associated with items not of the particular category;
sampling the training data by oversampling the positive class with respect to the negative class;
use the sampled training data to train a machine learned model using a machine learning algorithm, the training assigning weights to one or more query-based features;
obtaining one or more current candidate member profiles used as a basis for a search on member profiles in an online system having a plurality of members;
extracting one or more attributes from the one or more current candidate member profiles;
identifying a plurality of query intent clusters based on the extracted one or more attributes;
selecting, using the machine learned model, one of the plurality of identified query intent clusters based on one or more criteria evaluated for the plurality of identified query intent clusters;
generating a search query based on the selected one of the one or more identified query intent clusters;
causing a search to be performed based on the search query, returning a recommended member profile;
receiving one or more refinements of the search query; and
retraining the machine learned model based on the one or more refinements, altering the weights assigned to the one or more query-based features.
18. The machine-readable storage medium of claim 17 , wherein the online system hosts a social networking service including the member profiles, and wherein the identified query intent clusters include skills clusters.
19. The machine-readable storage medium of claim 17 , wherein identifying the query intent clusters comprises:
performing query intent clustering for the skill identifiers;
creating a skills taxonomy for the skill identifiers; and
representing each skill identifier in the skills taxonomy as a categorical variable that is either absent or present in a member profile.