Query building for search by ideal candidates

1. A computer-implemented method, comprising:

causing display, in a graphical user interface, of identifications of a plurality of member profiles, each of the member profiles comprising a page, renderable by a user interface, containing attributes of a different member of an online service:

receiving, via the graphical user interface, a selection of a plurality of the identifications, each of the plurality of identifications corresponding to a different member profile from the plurality of member profiles and identifying the corresponding member profile as one of a plurality of ideal candidate member profiles;

extracting, by a hardware processor, one or more attributes from the plurality of ideal candidate member profiles;

generating, by the hardware processor, a search query based on the extracted one or more attributes, the generating comprising:

forming an expertise matrix, the expertise matrix containing an expertise score for each of a plurality of combinations of member profiles and skills for members of the online service, the expertise score for a given combination of member profile and skill indicating a level of expertise for a corresponding member in a corresponding skill;

factorizing the expertise matrix into a member matrix and a skills matrix;

ranking the one or more attributes from the plurality of ideal candidate member profiles based on the expertise matrix and the skills matrix; and

selecting top N attributes based on the ranking;

modifying the generated search query to include the selected top N attributes; and

performing a search on member profiles in the online service using the modified generated search query, returning one or more result member profiles.

2. The method of claim 1 , wherein the expertise scores are generated using a supervised machine learning algorithm that combines a plurality of signals, the signals comprising a skill-endorsement graph page rank, a skill-profile textual similarity and member seniority.

3. The method of claim 2 , wherein the machine learning algorithm uses logistic regression.

4. The method of claim 1 , wherein the factorizing includes mapping the member matrix and the skill matrix in a K-dimensional latent space.

5. The method of claim 1 , further comprising assigning an expertise score of zero to a particular combination of member profile and skill if a member corresponding to the member profile has a calculated expertise score in the skill below a threshold.

6. The method of claim 5 , further comprising training a machine learning model to make suggestions of query modifications based on one or more facets, the training using usage data indicative of whether users added a suggestion, deleted a suggestion, or ignored a suggestion.

7. The method of claim 2 , wherein the supervised machine learning algorithm classifies each member profile as either an expert or a non-expert in a corresponding skill, based at least partially on publications in respective fields.

8. The method of claim 6 , wherein the training includes using a log from a query-based search.

9. The method of claim 1 , further comprising: calculating a dot product of the expertise matrix and the skill matrix.

10. The method of claim 1 , further comprising:

ranking the one or more result member profiles by:

computing a career path similarity between each of the plurality of ideal candidate member profiles and each of the one or more result member profiles;

feeding the computed career path similarity and each of the one or more result member profile into a machine learned model trained by a machine learning algorithm to output a similarity score based on career path similarity, the similarity score based on trajectory information including a model of a member profile as a sequence of nodes, wherein each node records company, title, industry, and time duration information for a particular position of a corresponding member's career; and

ordering the one or more result member profile by their corresponding similarity score.

11. The method of claim 10 , wherein the machine learned model is a generalized linear model.

12. The method of claim 10 , wherein the machine learned model models each member profile as a sequence of nodes, wherein each node records information about a different position in a members career.

13. A system comprising:

a computer-readable medium having instructions stored thereon, which, when executed by a processor, cause the system to:

receive, via the graphical user interface, a selection of a plurality of the identifications, each of the plurality of identifications corresponding to a different member profile from the plurality of member profiles and identifying the corresponding member profile as one of a plurality of ideal candidate member profiles;

extract, by a hardware processor, one or more attributes from the plurality of ideal candidate member profiles;

generate, by the hardware processor, a search query based on the extracted one or more attributes, the generating comprising:

form an expertise matrix, the expertise matrix containing an expertise score for each of a plurality of combinations of member profiles and skills for members of the online service, the expertise score for a given combination of member profile and skill indicating a level of expertise for a corresponding member in a corresponding skill;

factorize the expertise matrix into a member matrix and a skills matrix;

rank the one or more attributes from the plurality of ideal candidate member profiles based on the expertise matrix and the skills matrix; and

select top N attributes based on the ranking;

modify the generated search query to include the selected top N attributes; and

perform a search on member profiles in the online service using the modified generated search query, returning one or more result member profiles.

14. The system of claim 13 , wherein the generating a search query comprises:

aggregating the one or more attributes across the one or more ideal candidate member profiles; and

selecting top attributes most similar to attributes of all of the one or more ideal candidate member profiles.

15. The system of claim 13 , wherein the generating includes:

calculating a set of expertise scores using a statistical model and a set of features regarding skills of the one or more ideal candidate member profiles, the statistical model trained using a machine learning algorithm;

using the expertise scores to rank skills of the one or more ideal candidate member profiles; and

adding one or more top ranked skills to the search query.

16. The system of claim 15 , wherein the statistical model is a logistic regression model.

17. The system of claim 13 , wherein the generating includes:

referencing a second browse map, the second browse map identifying companies having employees whose corresponding member profiles have been viewed during a same browsing session, to identify co-viewed companies; and

adding, to the search query, one or more companies that have been co-viewed during the same browsing session as a company identified in one or more of the ideal candidate member profiles, using the second browse map.

18. The system of claim 13 , wherein the instructions further cause the system to submit the search query to a searcher for editing.

19. A non-transitory machine-readable storage medium comprising instructions, which when implemented by one or more machines, cause the one or more machines to perform operations comprising:

causing display, in a graphical user interface, of identifications of a plurality of member profiles, each of the member profiles comprising a page, renderable by a user interface, containing attributes of a different member of an online service;

receiving, via the graphical user interface, a selection of a plurality of the identifications, each of the plurality of identifications corresponding to a different member profile from the plurality of member profiles and identifying the corresponding member profile as one of a plurality of ideal candidate member profiles;

extracting, by a hardware processor, one or more attributes from the plurality of ideal candidate member profiles;

generating, by the hardware processor, a search query based on the extracted one or more attributes, the generating comprising

forming an expertise matrix, the expertise matrix containing an expertise score for each of a plurality of combinations of member profiles and skills for members of the online service, the expertise score for a given combination of member profile and skill indicating a level of expertise for a corresponding member in a corresponding skill;

factorizing the expertise matrix into a member matrix and a skills matrix;

ranking the one or more attributes from the plurality of ideal candidate member profiles based on the expertise matrix and the skills matrix; and

selecting top N attributes based on the ranking;

modifying the generated search query to include the selected top N attributes; and

performing a search on member profiles in the online service using the modified generated search query, returning one or more result member profiles.

20. The non-transitory machine-readable storage medium of claim 19 , wherein the generating the search query comprises:

aggregating the one or more attributes across the one or more ideal candidate member profiles; and

selecting top attributes most similar to attributes of all of the one or more ideal candidate member profiles.

21. The non-transitory machine-readable storage medium of claim 19 , wherein the generating includes:

calculating a set of expertise scores using a statistical model and a set of features regarding skills of the one or more ideal candidate member profiles, the statistical model trained using a machine learning algorithm;

using the expertise scores to rank skills of the one or more ideal candidate member profiles; and

adding one or more top ranked skills to the search query.

22. The non-transitory machine-readable storage medium of claim 21 , wherein the statistical model is a logistic regression model.

23. The non-transitory machine-readable storage medium of claim 19 , wherein the generating includes:

referencing a second browse map, the second browse map identifying companies having employees whose corresponding member profiles have been viewed during a same browsing session, to identify co-viewed companies; and

adding, to the search query, one or more companies that have been co-viewed during the same browsing session as a company identified in one or more of the ideal candidate member profiles, using the second browse map.