RUNTIME EVALUATION OF THE OFFLINE-GENERATED RESULTS
A recommendation system is configured to generate relevance scores for job postings with respect to a subject member profile preemptively, prior to detecting that the associated member is logged in into the online social network system. The online ranker part of the recommendation system selectively utilizes these pre-computed values at runtime only if they have been generated within a predetermined period of time. The online ranker discards or ignores the offline relevance values generated with respect to a subject member profile if it detects changes to the subject member profile that occurred subsequent to the generation of the offline relevance values.
1 . A computer-implemented method comprising:
generating offline relevance scores for job postings provided in an online social network system with respect to a subject member profile maintained in the online social network system, a relevance score generated with respect to the subject member profile and a certain job posting reflecting probability of a member represented by the subject member profile applying for a job represented by he certain job posting;
storing the offline relevance scores in a key-value store, the key-value store accessible to a presentation set generator, the presentation set generator configured to generate a presentation set of job postings for the subject member profile in response to detecting that a member represented by the subject member profile is logged in into the online social network system;
using at least one processor, in response to detecting that the member represented by the subject member profile is logged in into the online social network system, by the presentation set generator:
determining a staleness value, the staleness value reflecting a difference between a current date and a creation date for the offline relevance scores, and
using the offline relevance scores to generate a presentation set of job postings only if the staleness value is less than a value of a staleness indicator; and
causing displaying of a reference to a job posting from the presentation set of job postings on a display device associated with the member.
2 . The method of claim 1 , comprising:
monitoring arrival of new job postings into the online social network system;
based on the monitoring, determining frequency of arrival of new job postings into the online social network system; and
setting the value of the staleness indicator based on the determined frequency of arrival of new job postings, the value of the staleness indicator being set to a lesser value as the frequency of arrival of new job postings increases.
3 . The method of claim 2 , wherein the monitoring of arrival of new job postings into the online social network system is only with respect to new job posting characterized by a predetermined aspect.
4 . The method of claim 3 , wherein the predetermined aspect is a geographic location that characterizes the subject member profile.
5 . The method of claim 3 , wherein the predetermined aspect is an industry that characterizes the subject member profile.
6 . The method of claim 1 , comprising, in response to detecting that the member represented by the subject member profile is logged in into the online social network system:
determining that the subject member profile has been changed subsequent to the generating of the offline relevance scores; and
discarding the offline relevance scores generated with respect to the subject member profile.
7 . The method of claim 1 , comprising, in response to determining that the subject member profile has been changed subsequent to the generating of the offline relevance scores:
using the presentation set generator to generate runtime relevance values for job postings with respect to the subject member profiles; and
using the runtime relevance values to select job postings for inclusion in the presentation set of job postings.
8 . The method of claim 1 , wherein the using the offline relevance scores to generate a presentation set of job postings comprises selecting a job posting for inclusion into the presentation set of job postings based on its associated score from the offline relevance scores.
9 . The method of claim 8 , wherein the using of the presentation set generator to generate runtime relevance values for job postings with respect to the subject member profiles comprises utilizing a runtime relevance model.
10 . The method of claim 9 , wherein the runtime relevance model utilizes an approach that is different from an approach utilized by an offline relevance model, the offline relevance model used to generate the offline relevance scores.
11 . A computer-implemented system comprising:
an offline relevance scores generator, implemented using at least one processor, to generate offline relevance scores for job postings provided in an online social network system with respect to a subject member profile maintained in the online social network system, a relevance score generated with respect to the subject member profile and a certain job posting reflecting probability of a member represented by the subject member profile applying for a job represented by the certain job posting;
a key-value store, implemented using at least one processor, to store the offline relevance scores;
a presentation set generator, implemented using at least one processor, to, in response to detecting that the member represented by the subject member profile is logged in into the online social network system, to:
access the offline relevance scores stored in the key-value store;
determine a staleness value, the staleness value reflecting a difference between a current date and a creation date for the offline relevance scores, and
use the offline relevance scores to generate a presentation set of job postings only if the staleness value is less than a value of a staleness indicator; and
a presentation module, implemented using at least one processor, to cause displaying of a reference to a job posting from the presentation set of job postings on a display device associated with the member.
12 . The system of claim 11 , comprising a new jobs arrival monitor, implemented using at least one processor, to:
monitor arrival of new job postings into the online social network system;
based on the monitoring, determine frequency of arrival of new job postings into the online social network system; and
set the value of the staleness indicator based on the determined frequency of arrival of new job postings, the value of the staleness indicator being set to a lesser value as the frequency of arrival of new job postings increases.
13 . The system of claim 12 , wherein the monitoring of arrival of new job postings into the online social network system is only with respect to new job posting characterized by a predetermined aspect.
14 . The system of claim 13 , wherein the predetermined aspect is a geographic location that characterizes the subject member profile.
15 . The system of claim 13 , wherein the predetermined aspect is an industry that characterizes the subject member profile.
16 . The system of claim 11 , wherein the presentation set generator, in response to detecting that the member represented by the subject member profile is logged in into the online social network system, is to:
determine that the subject member profile has been changed subsequent to the generating of the offline relevance scores; and
discard the offline relevance scores generated with respect to the subject member profile.
17 . The system of claim 11 , wherein the presentation set generator, in response to determining that the subject member profile has been changed subsequent to the generating of the offline relevance scores, is to:
generate runtime relevance values for job postings with respect to the subject member profiles; and
use the runtime relevance values to select job postings for inclusion in the presentation set of job postings.
18 . The system of claim 11 , wherein the using the offline relevance scores to generate a presentation set of job postings comprises selecting a job posting for inclusion into the presentation set of job postings based on its associated score from the offline relevance scores.
19 . The system of claim 18 , wherein the presentation set generator executes a runtime relevance model to generate the runtime relevance values, the runtime relevance model utilizes an approach that is different from an approach utilized by an offline relevance model, the offline relevance model used to generate the offline relevance scores.
20 . A machine-readable non-transitory storage medium having instruction data executable by a machine to cause the machine to perform operations comprising:
generating offline relevance scores for job postings provided in an online social network system with respect to a subject member profile maintained in the online social network system, a relevance score generated with respect to the subject member profile and a certain job posting reflecting probability of a member represented by the subject member profile applying for a job represented by the certain job posting;
storing the offline relevance scores in a key-value store, the key-value store accessible to a presentation set generator, the presentation set generator configured to generate a presentation set of job postings for the subject member profile in response to detecting that a member represented by the subject member profile is logged in into the online social network system;
in response to detecting that the member represented by the subject member profile is logged in into the online social network system, by the presentation set generator:
determining a staleness value, the staleness value reflecting a difference between a current date and a creation date for the offline relevance scores, and
using the offline relevance scores to generate a presentation set of job postings only if the staleness value is less than a value of a staleness indicator; and
causing displaying of a reference to a job posting from the presentation set of job postings on a display device associated with the member.