Information Technology Reference
In-Depth Information
refer to requests participants received. Clicks represent recommendations which
users followed by clicking. CTR describes the ratio of clicks to impressions. ORP's
goal is to discover recommendation algorithms which maximize the CTR. Figure
6.2
shows an exemplary dashboard illustrating the clicks, impressions, and CTR
graphically and as table. Additionally, ORP provides a leaderboard where partic-
ipants can compare their performance to others.
6.3.5 Participation
There are plenty of reasons for both researchers and practitioners to contribute to
ORP. There are hardly any opportunities to get access to actual interactions between
users and items. Thus, ORP provides a unique way to evaluate recommendation
algorithms. Existing implementations of application programming interfaces facili-
tate getting started. Plista as well as members of different research institutions have
contributed implementations in Java,
1
PHP,
2
python,
3
and Node.js.
4
In addition, we
have organized a variety of workshops and competition where researchers along
with practitioners published results obtained through ORP. These events include the
“International News Recommendation Workshop and Challenge”
5
[
59
], the “Work-
shop on Benchmarking Adaptive Retrieval and Recommender Systems”
6
[
15
], and
CLEF NEWSREEL
, the “News Recommendation Evaluation Lab”
7
[
10
,
28
,
31
].
The Open Recommendation Platform provides a unique chance for researchers to
evaluate recommendation algorithms with actual user feedback. We have seen which
technical requirements it entails. Systems have to reply to request within 100 ms. This
prevents plista's performance from dropping below a level where customers suffer
substantial losses. ORP commits to open standards with respect to data interchange
and interfaces. Researchers and practitioners have already contributed implementa-
tions in a variety of programming languages. We encourage researchers to start or
continue contributing recommendation algorithms to discover new ways to support
users struggling to find relevant news.
Search WWH ::
Custom Search