You are on page 1of 6

User Id

No

Use Algorithm for


user without
Confirmations and
Rejections

User has
confirmed & rejected
publications

Yes

Use Algorithm for


user with
Confirmations and
Rejections

Filter By Accepted
NameVariants
Combinations

Exclude by Date
of Birth

Seeded by User name variants + AuthorNames


taken from already confirmed publications (The
more occurrences in confirmed publications the
higher the weight of the name variant)

If PublicationDate not null suggest only if later


than Users's Date of Birth

Phase 1

Boost by Email

Boost by User
Affiliation /
Position

Yes

User has
Domain
+ Field
+ Specialty

Verified user emails

As affiliations are defined and stored in different


manners depending on the publication source, this
requires a non-exact match search, giving more
relevancy to the more specific terms
(depending on the term frequency and the inverted
document frequency)

No
Requires us to
process all
publications in order
to extract their topics

Phase 2
As this process
requires further
analysis, it
would be
included in a
2nd iteration

Search Domain +
Field + Specialty
in Pub. Topics

User is
following other
users

No

Yes

Search topics of
following users in
Pub. Topics

Take 20 most
relevant
publications

These topics are then


matched to a user's
domain, field or
specialty.
In the cases where a
user has not filled in
the domain field and
specialty, but has
followed other users,
then we can search
using the topics of
these users.

We provide the 'n' most relevant publications per time


period (e.g: per month)

Phase 1

Input: UserId

Exclude
Confirmed/Rejected

Exclude by Date
of Birth

Filter By Accepted
NameVariants
Combinations

Boost by Email

If PublicationDate not null suggest only if later


than Users's Date of Birth

Seeded by User name variants +


AuthorNames taken from already
confirmed publications (The
more occurrences in confirmed
publications the higher the
weight of the name variant)

User email + Confirmed


publications email

Boost by
Affiliations

User Affiliations (Affiliation + Positions) +


ConfirmedPublications affiliations
(Weighted by frequency) RejectedPublications Affiliations (Weighted
by frequency)

Boost by
Keywords

ConfirmedPublications Keywords
(Weighted by frequency) RejectedPublications Affiliations (Weighted
by frequency)

Boost by
Coauthors

Take user's coauthors from confirmed


publications (UserId and particular name
linked) and boost publications confirmed by
them (UserId) or containing author
matching those names

Boost by External
Repository
AuthorId

Scopus associates Ids to authors, take Ids


from confirmed and rejected publications
and boost or downgrade publications if the
contain those Ids.

Phase 1

BoostingK = F (If, Uf) Where If


(Keyword inverted frequency
in all System and Uf is User
keyword frequency in
confirmed/Rejected
publications)

You might also like