Building Multilingual Semantic Networks with Non-Expert Contributions over the Web Page: 4
The following text was automatically extracted from the image on this page using optical character recognition software:
facility that allows the user to request another random
list of synsets for her to choose from. Once a synset
is selected, and the user constructs the equivalent Ro-
manian synset, the synset is removed from the pool of
"available" synsets and moved into a different set con-
taining synsets to be validated.
Phase I directs the user to specify the Romanian words
that belong in the chosen synset either by translating
the words in the English synset or by providing words
that are not direct translations of any of the English
words. To speed up the process, RSDNET uses an in-
ternal English to Romanian dictionary (Section 3.2) to
translate the English words. These are only suggestions
for the user because the system cannot determine if the
translations are correct in the context of the synset. For
example, RSDNET translates the word plant as planta
which is correct if the synset refers to living organism,
but not if the synset refers to industrial plant. It is the
role of the contributor to decide on the right translation
for a given synset word. In the example described above
for Phase 0, RSDNET correctly suggests the words de-
viere, deviere, digresiune, abatere respectively for the
first four words.
Figure 1 shows a screen shot of Phase 1. As an added
benefit of this phase, the translations validated by the
user create relationships between English and Roma-
nian words within the context of the synset. These rela-
tionships are stored in a database and could eventually
lead to a semantic English-Romanian dictionary.
Please Validate These
Translations 0tunon' p ssi ,a.
Add New Word
Figure 1: The top section of Phase 1 keeps the
English synset in view while the bottom section
directs the user to make changes to Romanian
translations if necessary.
In Phase 2, RSDNET uses an internal Romanian Dic-
tionary (Section 3.3) and a textual corpus (Section 3.4)
to suggest definitions and samples respectively. As in
Phase 1, these suggestions may be out of context and
the user needs to validate them or add new entries.
Phase 2 also provides a facility that allows the user
to use popular search engines on the Web to retrieve
In the earlier "diversion" example, RSDNET does not
suggest any definitions, so the user enters one: o schim-
bare (in aten tie, a drumului, etc.). The system provides
some samples from which the user selects o deviere a
drumului, o digresiune de la subiect, and o abatere de la
calea cea dreapta.
Phase 3 directs the user to review the synset he or
she has created to look for errors. Additionally, since
Phases 1 and 2 allow the contributor to use simple html
markup and to represent special characters using html,
this phase gives a visual confirmation that the right
markup has been used. From this phase, the user can
finally submit his or her contribution to the RSDNET
In each of these phases, an online help system provides
instructions for the non-expert user. This system also
provides a reference for escape sequences that can be
used to represent special Romanian characters. For ex-
ample, the sequence '\a' is used to represent "i.
4.2 The Administrative System
A major concern with Web based data collection is the
introduction of errors into the database because of a
user's oversight, malicious intent or limited knowledge
of the language. A color-coded administrative Web
page was designed to allow select individuals to review
and validate the entries into RSDNET, correcting or
deleting them if necessary. Figure 2 shows a screenshot
of this page, which also shows the administrator the
original English synset and the relationships that have
been created between English and Romanian words. A
field in the database indicates which synsets have been
4.3 The Scoring System
RSDNET, like similar projects at teach-computers.org,
uses a rewards program to motivate users to make more
contributions. When the user submits his or her synset
in Phase 3, a score is computed based on the number
of words in the synset. This is a very simple measure
and can be thought of as a measure of the size of the
synset. Future scoring schemes may consider the num-
ber of definitions and samples provided and penalize the
user for incorrect entries. To attract new users and in-
crease retention, we are giving away prizes, on a weekly
or monthly basis.
4.4 Other Interface Features
Only users that are registered with RSDNET can im-
prove their scores and win prizes. RSDNET provides a
simple interface for registering with the system and up-
dating personal information such as an email address.
Here’s what’s next.
This paper can be searched. Note: Results may vary based on the legibility of text within the document.
Tools / Downloads
Get a copy of this page or view the extracted text.
Citing and Sharing
Basic information for referencing this web page. We also provide extended guidance on usage rights, references, copying or embedding.
Reference the current page of this Paper.
Ayewah, Nathanial; Mihalcea, Rada, 1974- & Nastase, Vivi. Building Multilingual Semantic Networks with Non-Expert Contributions over the Web, paper, November 2003; (digital.library.unt.edu/ark:/67531/metadc30947/m1/4/: accessed February 19, 2019), University of North Texas Libraries, Digital Library, digital.library.unt.edu; crediting UNT College of Engineering.