A key ambition of AI is to render computers able to evolve and interact
with the real world. This can be made possible only if the machine is able
to produce an interpretation of its available modalities (image, audio,
text, etc.) which can be used to support reasoning and taking appropriate
actions. Computational linguists use the term “semantics” to refer to the
possible interpretations of natural language expressions and there is
recent work in “learning semantics” – finding (in an automated way) these
interpretations. However, “semantics” are not restricted to the natural
language (and speech) modality, and are also pertinent to visual
modalities. Hence, knowing visual concepts and common relationships
between them would certainly provide a leap forward in scene analysis and
in image parsing akin to the improvement that language phrase
interpretations would bring to data mining, information extraction or
automatic translation, to name a few.
Progress in learning semantics has been slow mainly because this involves
sophisticated models which are hard to train, especially since they seem
to require large quantities of precisely annotated training data. However,
recent advances in learning with weak, limited and indirect supervision
led to the emergence of a new body of research in semantics based on
multi-task/transfer learning, on learning with semi/ambiguous/indirect
supervision or even with no supervision at all. Hence, this special issue
invites paper submissions on recent work for learning semantics of natural
language, vision, speech, etc.
Papers should address at least some of the following questions:
– How should meaning representations be structured to be easily
interpretable by a computer and still express rich and complex knowledge?
– What is a realistic supervision setting for learning semantics?
– How can we learn sophisticated representations with limited supervision?
– How can we jointly infer semantics from several modalities?
Submission deadline: May 1, 2012
First review results: July 30, 2012
Final drafts: September 30, 2012
Papers must be submitted online, selecting the article type that indicates
this special issue. Peer reviews will follow the standard Machine Learning
journal review process. It is the policy of the Machine Learning journal
that no submission, or substantially overlapping submission, be published
or be under review at another journal or conference at any time during the
review process. Papers extending previously published conference papers
are acceptable, as long as the journal submission provides a significant
contribution beyond the conference paper, and the overlap is described
clearly at the beginning of the journal submission. Complete manuscripts
of full length are expected, following the MLJ guidelines in
Antoine Bordes (email@example.com)
Léon Bottou (firstname.lastname@example.org)
Ronan Collobert (email@example.com)
Dan Roth (firstname.lastname@example.org)
Jason Weston (email@example.com)
Luke Zettlemoyer (firstname.lastname@example.org)