Jan 6
=====

tenBosch/etal:2008a,b
---------------------

In the 'Computational model of language acquisition' paper the
database is said to include both adult directed and infant directed
speech examples. Were the learner models in the experiments given both
forms of speech and if so, how would this affect the learner's ability
to recall target words when given adult directed vs infant directed
speech?

In figure 2, what is meant by "speaker-blocked order"?

According to the experiment the critical mass for word representation
is 30-50 words. The authors focus on the number of tokens but isn't
the time span also and important factor, i.e. the period of time that
these 50 tokens have spread out? I think it makes sense to assume that
if a learner hears 50 tokens in a very short period of time, that
critical mass alone won’t be enough.

When more abstract general word representations are formed, do more
specific fade away or form complex structures such as inheritance
trees, that is hierarchical networks in which lower nodes, can be
assumed to inherit information from more abstract dominating higher
nodes?

What is meant by gated input ?

How do the two speech modes (ADS and IDS) differ?

In their models, they involved "tag" in the input that represent
referent of words. As it is described in the paper, it seems that they
used only words that have a visual representation, such as physical
objects - ball. However, infants also hear and learn number of words
that have no visual referent but refer to a concept, e.g., Mama loves
you. How could we design then a tag for conceptual words?