Information Technology Reference
In-Depth Information
the system because it shows how the incremental model is able to maintain the
previous lexicon consensus. We think this is a key factor in the language develop-
ing and acquisition process because the language probably would not evolve as
quick and reliable as it really does if the consensus were not maintained through
generations.
6 Conclusions
Multi-agent systems would probably improve their skills and usefulness if they
were able to communicate each other with a complex communication system.
First attempts in artificial communication is being focused in developing a vo-
cabulary or lexicon. With a consensual lexicon we can address more complex and
more ambitious goals about artificial language. In this work we propose an incre-
mental model to develop a consensus lexicon in a population of simulated agents
by means of a hybrid algorithm which merge grammatical evolution with seman-
tic rules and reinforcement learning. Although the experiments we addressed in
this document could be developed with other methods we use grammar because
in a future we think to enhance the system in order to generate a more com-
plicated language with a syntactical structure. In this case, we think grammars
are the most ecient mechanism. Meanwhile, this work allows us to stablish an
starting point.
The proposed incremental model allows to add new objects and agents to
the environment when a consensus lexicon has emerged for an initial set of
objects. Each object is named with a symbol and a meaning for the object is
associated with the symbol by means of language games. Agents are endowed
of basic behaviours for sensing and acting and two specialized behaviours for
associating meaning and symbols and for playing language games. Language
games are only developed when two agents find each other in order to simulate
a realistic situation. Results show how the incremental model is able to find
solutions for consensual vocabularies with 4, 6, 8, 10, 12 and 14 symbols and
agents with a limited set of resources (a low value of individuals in the population
and few generations). Thanks to grammars the solutions are more readable and
understandable as well. The incremental model shows how the consensual lexicon
can be maintained through configurations and how the previous symbol-meaning
pairs are not lost.
Acknowledgements
This work has been funded by the Spanish Ministry of Science under contract
TIN2008-06491-C04-03.
References
1. Briscoe, E.J.: Linguistic Evolution through Language Acquisition: formal and com-
putational models. Cambridge University Press, Cambridge (2002)
Search WWH ::
Custom Search