Steels, L.: Self-organizing vocabularies. In: Langton, C. (ed.): Proceedings of the Conference on Artificial Life V (Alife V) (Nara, Japan) (1996)  Home/Search   Document Not in Database   Summary   Related Articles   Check

....appropriate for direct translation into human thought and reasoning based on a system that organizes its own memories based on instinctive, hardwired spatial perception. ASPARC develops these concepts as a foundation for communication and spatial reasoning as an emergent phenomenon (Steels [13], Steels[14] but based on the assumption that the framework for spatial perception is biological in people and should be endemic to the perception and communication systems of artificial entities. By framing their perceptions in the cognitive model (Davidson [15] to match our own, we frame their ....

L. Steels, "Self-organizing vocabularies," in Proceedings of Alife V, C. Langton, Ed., Nara Japan, 1996.

....paradigms is clearly huge. A related discussion of modelcoordination can be found in [AdJM00, Ago00] Our approach is closed in spirit to the framework of the language games, namely, models of language change and language evolution in populations of communicating agents. Some reference are [Ste96, Bat00, Kir00] Language games, however, focus much more on language creation and evoution than our models. Moreover, a formal comparison with language games would only be possible if we enriched our basic paradigm with an explicit system of pay o s, so as to explicitly relate agent s actions ....

....with an explicit system of pay o s, so as to explicitly relate agent s actions and beliefs. Other related models include semantic evaluation games (see for instance [HS97] for a survey) dialogue games for validity, and especially naming games a kind of language games rst introduced by Steels [Ste96] As in our paradigm, in naming games the meaning play an important role. A speaker (the source) chooses a meaning and a form to express that mean11 ing, and a hearer (the seeker) makes, based on the received form, a guess of what is meant. The hearer receives direct feedback on whether its ....

L. Steels. Self-organizing vocabularies. In C. Langton, editor, Proceedings of the V Alife Conference, Nara, Japan, 1996.

....transmission and production, tolerance of minor linguistic variation, and a certain rate of population change. The authors first produced completely deterministic agents, showing that linguistic variation was not tolerated. As a result, the final model (reported in [8] 9] and prototypically in [7]) of agent internal linguistic processing is something of a hybrid, with probabilistic measures grafted onto what is essentially a deterministic backbone, while linguistic utterances in the system are represented by a random sequence of characters. We present a new system for investigating lexicon ....

....undertake research involving multiple software agents without a principled framework within which the agents can interact; that is, the precise details of all possible interactions between agents and their environment (and or each other) must be specified in an unambiguous fashion. To this end, [7] introduces the naming game, an austere paradigm for interaction tailored especially for the development, transmission, and evolution of a lexicon in either a static or dynamic population of agents. The naming game is appropriate for a population of agents and a number objects 1 ; an ....

Steels, Luc. "Self-organizing vocabularies" In: Langton, C. (ed.) (1996) Proceedings of Alife V, Nara Japan.

....by this group ( 8] four factors are postulated as necessary for effective lexical change to occur; the last of these is population flux. We perform our investigations into the nature of lexical change within the framework of a multi agent simulation using the naming game (first introduced in [6]) as a paradigm for agent interaction; however, we have augmented our implementation with a more realistic phonetic model than is present in [8] which was the starting point for our simulation. Furthermore, the behavioural mechanisms of our agents are based on an internal linguistic model which ....

....mechanisms of our agents are based on an internal linguistic model which is connectionist in nature. 2 The Naming Game Research involving multiple software agents requires unambiguously specifying the precise details of all possible agent agent and agent environment interactions. To this end, [6] introduces the naming game, an austere paradigm for interaction tailored especially for the development, transmission, and evolution of a lexicon. The naming game is appropriate for any simulation with a fixed number of agents and objects; an interaction proceeds as follows: Two agents are ....

[Article contains additional citation context not shown here]

Steels, Luc. "Self-organizing vocabularies" In: Langton, C(ed.) Proceedings of Alife V, Nara Japan.1996.

....ways to present the concept of elementary equivalence in model theory in terms of Ehrenfeucht games. Other games in linguistic include semantic evaluation games (see for instance [HS97] for a survey) dialogue games for validity, and naming games kind of language games rst introduced by Steels [Ste96] In dialogue games, the validity of some given formula is examined in terms of a two person, perfect information game. Naming games are interactions between two agents, a speaker and a hearer, in which the speaker identi es an object using a name, and the hearer eventually agrees on the name as ....

L. Steels. Self-organizing vocabularies. In C. Langton, editor, Proceedings of the V Alife Conference, Nara, Japan, 1996.

....a simple abstract world. Walker and Wooldridge [23] study the emergence of conventions in multi agent systems as a function of various hard coded strategy update functions, including update functions where agents communicate to exchange memories of observed strategies by other agents. Luc Steels [20] show that vocabulary can evolve through the principle of self organization. A set of agents create their own vocabulary in a random manner, yet self organization occurs because the agents are coupled in the sense that they must conform to a common vocabulary in order to cooperate through ....

....action at each time step from the set N, S, E, W using a uniform random distribution. The Linear Prey picks a random direction at the beginning of a trial and continues in that direction for the duration of the scenario. It has been shown that the Linear Prey can be a difficult prey to capture [20], 10] because it does not stay localized in an area. In our simulations this is an even more difficult prey to capture because the prey and predators move at the same speed. 3 Communication This paper studies a simple framework in which all predator agents communicate simultaneously to a ....

[Article contains additional citation context not shown here]

Luc Steels. Self-organizing vocabularies. In Proceedings of Alife V, 1996.

....and (iii) a learning algorithm. Linguistic Representation With representation we mean here a formalism to represent the linguistic abilities of agents, ranging from recurrent neural networks (Batali, 1998) or rewriting grammars (Kirby, 2000) to a simple associative memory (Hurford, 1989; Steels, 1996; Oliphant Batali, 1996; De Boer, 1999; Kaplan, 2000) In the model described in this paper, we will use a simple list of associations between linguistics forms (words) and their meanings. Each association has a score that represents the cost (or inversed strength) of that association and ....

....this form is the form that it originally used, the game is successful. Otherwise it is a failure. In the imitation game meanings play no role. It serves as a model system for studying the interaction between forms, and the emergent maximisation of the distance between them. In the naming game (Steels, 1996), the meanings do play a role. The speaker chooses a meaning and a form to express that meaning, and the hearer makes, based on the received form, a guess of what is meant. The hearer then receives feedback on the intended meaning, i.e. whether its guess was correct. The game is a success if the ....

[Article contains additional citation context not shown here]

STEELS, L. (1996). Self-organizing vocabularies. In: Proceedings of Alife V ( Langton, C., ed.).

....speci cs of natural syntax, such as the position of auxiliary verbs in an English sentence. Rather, these models can give insights in the origins of some general but fundamental aspects of natural language. e.g. the facts that human language is (in nitely) expressive; for a large part speci c [10, 21, 18] and distinctive [6] compositional [1, 16, 2, 12] recursive [8, 11, 24] diverse on a global scale, but uniform on a local scale; dynamic, constantly subject to innovations. And that languages share universal tendencies [13, 6] and are used for very diverse purposes, including information ....

Luc Steels. Self-organizing vocabularies. In Chris Langton, editor, Proceedings of Alife V, 1996.

....of human grammars. In artificial intelligence, the problem arises in many di#erent settings. A number of studies have emerged where linguistic agents interact with each other in simulated worlds and one studies whether coherent or coordinated communication ultimately emerges (see, for example, [Steels 1996, Steels Vogt 1997, Steels Kaplan 1998, Oliphant Batali 1996, Oliphant 1997, Briscoe 2000, Kirby 1999] Much of this kind of research employs the simulation methodology of Artificial Life. In this paper, we create a mathematical framework for these kinds of problems and derive a number of ....

Steels, L. (1996) Self-organizing vocabularies. In: Proceedings of Alife V, (Langston, C. ed). Nara, Japan.

....The next part introduces the linguistic behavior and the learning process of a single agent. Then the typical behavior of such a system is studied without changes in language contact. Finally an experiment is discussed related to the dynamics of language contact. The , as introduced by Steels [7], is a specific kind of language game which is played between two agents. It is a theoretical model for studying how a shared vocabulary may emerge in a distributed group of agents based only on local interaction. One agent (the speaker) identifies an object by using a name. The game succeeds if ....

....use them themselves, and vice versa. 4. The is equal to the average percentage of shared word meaning pairs between two groups of agents and at time t. A wordmeaning pair is shared if both agents have adopted it as the most prefered word for expressing a particular meaning. In an earlier paper [7], the behavior of groups of agents engaged in naming games has been studied. In closed naming game systems, i.e. where the set of agents and possible meanings is constant, total communicative success is quickly reached (figure 1) An inspection of the lexicon of the invididual agents shows that ....

Steels, L. (1996) Self-organizing Vocabularies. In: Langton, C. (ed.) (1996) Proceedings of the V Alife Conference, Nara Japan.

.... paradigm seems well adapted to the study of these phenomena (see discussion in [6] Interesting results have already been obtained for different areas of language: simple communication codes [2] 13] 15] 22] coordination [4] 5] meaning and class formation [16] conventional lexicons [10][17][20] phonetics [3] and syntax [1] 9] 18] An overview of this approach is given in [19] This paper focuses on the links between class formation and lexicon building. This topic is a fundamental issue in cognitive science, linguistics and philosophy (see [12] Most of the existing ....

....in [19] This paper focuses on the links between class formation and lexicon building. This topic is a fundamental issue in cognitive science, linguistics and philosophy (see [12] Most of the existing computational models study how agents can associate a single word with a referent [13] 15][17][21] 22] The referents can be categories, classes of objects, concepts either predefined in the model or evolving from processes distinct from the language formation itself. This amounts to assuming that an object table corresponds to a concept table independently of the agents naming it a ....

[Article contains additional citation context not shown here]

L. Steels. Self-organizing vocabularies. In C. Langton, editor, Proceeding of Alife V, Nara, Japan, 1996.

....for every meaning. An example of an optimally communicating population is shown in Figure 2. 3 Observational learning The majority of simulation approaches to the study of learned communication use some form of reinforcement learning paradigm (Yanco and Stein, 1993; Hutchins and Hazelhurst, 1995; Steels, 1996; Murciano and Millan, 1996; Schmajuk, 1997) While it seems likely that reinforcement learning is involved to some degree, questions such as how costly the reinforcement signal is, whether one is available at all make it problematic as a primary s a b c 1 1.0 0.0 0.0 Transmission 2 0.0 0.6 ....

Steels, L. (1996). Self-organizing vocabularies. In Proceedings of the V Alife Conference, Nara, Japan.

No context found.

Steels, L.: Self-organizing vocabularies. In: Langton, C. (ed.): Proceedings of the Conference on Artificial Life V (Alife V) (Nara, Japan) (1996)

....as well, for example by using k nearest neighbor, population coding, radial basis functions, etc. Witten and Eibe, 2000] A more sophisticated model of word learning could be used based on maintaining a score between word meaning associations that reflects the success in using a word [Steels, 1996]. Instead we have adopted the simplest possible solutions in order to make the experiments which involve real time interaction with humans possible. If more complex methods would have been adopted they would not fit on the available hardware and the dialog would no longer have a real time ....

Steels, L. (1996). Self-organizing vocabularies. In Langton, C. and Shimohara, T., editors, Proceeding of Alife V, Cambridge, MA. The MIT Press.

....capable to acquire perceptually grounded concepts, a lexicon, and a grammar, and participate in their collective construction. 2 The Talking Heads Experiment We have been working in this direction and have reached already some important milestones, particularly in the area of lexical semantics (Steels, 1996a; 1996b; 1997a) These results have been integrated in the Talking Heads experiment 1 , a large scale public experiment in which visually grounded and situated robotic agents autonomously develop a shared lexicon and ontology from scratch (Steels and Kaplan, 1999b) The agents use vision to make ....

Steels, L. (1996). Self-organizing vocabularies. In: Langton, C. (ed.) (1996) Proceedings of Alife V, Nara Japan. The MIT Press, Cambridge.

....Each game involves a linguistic as well as a non linguistic interaction. The agents have feedback about success and failure and adapt so as to be more successful in future games. We have extensively experimented with a particular type of such a game, called the naming game, first introduced in [Steels, 1996b] The game is played between a speaker and a hearer, randomly drawn from a population of agents. The speaker attempts to identify an object to the hearer, based on pointing and based on using a name. The game succeeds if the hearer guesses correctly the object chosen by the speaker. A speaker may ....

....and the topic, second in terms of noise on the message being transmitted, and finally memory access. Some conclusions and suggestions for further work end the paper. 2 The Naming Game Model The Naming Game, as used in the present paper, is an enriched version of a model first presented in [Steels, 1996b] We assume a set of agents A where each agent a 2 A has contact with a set of objects. These objects constitute a set of meanings to be expressed M = fm 1 ; mng. All the experiments in this paper involve a population of 20 agents and 10 meanings. A form is a sequence of letters drawn from a ....

L. Steels. Self-organizing vocabularies. In C. Langton, editor, Proceeding of Alife V, Nara, Japan, 1996.

....picks the most commonly used successful association. This introduces a positive feedback loop: The more a word gets used, the more successful it will be and therefore the more it solidifies. When there are multiple possibilities a temporary struggle goes on until one association wins (see fig. 1) [11]. So we see three mechanisms: propagation, when agents adopt wordmeaning associations from others, creation, when an agent generates a new word and associates it with an uncovered feature set, and self organisation, due to the positive feedback mechanism between association selection and success ....

Steels, L. (1996a) Self-organizing vocabularies. In: Langton, C. (ed.) Artificial Life V. Nara, Japan. Addison-Wesley Pub. Co. Redwood City, Ca.

....next three sections of the paper. 2 How a coherent lexicon may arise To investigate concretely how a lexicon may originate, be transmitted from one generation to the next, and evolve, we have developed a minimal model of language use in a dynamically evolving population, called the naming game (Steels, 1996). The naming game has been explored through computational simulations and is related to systems proposed and investigated by (Oliphant, 1996) MacLennan, 1991) Werner and Dyer, 1991) a.o. It has even been implemented on robotic agents who develop autonomously a shared lexicon grounded in their ....

L. Steels. 1996. Self-organizing vocabularies.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

L. Steels. Self-organizing vocabularies. In C. Langton and T. Shimohara, editors, Proceedings of the V Alife Conference, Cambridge, MA, 1996. The MIT Press.

No context found.

Steels, Luc, "Self-organizing vocabularies", In: Langton, C. (ed.) Proceedings of Alife V, Nara Japan, 1996.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC