A Practical Guide to Studying Emergent Communication through Grounded Language Games
This web demonstration accompanies the following paper:
Nevens J., Van Eecke, P. and Beuls K. A Practical Guide to Studying Emergent Communication through Grounded Language Games (2019), In AISB Language Learning for Artificial Agents Symposium , pages 1–8.
Abstract
The question of how an effective and efficient communication system can emerge in a population of agents that need to solve a particular task attracts more and more attention from researchers in many fields, including artificial intelligence, linguistics and statistical physics. A common methodology for studying this question consists of carrying out multi-agent experiments in which a population of agents takes part in a series of scripted and task-oriented communicative interactions, called ‘language games’. While each individual language game is typically played by two agents in the population, a large series of games allows the population to converge on a shared communication system. Setting up an experiment in which a rich system for communicating about the real world emerges is a major enterprise; as it requires a variety of software components for running multi-agent experiments, for interacting with sensors and actuators, for conceptualising and interpreting semantic structures, and for mapping between these semantic structures and linguistic utterances. The aim of this paper is twofold. On the one hand, it introduces a high-level robot interface that extends the Babel software system, presenting for the first time a toolkit that provides flexible modules for dealing with each subtask involved in running advanced grounded language game experiments. On the other hand, it provides a practical guide to using the toolkit for implementing such experiments, taking a grounded colour naming game experiment as a didactic example.
The Experiment
The goal of the colour naming game experiment is to show how a shared communication system for referring to objects by their colour can emerge in a population of autonomous agents. The agents start without any concepts or words, perceiving only the average colour values of the objects in the scene. In a real-world setting, transmitting raw sensor values does not lead to successful communication, because the sensors of each agent will always record slightly different values due to differences in the agents’ perspectives on the scene, changes in lighting conditions and in some cases differences in robot morphology. Therefore, concepts and words form the necessary layers to abstract away from sensor data, in order to achieve more robust communication.
In essence, the emergence of colour categories and words through self-organisation can be achieved by the agents dividing their continuous colour space into convex regions that correspond to colour categories that are functional in the world, and by establishing a shared lexicon to refer to each region. An operationalisation of this solution has been proposed in the form of the grounded colour naming game experiment.
Web Demonstration
This web demonstration shows the result of running the grounded colour naming game experiment. The experiment was ran for 1000 interactions, of which every 100th interaction is shown below. The following parameter values were used to generate this demonstration:
Parameter Value
Number of agents
10
Min. context size
2
Max. context size
5
Perceptual deviation
No
Mode
Simulation
Jump to:
Interaction 1 Agent 6 is the speaker Agent 1 is the hearer Running in simulation mode Processed scene:
lab = (85.50, -74.84, -19.40)
lab = (66.66, 104.42, -51.89)
The speaker chooses SENSORY-OBJECT-2 as the topic The speaker tries to conceptualise the topic Conceptualisation failed A new color category is created
lab = (66.66, 104.42, -51.89)
The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-1
lab = (66.66, 104.42, -51.89)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-1) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-1 source-set-1 color-category-2) struct1 (get-context source-set-1) struct0:SOURCEdashSETdash11->struct1:SOURCEdashSETdash12 struct2 (bind color-category color-category-2 color-category-1) struct0:COLORdashCATEGORYdash21->struct2:COLORdashCATEGORYdash23 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: ""
Formulation failed A new construction is created The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-2 source-set-2 color-category-3) struct1 (get-context source-set-2) struct0:SOURCEdashSETdash21->struct1:SOURCEdashSETdash22 struct2 (bind color-category color-category-3 color-category-1) struct0:COLORdashCATEGORYdash31->struct2:COLORdashCATEGORYdash33 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "bipura"
Formulation succeeded The speaker utters the word "bipura" The utterance "bipura" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "bipura" initial structure application process constructional dependencies applied constructions resulting structure
Meaning:
Parsing failed The hearer does not know the word "bipura" The speaker started alignment The speaker finished alignment The hearer started alignment The hearer will learn a new word The hearer looks for a discriminating color category The hearer could not find a discriminating category A new color category is created
lab = (66.66, 104.42, -51.89)
The hearer adopted a new word The hearer finished alignment Interaction failed The speaker's lexicon after 1 interactions: g cluster_cluster6 agent 6 (interaction 1) colour1 bipura (0.5) The hearer's lexicon after 1 interactions: g cluster_cluster1 agent 1 (interaction 1) colour1 bipura (0.5) Jump to top Interaction 100 Agent 2 is the speaker Agent 9 is the hearer Running in simulation mode Processed scene:
lab = (83.02, -137.58, 90.57)
lab = (94.49, -16.24, 103.07)
The speaker chooses SENSORY-OBJECT-357 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-137
lab = (95.26, -33.94, 103.51)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-137) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-125 source-set-125 color-category-174) struct1 (get-context source-set-125) struct0:SOURCEdashSETdash1251->struct1:SOURCEdashSETdash1252 struct2 (bind color-category color-category-174 color-category-137) struct0:COLORdashCATEGORYdash1741->struct2:COLORdashCATEGORYdash1743 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "dufami"
Formulation succeeded The speaker utters the word "dufami" The utterance "dufami" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "dufami" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-1446 ?source-set-479 ?color-category-479) struct1 (get-context ?source-set-479) struct0:varSOURCEdashSETdash4791->struct1:varSOURCEdashSETdash4792 struct2 (bind color-category ?color-category-479 color-category-5) struct0:varCOLORdashCATEGORYdash4791->struct2:varCOLORdashCATEGORYdash4793
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-1446 ?source-set-479 ?color-category-479) struct1 (get-context ?source-set-479) struct0:varSOURCEdashSETdash4791->struct1:varSOURCEdashSETdash4792 struct2 (bind color-category ?color-category-479 color-category-5) struct0:varCOLORdashCATEGORYdash4791->struct2:varCOLORdashCATEGORYdash4793 The hearer tries to interpret the utterance using COLOR-CATEGORY-5
lab = (91.64, -50.28, 99.86)
Interpretation succeeded The hearer points to SENSORY-OBJECT-357 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-137 The agent rewarded the construction: The agent punished the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-5 The agent rewarded the construction: The agent punished the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 100 interactions: g cluster_cluster2 agent 2 (interaction 100) colour1 fedawa (0.5) colour2 bipura (0.4) colour3 dozize (0.4) colour4 dufami (0.7) pupibe (0.3) The hearer's lexicon after 100 interactions: g cluster_cluster9 agent 9 (interaction 100) colour1 sakalu (0.4) colour2 bipura (0.5) colour3 lokolo (0.5) vapufo (0.5) colour4 dufami (0.9) pikono (0.4) Jump to top Interaction 200 Agent 6 is the speaker Agent 1 is the hearer Running in simulation mode Processed scene:
lab = (96.69, -22.51, 105.00)
lab = (61.20, 89.30, 74.92)
lab = (80.72, -134.32, 88.16)
lab = (66.85, 104.60, -51.56)
lab = (86.58, -81.11, -16.63)
The speaker chooses SENSORY-OBJECT-695 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-237
lab = (80.43, -133.97, 88.20)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-237) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-232 source-set-232 color-category-290) struct1 (get-context source-set-232) struct0:SOURCEdashSETdash2321->struct1:SOURCEdashSETdash2322 struct2 (bind color-category color-category-290 color-category-237) struct0:COLORdashCATEGORYdash2901->struct2:COLORdashCATEGORYdash2903 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "salela"
Formulation succeeded The speaker utters the word "salela" The utterance "salela" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "salela" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-3995 ?context-2870 ?x-2989) struct1 (get-context ?context-2870) struct0:varCONTEXTdash28701->struct1:varCONTEXTdash28702 struct2 (bind color-category ?x-2989 color-category-74) struct0:varXdash29891->struct2:varXdash29893
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-3995 ?context-2870 ?x-2989) struct1 (get-context ?context-2870) struct0:varCONTEXTdash28701->struct1:varCONTEXTdash28702 struct2 (bind color-category ?x-2989 color-category-74) struct0:varXdash29891->struct2:varXdash29893 The hearer tries to interpret the utterance using COLOR-CATEGORY-74
lab = (83.02, -137.58, 90.57)
Interpretation succeeded The hearer points to SENSORY-OBJECT-695 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-237 The agent rewarded the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-74 The agent rewarded the construction: The agent punished the construction: The agent punished the construction: The agent punished the construction: The agent punished the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 200 interactions: g cluster_cluster6 agent 6 (interaction 200) colour1 xuxixi (0.7) sakalu (0.5) ziwomo (0.3) sudexo (0.3) xotosu (0.3) colour2 zuvoba (0.6) fufiki (0.3) colour3 bipura (0.8) xeluwi (0.3) colour4 salela (0.5) colour5 dozize (0.6) puwaso (0.4) buboga (0.4) colour6 dufami (0.7) The hearer's lexicon after 200 interactions: g cluster_cluster1 agent 1 (interaction 200) colour1 sudexo (0.6) fefape (0.5) sakalu (0.5) ziwomo (0.5) colour2 fufiki (0.5) dufami (0.5) muzazu (0.3) rimixa (0.3) colour3 bipura (0.9) nuvani (0.3) colour4 salela (0.6) letave (0.4) tuxipi (0.4) mozave (0.4) zovune (0.4) colour5 buboga (0.6) puwaso (0.2) Jump to top Interaction 300 Agent 0 is the speaker Agent 3 is the hearer Running in simulation mode Processed scene:
lab = (81.30, -135.12, 89.00)
lab = (84.44, -72.34, -21.24)
lab = (60.53, 88.53, 74.28)
lab = (32.41, 79.38, -108.12)
The speaker chooses SENSORY-OBJECT-1027 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-110
lab = (59.84, 87.71, 73.57)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-110) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-333 source-set-333 color-category-392) struct1 (get-context source-set-333) struct0:SOURCEdashSETdash3331->struct1:SOURCEdashSETdash3332 struct2 (bind color-category color-category-392 color-category-110) struct0:COLORdashCATEGORYdash3921->struct2:COLORdashCATEGORYdash3923 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "muzazu"
Formulation succeeded The speaker utters the word "muzazu" The utterance "muzazu" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "muzazu" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-6996 ?context-5230 ?x-5378) struct1 (get-context ?context-5230) struct0:varCONTEXTdash52301->struct1:varCONTEXTdash52302 struct2 (bind color-category ?x-5378 color-category-73) struct0:varXdash53781->struct2:varXdash53783
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-6996 ?context-5230 ?x-5378) struct1 (get-context ?context-5230) struct0:varCONTEXTdash52301->struct1:varCONTEXTdash52302 struct2 (bind color-category ?x-5378 color-category-73) struct0:varXdash53781->struct2:varXdash53783 The hearer tries to interpret the utterance using COLOR-CATEGORY-73
lab = (60.87, 88.91, 74.52)
Interpretation succeeded The hearer points to SENSORY-OBJECT-1027 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-110 The agent rewarded the construction: The agent punished the construction: The agent punished the construction: The agent punished the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-73 The agent rewarded the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 300 interactions: g cluster_cluster0 agent 0 (interaction 300) colour1 xuxixi (0.8) fefape (0.1) ziwomo (0.1) sakalu (0.1) sudexo (0.1) colour2 muzazu (0.7) fufiki (0.4) zuvoba (0.2) duvuzo (0.2) colour3 bipura (0.5) xeluwi (0.2) nuvani (0.2) colour4 gomudo (0.5) tuxipi (0.5) zovune (0.5) pikono (0.4) letave (0.4) salela (0.3) colour5 vapufo (0.3) colour6 dufami (1.0) The hearer's lexicon after 300 interactions: g cluster_cluster3 agent 3 (interaction 300) colour1 xuxixi (0.5) ziwomo (0.1) colour2 muzazu (1.0) colour3 mufume (0.5) bipura (0.4) colour4 gomudo (0.5) mozave (0.5) tuxipi (0.1) colour5 buboga (1.0) puwaso (0.5) vapufo (0.3) colour6 dufami (1.0) Jump to top Interaction 400 Agent 6 is the speaker Agent 3 is the hearer Running in simulation mode Processed scene:
lab = (81.58, -135.58, 89.22)
lab = (85.51, -74.81, -19.39)
lab = (61.20, 89.30, 74.93)
The speaker chooses SENSORY-OBJECT-1380 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-19
lab = (86.52, -77.25, -17.65)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-19) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-433 source-set-433 color-category-492) struct1 (get-context source-set-433) struct0:SOURCEdashSETdash4331->struct1:SOURCEdashSETdash4332 struct2 (bind color-category color-category-492 color-category-19) struct0:COLORdashCATEGORYdash4921->struct2:COLORdashCATEGORYdash4923 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "buboga"
Formulation succeeded The speaker utters the word "buboga" The utterance "buboga" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "buboga" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-10017 ?source-set-2286 ?color-category-2286) struct1 (get-context ?source-set-2286) struct0:varSOURCEdashSETdash22861->struct1:varSOURCEdashSETdash22862 struct2 (bind color-category ?color-category-2286 color-category-64) struct0:varCOLORdashCATEGORYdash22861->struct2:varCOLORdashCATEGORYdash22863
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-10017 ?source-set-2286 ?color-category-2286) struct1 (get-context ?source-set-2286) struct0:varSOURCEdashSETdash22861->struct1:varSOURCEdashSETdash22862 struct2 (bind color-category ?color-category-2286 color-category-64) struct0:varCOLORdashCATEGORYdash22861->struct2:varCOLORdashCATEGORYdash22863 The hearer tries to interpret the utterance using COLOR-CATEGORY-64
lab = (86.49, -77.91, -17.53)
Interpretation succeeded The hearer points to SENSORY-OBJECT-1380 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-19 The agent rewarded the construction: The agent punished the construction: The agent punished the construction: The agent punished the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-64 The agent rewarded the construction: The agent punished the construction: The agent punished the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 400 interactions: g cluster_cluster6 agent 6 (interaction 400) colour1 xuxixi (0.4) sakalu (0.3) ziwomo (0.1) colour2 zuvoba (0.7) muzazu (0.5) fufiki (0.1) colour3 bipura (1.0) colour4 gomudo (0.7) salela (0.2) tuxipi (0.2) colour5 buboga (0.8) vapufo (0.4) dozize (0.2) colour6 dufami (1.0) The hearer's lexicon after 400 interactions: g cluster_cluster3 agent 3 (interaction 400) colour1 sakalu (0.7) xuxixi (0.3) colour2 muzazu (1.0) colour3 bipura (0.5) mufume (0.3) colour4 gomudo (0.7) mozave (0.2) colour5 buboga (1.0) puwaso (0.2) colour6 dufami (1.0) Jump to top Interaction 500 Agent 3 is the speaker Agent 8 is the hearer Running in simulation mode Processed scene:
lab = (61.24, 89.18, 74.96)
lab = (95.52, -25.21, 103.85)
The speaker chooses SENSORY-OBJECT-1718 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-73
lab = (60.62, 88.59, 74.30)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-73) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-533 source-set-533 color-category-592) struct1 (get-context source-set-533) struct0:SOURCEdashSETdash5331->struct1:SOURCEdashSETdash5332 struct2 (bind color-category color-category-592 color-category-73) struct0:COLORdashCATEGORYdash5921->struct2:COLORdashCATEGORYdash5923 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "muzazu"
Formulation succeeded The speaker utters the word "muzazu" The utterance "muzazu" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "muzazu" initial structure application process constructional dependencies applied constructions resulting structure
Meaning:
Parsing failed The hearer does not know the word "muzazu" The speaker started alignment The agent punished the construction: The speaker finished alignment The hearer started alignment The hearer will learn a new word The hearer looks for a discriminating color category The hearer found a discriminating category
lab = (61.03, 89.07, 74.66)
The hearer adopted a new word The hearer finished alignment Interaction failed The speaker's lexicon after 500 interactions: g cluster_cluster3 agent 3 (interaction 500) colour1 sakalu (0.9) xuxixi (0.1) colour2 muzazu (0.9) colour3 bipura (0.9) colour4 gomudo (0.8) mozave (0.1) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 500 interactions: g cluster_cluster8 agent 8 (interaction 500) colour1 sakalu (1.0) colour2 zuvoba (1.0) muzazu (0.5) colour3 bipura (1.0) colour4 gomudo (1.0) tuxipi (0.2) colour5 buboga (1.0) colour6 dufami (1.0) Jump to top Interaction 600 Agent 6 is the speaker Agent 0 is the hearer Running in simulation mode Processed scene:
lab = (32.84, 80.07, -109.07)
lab = (94.93, -17.50, 103.46)
lab = (67.24, 104.97, -50.90)
lab = (83.02, -137.58, 90.57)
lab = (86.70, -79.54, -17.33)
The speaker chooses SENSORY-OBJECT-2057 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-84
lab = (32.42, 79.32, -108.06)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-84) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-633 source-set-633 color-category-693) struct1 (get-context source-set-633) struct0:SOURCEdashSETdash6331->struct1:SOURCEdashSETdash6332 struct2 (bind color-category color-category-693 color-category-84) struct0:COLORdashCATEGORYdash6931->struct2:COLORdashCATEGORYdash6933 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "sakalu"
Formulation succeeded The speaker utters the word "sakalu" The utterance "sakalu" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "sakalu" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-14362 ?context-11576 ?x-11762) struct1 (get-context ?context-11576) struct0:varCONTEXTdash115761->struct1:varCONTEXTdash115762 struct2 (bind color-category ?x-11762 color-category-56) struct0:varXdash117621->struct2:varXdash117623
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-14362 ?context-11576 ?x-11762) struct1 (get-context ?context-11576) struct0:varCONTEXTdash115761->struct1:varCONTEXTdash115762 struct2 (bind color-category ?x-11762 color-category-56) struct0:varXdash117621->struct2:varXdash117623 The hearer tries to interpret the utterance using COLOR-CATEGORY-56
lab = (32.15, 78.82, -107.43)
Interpretation succeeded The hearer points to SENSORY-OBJECT-2057 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-84 The agent rewarded the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-56 The agent rewarded the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 600 interactions: g cluster_cluster6 agent 6 (interaction 600) colour1 sakalu (1.0) colour2 zuvoba (0.5) muzazu (0.4) colour3 bipura (1.0) colour4 gomudo (0.7) salela (0.2) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 600 interactions: g cluster_cluster0 agent 0 (interaction 600) colour1 sakalu (1.0) colour2 muzazu (1.0) zuvoba (0.2) colour3 bipura (1.0) colour4 salela (0.4) gomudo (0.1) colour5 buboga (1.0) colour6 dufami (1.0) Jump to top Interaction 700 Agent 4 is the speaker Agent 8 is the hearer Running in simulation mode Processed scene:
lab = (66.95, 104.25, -48.70)
lab = (82.16, -136.36, 89.78)
lab = (32.74, 79.80, -108.69)
lab = (61.25, 89.15, 74.97)
The speaker chooses SENSORY-OBJECT-2410 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-192
lab = (66.86, 104.32, -49.99)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-192) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-733 source-set-733 color-category-793) struct1 (get-context source-set-733) struct0:SOURCEdashSETdash7331->struct1:SOURCEdashSETdash7332 struct2 (bind color-category color-category-793 color-category-192) struct0:COLORdashCATEGORYdash7931->struct2:COLORdashCATEGORYdash7933 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "bipura"
Formulation succeeded The speaker utters the word "bipura" The utterance "bipura" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "bipura" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-16082 ?context-13137 ?x-13328) struct1 (get-context ?context-13137) struct0:varCONTEXTdash131371->struct1:varCONTEXTdash131372 struct2 (bind color-category ?x-13328 color-category-15) struct0:varXdash133281->struct2:varXdash133283
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-16082 ?context-13137 ?x-13328) struct1 (get-context ?context-13137) struct0:varCONTEXTdash131371->struct1:varCONTEXTdash131372 struct2 (bind color-category ?x-13328 color-category-15) struct0:varXdash133281->struct2:varXdash133283 The hearer tries to interpret the utterance using COLOR-CATEGORY-15
lab = (66.50, 104.07, -51.17)
Interpretation succeeded The hearer points to SENSORY-OBJECT-2410 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-192 The agent rewarded the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-15 The agent rewarded the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 700 interactions: g cluster_cluster4 agent 4 (interaction 700) colour1 sakalu (1.0) colour2 muzazu (1.0) zuvoba (0.2) colour3 bipura (1.0) colour4 gomudo (1.0) salela (0.5) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 700 interactions: g cluster_cluster8 agent 8 (interaction 700) colour1 sakalu (1.0) colour2 muzazu (0.7) zuvoba (0.7) colour3 bipura (1.0) colour4 gomudo (0.8) salela (0.6) colour5 buboga (1.0) colour6 dufami (1.0) Jump to top Interaction 800 Agent 2 is the speaker Agent 8 is the hearer Running in simulation mode Processed scene:
lab = (85.24, -74.20, -19.85)
lab = (81.01, -134.73, 88.49)
The speaker chooses SENSORY-OBJECT-2758 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-209
lab = (82.02, -136.11, 89.62)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-209) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-833 source-set-833 color-category-893) struct1 (get-context source-set-833) struct0:SOURCEdashSETdash8331->struct1:SOURCEdashSETdash8332 struct2 (bind color-category color-category-893 color-category-209) struct0:COLORdashCATEGORYdash8931->struct2:COLORdashCATEGORYdash8933 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "salela"
Formulation succeeded The speaker utters the word "salela" The utterance "salela" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "salela" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-17755 ?context-14638 ?x-14833) struct1 (get-context ?context-14638) struct0:varCONTEXTdash146381->struct1:varCONTEXTdash146382 struct2 (bind color-category ?x-14833 color-category-77) struct0:varXdash148331->struct2:varXdash148333
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-17755 ?context-14638 ?x-14833) struct1 (get-context ?context-14638) struct0:varCONTEXTdash146381->struct1:varCONTEXTdash146382 struct2 (bind color-category ?x-14833 color-category-77) struct0:varXdash148331->struct2:varXdash148333 The hearer tries to interpret the utterance using COLOR-CATEGORY-77
lab = (82.55, -136.91, 90.11)
Interpretation succeeded The hearer points to SENSORY-OBJECT-2758 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-209 The agent rewarded the construction: The agent punished the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-77 The agent rewarded the construction: The agent punished the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 800 interactions: g cluster_cluster2 agent 2 (interaction 800) colour1 sakalu (1.0) colour2 muzazu (1.0) zuvoba (0.5) colour3 bipura (1.0) colour4 salela (1.0) gomudo (0.4) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 800 interactions: g cluster_cluster8 agent 8 (interaction 800) colour1 sakalu (1.0) colour2 muzazu (0.9) zuvoba (0.5) colour3 bipura (1.0) colour4 salela (1.0) gomudo (0.4) colour5 buboga (1.0) colour6 dufami (1.0) Jump to top Interaction 900 Agent 8 is the speaker Agent 5 is the hearer Running in simulation mode Processed scene:
lab = (83.03, -137.43, 90.58)
lab = (86.83, -77.93, -17.11)
The speaker chooses SENSORY-OBJECT-3118 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-77
lab = (82.52, -136.83, 90.04)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-77) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-933 source-set-933 color-category-993) struct1 (get-context source-set-933) struct0:SOURCEdashSETdash9331->struct1:SOURCEdashSETdash9332 struct2 (bind color-category color-category-993 color-category-77) struct0:COLORdashCATEGORYdash9931->struct2:COLORdashCATEGORYdash9933 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "salela"
Formulation succeeded The speaker utters the word "salela" The utterance "salela" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "salela" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-19337 ?context-16059 ?x-16255) struct1 (get-context ?context-16059) struct0:varCONTEXTdash160591->struct1:varCONTEXTdash160592 struct2 (bind color-category ?x-16255 color-category-232) struct0:varXdash162551->struct2:varXdash162553
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-19337 ?context-16059 ?x-16255) struct1 (get-context ?context-16059) struct0:varCONTEXTdash160591->struct1:varCONTEXTdash160592 struct2 (bind color-category ?x-16255 color-category-232) struct0:varXdash162551->struct2:varXdash162553 The hearer tries to interpret the utterance using COLOR-CATEGORY-232
lab = (82.49, -136.74, 89.99)
Interpretation succeeded The hearer points to SENSORY-OBJECT-3118 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-77 The agent rewarded the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-232 The agent rewarded the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 900 interactions: g cluster_cluster8 agent 8 (interaction 900) colour1 sakalu (1.0) colour2 muzazu (1.0) colour3 bipura (1.0) colour4 salela (1.0) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 900 interactions: g cluster_cluster5 agent 5 (interaction 900) colour1 sakalu (1.0) colour2 muzazu (1.0) colour3 bipura (1.0) colour4 salela (1.0) colour5 buboga (1.0) colour6 dufami (1.0) Jump to top Interaction 1000 Agent 1 is the speaker Agent 6 is the hearer Running in simulation mode Processed scene:
lab = (93.86, -23.12, 102.27)
lab = (86.03, -76.11, -18.49)
The speaker chooses SENSORY-OBJECT-3471 as the topic The speaker tries to conceptualise the topic Conceptualisation succeeded The speaker conceptualises the topic using COLOR-CATEGORY-308
lab = (96.26, -22.57, 104.57)
Meaning representation: g struct0 (filter-by-closest-color ?topic ?source-set ?color-category) struct1 (get-context ?source-set) struct0:varSOURCEdashSET1->struct1:varSOURCEdashSET2 struct2 (bind color-category ?color-category color-category-308) struct0:varCOLORdashCATEGORY1->struct2:varCOLORdashCATEGORY3 The speaker tries to formulate Formulating g struct0 (filter-by-closest-color topic-1033 source-set-1033 color-category-1093) struct1 (get-context source-set-1033) struct0:SOURCEdashSETdash10331->struct1:SOURCEdashSETdash10332 struct2 (bind color-category color-category-1093 color-category-308) struct0:COLORdashCATEGORYdash10931->struct2:COLORdashCATEGORYdash10933 initial structure application process constructional dependencies applied constructions resulting structure
Utterance: "dufami"
Formulation succeeded The speaker utters the word "dufami" The utterance "dufami" is passed from speaker to hearer The hearer tries to parse the utterance Comprehending "dufami" initial structure application process constructional dependencies applied constructions resulting structure
Meaning: g struct0 (filter-by-closest-color ?topic-20853 ?context-17422 ?x-17618) struct1 (get-context ?context-17422) struct0:varCONTEXTdash174221->struct1:varCONTEXTdash174222 struct2 (bind color-category ?x-17618 color-category-14) struct0:varXdash176181->struct2:varXdash176183
Parsing succeeded The hearer could parse the utterance: Meaning representation: g struct0 (filter-by-closest-color ?topic-20853 ?context-17422 ?x-17618) struct1 (get-context ?context-17422) struct0:varCONTEXTdash174221->struct1:varCONTEXTdash174222 struct2 (bind color-category ?x-17618 color-category-14) struct0:varXdash176181->struct2:varXdash176183 The hearer tries to interpret the utterance using COLOR-CATEGORY-14
lab = (95.75, -22.88, 104.10)
Interpretation succeeded The hearer points to SENSORY-OBJECT-3471 as the topic This is correct. The speaker nods The speaker started alignment The agent shifted COLOR-CATEGORY-308 The agent rewarded the construction: The speaker finished alignment The hearer started alignment The agent shifted COLOR-CATEGORY-14 The agent rewarded the construction: The hearer finished alignment Interaction succeeded The speaker's lexicon after 1000 interactions: g cluster_cluster1 agent 1 (interaction 1000) colour1 sakalu (1.0) colour2 muzazu (1.0) colour3 bipura (1.0) colour4 salela (1.0) colour5 buboga (1.0) colour6 dufami (1.0) The hearer's lexicon after 1000 interactions: g cluster_cluster6 agent 6 (interaction 1000) colour1 sakalu (1.0) colour2 muzazu (1.0) colour3 bipura (1.0) colour4 salela (1.0) colour5 buboga (1.0) colour6 dufami (1.0)