Design of Human-GIS Dialogue for Communication of Vague Spatial Concepts. Based on Human Communication Framework. Hongmei Wang1,3. Alan M.
Design of Human-GIS Dialogue for Communication of Vague Spatial Concepts Based on Human Communication Framework Hongmei Wang1,3 Alan M. MacEachren2,3 Guoray Cai1,3 1 School of Information Sciences and Technology 2 Department of Geography 3 GeoVISTA Center Pennsylvania State University 1 311B IST Building, University Park, PA 16802, U. S. A. 2, 3 310 Walker Building, University Park, PA 16802, U. S. A. {hzw102, maceachren, gxc26}@ psu.edu Spoken language interfaces have the potential of achieving interface transparency for GIS [1, 2]. One of the challenges for designing such an interface is that natural language is vague and ambiguous when used in the communication of spatial concepts. This challenge has been recognized by the GIS community[1-4]; yet limited progress has been made[5-9]. The goal of this study is to facilitate human-GIS conversation that includes vague spatial concepts. Our approach integrates a human communication framework with agent-based computational methods to enable a collaborative dialogue. The human-GIS dialogue is designed to overcome vagueness by emulating a human-human collaborative dialogue. With this approach, each vague spatial concept is understood within a shared context (established through human-computer dialogues), followed by dialogue-assisted feedback and a negotiation process that grounds the meaning of the spatial concept in a common understanding. Our previous work[10-12] has focused on the design of a dialogue management agent, GeoDialogue, to enable mixed-initiative interactions during human-GIS dialogues. The agent, GeoDialogue, maintains the knowledge of the dynamic discourse context through an agent-based computational model PlanGraph and understands the user’s request within the current discourse context. Although our previous work introduced preliminary ideas about communicating vague spatial concepts through collaborative dialogues, methods used to share contextual knowledge and to ground the meaning of vague spatial concepts were simplistic, because the complex mechanisms through which vagueness is generated and managed during the communication process are not well understood. In the study reported here, we address the problem of communicating vague spatial concepts from a human cognition perspective. We propose a human communication framework to understand the effects of contexts on the communication of vague spatial concepts. This work has led to insights for resolving the vagueness problem during human-GIS communication. By integrating this framework with the agent-based computational method, this study models the human-GIS communication of a vague spatial concept as a collaborative planning process between the user and the system towards a shared understanding of the concept. We extend the system’s knowledge representation related to communication of vague spatial concepts by establishing a knowledge base which emulates the human knowledge involved in communication of
vague spatial concepts. The GIS can take different dialogue strategies encoded in the system knowledge base to handle specific issues involved in communication of vague spatial concepts. We also extend the agent-based computational model PlanGraph to manage the system’s finer mental belief statuses and associated reasoning algorithms specifically for the GIS to handle the vagueness problems during the communication process. The human communication framework built here draws upon theories related to human communication of meanings [13-17]. According to this framework, communication of a vague spatial concept between the system and the user subjects to three issues: 1) mental knowledge about a concept varies from person to person; 2) the spatial object or a spatial relationship referred by a spatial concept may be inherently fuzzy; 3) the contextual knowledge that are necessary for interpreting the meaning of a spatial concept may not be fully shared, causing discrepancies between what is signified and what is understood. To handle these issues, at first, human communicators usually use the existing and estimated contextual information to understand the vague concept communicated first. Next, they can use various collaborative dialogue strategies to handle these issues. Two major dialogue strategies are used here, including sharing contextual knowledge and negotiating meanings. These dialogue strategies are adopted in this study to facilitate the human-GIS conversation of vague spatial concepts. In this study, we extend the system’s knowledge representation on the vague spatial concept by simulating the human’s knowledge related to communication of vague concepts. We built the two human dialogue strategies as two computational ways in the system’s knowledge for the system to understand a vague spatial concept. One way encodes major contextual factors as optional parameters associated with this way for the system to understand that concept. With using this way, the system knows what contextual factors affect the meaning of that vague spatial concept, what contextual information should be extracted from the background context (represented as the root of PlanGraph), and what contextual information can be asked from the user or shared with the user. To enable the system to solve the first two issues discussed in the last paragraph, we build another computational way for the system to negotiate the meaning of the vague spatial concept by following the human dialogue strategy to negotiate the meaning of vague concepts. In case of communication of vague spatial concepts, the human communicators have different mental belief statuses and different reasoning processes and can take different actions or dialogue strategies from those involved in cases of communication of other types of concepts, e. g. a determined concept or a general concept. To enable the system to have such abilities necessary for communication of vague spatial concepts, we extend the computational representation of the system’s the mental belief statuses in the communication process and associated reasoning algorithms for the system to reason what to do next in terms of the discourse context representation PlanGraph. The reasoning algorithms specifically for the purpose of communication of vague spatial concepts are listed as follows: 1) In case the system believes that more contextual knowledge can be asked from the user to better understand a vague concept, it needs to prepare for asking for relevant contextual information from the user in the next step; 2) In
case that the system believes that contextual knowledge is enough, it needs to use the second computational way described above to further refine the concept meaning by negotiating it with the user in the next step; 3) In case the system believes that the user disagrees with the concept meaning represented by the system, the system needs to continue to negotiate it with the user. As the human-GIS dialogue proceeds, the background context information related to communication of a vague spatial concept is modeled as parameters associated with the root of the PlanGraph, which represents the final goal of the user’s request. The final goal can be a task that the user is going to do or an event that the user is going to participate in. These parameters associated with the root represent different properties of the final goal, which constraints the system and the user’s understanding of all vague concepts involved in the final goal. The system can re-use part of such contextual information in case that it needs to understand another vague concept in the same background context later. The work presented here can improve the system’s ability to successfully communicate vague spatial concepts with the user by enabling the system to reach a shared understanding of any vague concepts with the user. The human communication framework proposed here helps us to think about various problems in the human-GIS interaction from the human cognition perspective. This work also facilitates the GIS to have a more natural interaction with the user by simulating a human communicator’s behavior to communicate with the user. Our future work will focus on further implementation of the design about the humanGIS dialogue reported here and evaluation of a GIS incorporating the collaborative dialogue approach described here. The human communication framework proposed in this study can be used to understand the mechanisms about how other problems, e. g. ambiguity, are generated in the human-GIS communication and what different collaborative dialogue strategies can be used for the GIS to solve them. Acknowledgements This material is based upon work supported by the National Science Foundation under Grants No. BCS-0113030 and EIA-0306845. References 1. Frank, A. U. and Mark, D. M.: Language issues for GIS. In: Macguire, D., Goodchild, M. F. and Rhind, D.(eds): Geographical Information Systems: Principles and Applications, Wiley, New York (1991) 147-163 2. Mark, D. M., Svorou, S. and Zubin, D.: Spatial terms and spatial concepts: Geographic, cognitive, and linguistic perspectives. Proceedingsof International Geographic Information Systems (IGIS) Symposium: The Research Adgenda:, Arlington, VA. (1987) 101-112 3. Robinson, V. B.: Interactive Machine Acquisition of a Fuzzy Spatial Relation. Computers and Geosciences 6 (1990) 857-872
4. Montello, D. R., Goodchild, M. F., Gottsegen, J. and Fohl, P.: Where's Downtown?: Behavioral Methods for Determining Referents of Vague Spatial Queries. Spatial Cognition and Computation 2/3 (2003) 185 - 206 5. Wang, F.: A fuzzy grammar and possibility theory-based natural language user interface for spatial queries. Fuzzy Sets and Systems 1 (2000) 147-159 6. Wang, F.: Towards a natural language user interface: An approach of fuzzy query. International Journal of Geographical Information Systems 2 (1994) 143-162 7. Wang, F.: Handling Grammatical Errors, Ambiguity and Impreciseness in GIS Natural Language Queries. Transactions in GIS 1 (2003) 103-121 8. Kollias, V. J. and Voliotis, A.: Fuzzy reasoning in the development of geographical information systems FRSIS: a prototype soil information system with fuzzy retrieval capabilities. International Journal of Geographical Information Science 2 (1991) 209223 9. Duckham, M., Mason, K., Stell, J. and Worboys, M. F.: A formal approach to imperfection in geographic information. Computers, Environments and Urban Systems (2001) 89-103 10. Cai, G., Wang, H. and MacEachren, A.: Communicating Vague Spatial Concepts in Human-GIS Interactions: A Collaborative Dialogue Approach. Conference on Spatial Information Theory 2003, Kartause Ittingen, Switzerland (2003) 304-319 11. Cai, G., Wang, H., MacEachren, A. M. and Fuhrmann, S.: Natural Conversational Interfaces to Geospatial Databases. Transactions in GIS (Accepted) 12. Wang, H. and Bolelli, L.: Agent-based collaboration approach for GIS to negotiate vague spatial concepts with the user. Eighteenth Annual Graduate Exhibition of Penn State University, University Park, PA (2003) 13. Barsalou, L. W.: Perceptual symbol systems. Behavioral and Brain Sciences (1999) 577-609 14. Barsalou, L. W.: Context-independent and context-dependent information in concepts. Memory & Cognition 1 (1982) 82-93 15. Heath, R. L. and Bryant, J.: Human communication theory and research: concepts, contexts, and challenges. 2. Lawrenece Erlbaum Associates, Publisher, Mahwah, New Jersey (2000) 16. Jackendoff, R. S.: Parts and boundaries. Cognition (1991) 9-44 17. Dimitrov, V. and Russell, D.: The Fuzziness of Communication. In: Fell, L., Russell, D. and Stewart, A.(eds): Seized by Agreement, Swamped by Understanding: Acollection of papers to celebrate the visit to Australia, in August 1994, Hawkesbury Printing, University of Western Sydney, Sydney, Australia (1994) 183-192