153 11 Estimation of Routes and Building Sites Described in Premodern Travel Accounts Through Spatial Reasoning Yasushi Asami, Takanori Kimura, Masashi Haneda, and Naoko Fukami CONTENTS 11.1 Introduction 153 11.2 Spatial Data 155 11.3 Search for the Route 156 11.3.1 Consistency with Building Use 156 11.3.2 Consistency of Branching Off of Road 156 11.3.3 Consistency of Turning Direction (Left or Right) 157 11.3.4 Consistency of Direction (North, East, South, and West) 157 11.4 Judgment of Neighborhood 157 11.5 Judgment of Priority in Route Choice 158 11.6 Route of Maximum Degree of Fitness and Estimation of Locations of Building 158 11.7 Conclusion 160 Acknowledgment 161 References 161 11. 1 Introduction Historical materials, such as documents and excavated evidences, are major sources of historical studies. These sources often show some spatial relation- ships at the era studied. To efficiently handle such information, geographical information systems (GIS) can be applied. Sometimes, the location of historical facilities is of particular interest when they are critical to revealing historical evidences. If excavated evidence is 2713_C011.fm Page 153 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC 154 GIS-based Studies in the Humanities and Social Sciences already known, there will be no need to search for the location. There are, however, a number of facilities whose locations are not known. In such cases, historical documents, if any, are the only source of inferring the location. For this purpose, spatial reasoning can be of potential use. Spatial reasoning is an attempt to infer the spatial location and relationship in spatial-information science. It involves the coding of information concern- ing spatial locations and inference concerning spatial relationship (Hernan- dez, 1994; Vieu, 1997). By coding the spatial information, specific spatial relationships can be focused on and topologically or quantitatively analyzed. This chapter illustrates the applicability of spatial reasoning in historical analysis by showing its application to infer the spatial locations of several facilities written in Jean Chardin’s travel account and his walking route in Isfahan in Iran in the 17th century, based on Asami, Kimura, Haneda and Fukami (2002). Isfahan, an old town located almost at the center of Iran, as shown in Figure 11.1, flourished for about 130 years from 1597, when Abbas I set the town as a capital, until early in the 18th century, when the Safavid dynasty fell. Jean Chardin, a French traveler, visited Isfahan late in 17th century and wrote a travel account based on his experience in the town. His account has been a first class and precious historical source full of detailed and vivid descriptions of the situation in the town (Haneda, 1996). Despite his detailed description, historians have not decisively located a number of buildings. A notable example is the location of la Maison de la Douze Tomans (Twelve Tomans’ house), where Chardin lived during his stay in Isfahan. Its location is of particular importance, for it was always the origin of his record. If its precise location can be identified, then locations of other historically important buildings can be inferred with much precision. Understanding the city structure of Isfahan — including blocks, streets, and locations of religious and urban facilities — in the premodern age is of historical importance, for it leads to the understanding of city planning and urban social structure (Sakamoto, 1980–1981). FIGURE 11.1 Isfahan in Iran. 2713_C011.fm Page 154 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC Estimation of Routes and Building Sites 155 11.2 Spatial Data To start the spatial-reasoning analysis, it is quite important to prepare a base map ready for GIS. It is well-known that the city structure of Isfahan was drastically changed in 1930s as a result of urban-modernization policy in Iran. The city had had no experience of change in large scale before this period. A most reliable map that shows the city structure before the 1930s is one made by Sayyed Reza Khan in 1923 to 1924. An old map, as is often the case, is not very precise in locations of road networks, as well as those of buildings. To make a more accurate map, several facilities in an aerial photograph map, which cannot be moved, are used. The transformed map is digitized to convert to GIS-ready spatial data. With this operation, 337 nodes are recorded. All the roads are expressed by links between nodes, and building use along the roads are input as attribute data of links (Figure 11.2). The next step for the spatial reasoning is the extraction of descriptions concerning spatial relationships in Isfahan. The travel account used for our analysis is that by Chardin, a French jeweler and traveler (Haneda, 1996). A part of Chardin’s travel account describing the route from Mirza Ashraf’s residence to “ l’Évêché (Bishop’s mansion)” is selected for our analysis, for the starting point and the destination of the route, as well as most of build- ings appearing in this description cannot be identified by the analysis of the text, and hence the route is not yet identified. Moreover, Chardin seems to have gone forward and then returned to a certain point on the route and walked in another direction. The spatial identification of this route is of great concern in historical studies. FIGURE 11.2 Digitization of old map (dots in the road network indicates nodes). 2713_C011.fm Page 155 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC 156 GIS-based Studies in the Humanities and Social Sciences There are 51 phrases that describe spatial relationship in this part. The spatial descriptions include relationships such as: (1) specific buildings are neighboring to each other; (2) buildings of particular building use are neigh- boring to each other; (3) roads branch off into a certain number of roads; (4) turning to left (or right); and (5) going to east (or other direction). These descriptions are related to characteristics of landmarks, crosses, and direc- tions (Haneda, 1996). The building use along the road indicated in the old map is assigned to the road as an attribute. The building name is not shown in the old map. For the details concerning the phrase and judgment of degree of coincidence, refer to Kimura (2002). 11.3 Search for the Route To derive a route that best fits to the description, two things have to be identified. One is the enumeration of the complete set of potential routes, among which the best route is identified. For this end, a computer program was coded to search all possible routes. It calculates all the possible routes conditional to a set of origins and destinations. The other important device is the evaluation measure of routes and how they are consistent with the description in the travel accounts. To measure the degree of coincidence of the route to the set of descriptions in the phrases, consistency function, C( m , n ), is defined, where m is the identification number of the phrase describing spatial relationship, and n is the order of the link specified. This function takes the value of 0, when the n th link is not consistent with phrase m , while it takes positive value if the link is somehow consistent with the phrase depending on the extent of coincidence. The sum of consis- tency function over the entire route is defined as degree of fitness with the description of the travel account. Note that even if a route is fixed, there are several ways to match the route to the set of phrases in general. The maximum value is assigned for the degree of fitness among all possible ways of matching. More specifically, consistency function is defined as follows: 11.3.1 Consistency with Building Use Since the existence of buildings at the time when Chardin traveled is not known now, consistency is judged based merely on the coincidence of build- ing use along the link. If a building at the age of the Safavids is confirmed, then one more point is assigned than the case that only the building use coincides to the building drawn in Sayyed Reza Khan’s map from the 1920s. 11.3.2 Consistency of Branching Off of Road If the road is branching off into the same or larger number of roads in Sayyed Reza Khan’s map, then the link is judged consistent with the phrase. This is 2713_C011.fm Page 156 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC Estimation of Routes and Building Sites 157 because the road network in the 1920s may develop more than that in the age of the Safavids. 11.3.3 Consistency of Turning Direction (Left or Right) If the link correctly turns to the left or right in accordance with the phrase description, then the link is judged consistent. 11.3.4 Consistency of Direction (North, East, South, and West) If the corresponding link is closest to the correct direction, then the link is judged consistent. Consistency function is weighted depending on the reliability by three values, 5, 3, and 1. For example, if the consistency of a building is judged by the coincidence of building use in the 1920s, the reliability of this judg- ment is rather low, and hence 1 is given for the weight. On the other hand, a road branching off and a description that the road leads to a certain facility are both highly reliable, and 5 is given for the weight. The value of 3 is given as a weight for the case of intermediate reliability (Kimura, 2002). 11.4 Judgment of Neighborhood Distance in historical documents is described in various forms, from distance measures to ambiguous expressions, such as “near” and “far.” A concrete distance measure in Chardin’s travel account is the usage of steps, how many steps it is from a certain place to another place. For example, the size of the Royal Square ( Meydan-e Shah ) is described as 440 steps in length and 160 steps in width. A caution should be paid, however, to these numbers. It could be an actual measurement by the author, but it could be also a measurement by impression or rumor. Thus, these numbers have to be examined against the actual size of the plaza. A vague distance expression appearing in Chardin’s travel account is the use of “neighborhood.” A place is described in the neighborhood of another place, if Chardin thought it was close enough. This kind of impression can change from situation to situation, but the threshold of usage for maximum distance in the “neighborhood” can be rather stable in a limited context, such as the parts of description in the current analysis. In the account, la Maison de la Douze Tomans is described as follows: “It was located in the neighborhood of the Royal Palace and the Royal Square, and close to British and Dutch East India Company’s offices as well as Capucin and Carme mission’s houses, and no other place is more conve- nient.” The terms, “neighborhood” and “close” ( à côté de in French for both), 2713_C011.fm Page 157 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC 158 GIS-based Studies in the Humanities and Social Sciences of course indicate that these facilities are near in distance. To estimate up to what distance Chardin used the term, similar expressions were searched for in the travel account. One description is that “Solimancan’s ( Soleyman Khan’s ) residence is located 150 steps from the two houses la Maison de la Douze Tomans and its next house), and Dutch East India Company’s office is next to it.” According to this state- ment, the distance between la Maison de la Douze Tomans and Dutch East India Company’s office is about 150 steps, but this distance cannot be judged the maximum distance for the range of neighborhood. Since no effective informa- tion is available in the description of the travel account, a naïve assumption was made that the maximum distance will be twice as much, i.e., 200 steps, to term “within the neighborhood” in the following. Chardin reported that the size of the Royal Square was 440 steps in length and 160 steps in width. The Royal Square still exists as it was then, and hence, by measuring the dimension of the plaza, the distance of one step can be estimated. By this procedure, one step is derived as 1.16 meters. By multiplying 300 steps, 347.7 meters can be regarded as the range of neigh- borhood. Using this distance, the possible area for la Maison de la Douze Tomans will be estimated. 11.5 Judgment of Priority in Route Choice It is natural to assign high priority in the route choice for roads that might have existed at the time of Chardin, such as roads connecting city gates, and the gate and the Royal Palace or the citadel, for more reasonable estimation of the route. To this effect, all the roads are ranked into four levels depending on the possibility of existence in Chardin’s age, based on Gaube’s estimation (Gaube and Wirth, 1978) and Chardin’s description. Moreover, the longer the route on the roads in that age, the more probable, it is judged, that Chardin actually took the route. Since the degree of fitness is defined as a linear form, as stated in Section 11.3, a longer route tends to gain a higher score than a shorter route. To adjust this tendency, the score for the route that looks longer than expected from the description of the travel account will be diminished. 11.6 Route of Maximum Degree of Fitness and Estimation of Locations of Building The route that scored the maximum degree of fitness is depicted in Figure 11.3. In the figure, the route consists of two arrows, for Chardin went back 2713_C011.fm Page 158 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC Estimation of Routes and Building Sites 159 to a certain point on the route. A route estimated by historians (i.e., Haneda and Fukami, coauthors of this chapter, based on descriptions only without using spatial-reasoning method) is shown in Figure 11.4. A discrepancy is found only in a part to the Royal Square. This is due to the lack of information of buildings in this area, and hence no score can be gained for the historians’ estimate route in this area. An examination of historian data in this area will make the estimate of the route more accurate, and the excavation in this area would be of great importance for further historical research. Location of la Maison de la Douze Tomans was estimated based on the phrase that “it was located in the neighborhood of the Royal Palace and the Royal Square, and close to British and Dutch East India Company’s offices as well as Capucin and Carme mission’s houses, and no other place is more conve- nient.” Using the information in this same phrase, the location of the Capucin mission’s house can be estimated, which in turn can be used to estimate the location of the Bazaar with the name of Mahamed Emin (Mohammad Emin), which is recorded to be 30 steps from the Capucin mission’s house. A buff- ering operation for la Maison de la Douze Tomans is devised for this process to estimate the location of the Capucin mission’s house with the use of the distance for “neighborhood.” The estimated locations of la Maison de la Douze Tomans are shown in Figures 11.3 and 11.4. Since this house is along the route, where spatial-reasoning results and historians’ estimations differ, the loca- tion is differently estimated. Figure 11.3 also indicates the location of the Capucin mission’s house, which was difficult to estimate by historians with- out help of spatial reasoning. FIGURE 11.3 Estimated location of “ la Maison de la Douze Tomans ” Estimated location of the Capucin mission’s house 2713_C011.fm Page 159 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC 160 GIS-based Studies in the Humanities and Social Sciences 11.7 Conclusion Spatial relationship is seldom a main focus of historical studies, partly due to the lack of analytical tools suited for such studies. Development of GIS, as well as that of methods for spatial analyses, has made it possible to cultivate spatial matters in history. This chapter illustrated an application of spatial reasoning using the GIS technique to estimate the route in Chardin’s description of his travels in the 17th century. The resulting route fairly matches historians’ estimated route, except for the area close to the Royal Square, where the lack of information for buildings is severe. Definition of degree of fitness and consistency function would depend on the quality of investigation. Judgment of importance and ranking of roads, for example, can be improved by using information from other various historical sources of the period. In a sense, the spatial-reasoning program can be thought of as a learning process from historical documents. Devel- opment of such learning-type evaluation functions can extend this kind of research. FIGURE 11.4 Estimated location of “ la Maison de la Douze Tomans ” 2713_C011.fm Page 160 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC Estimation of Routes and Building Sites 161 Acknowledgment Authors of this chapter were benefited from comments by Atsuyuki Okabe, Yukio Sadahiro, and Atsushi Masuyama. This research was supported by Grant-in-aid for Scientific Research by the Ministry of Education, Culture, Science, Sports and Technology of the Japanese government. These are grate- fully acknowledged. References Asami, Y., Kimura, T., Haneda, M., and Fukami, N., Estimation of Route and Building Sites Described in Pre-modern Travel Accounts Through Spatial Reasoning, papers and proceedings of the Geographic Information Systems Association, 11, 369–372, 2002. Gaube, H. and Wirth, E., Der Bazar in Isfahan ., Wiesbaden, 1978. Haneda, M., A Study of Jean Chardin’s Description of Isfahan , University of Tokyo Press, Tokyo, 1996. Hernandez, D., Qualitative Representation of Spatial Knowledge , Springer-Verlag, Berlin, 1996. Kimura, T., Restoration of Facility Allocation in an Old Map Through Spatial Rea- soning: Case Study of Isfahan in the 17th Century, graduation thesis, Depart- ment of Urban Engineering, University of Tokyo, 2002. Sakamoto, T., Urban Structure and Meydan in the Nineteenth Century Isfahan , (I), Shigaku (Historical Studies in Keio University), 50, 367_387,1980; (II), Shigaku, 51(1_2), 145_158, 1981; (III), Shigaku, 51(3), 43_79, 1981. Vieu, L., Spatial representation and reasoning in artificial intelligence, in Spatial and Temporal Reasoning , Stock, O., Ed., Kluwer Academic Publishers, Dordrecht, 1997, pp. 5–41. 2713_C011.fm Page 161 Thursday, September 8, 2005 5:29 AM Copyright © 2006 Taylor & Francis Group, LLC . building is judged by the coincidence of building use in the 1920s, the reliability of this judg- ment is rather low, and hence 1 is given for the weight. On the other hand, a road branching off and. judged based merely on the coincidence of build- ing use along the link. If a building at the age of the Safavids is confirmed, then one more point is assigned than the case that only the building. analysis, for the starting point and the destination of the route, as well as most of build- ings appearing in this description cannot be identified by the analysis of the text, and hence the route