Landmarks with Many "de" Cases 

Contexts in source publication

Context 1

... to the result obtained, wards in Kyoto have weak bond as communities. This is one characteristic of geographical object ”ward”, and is expressed on its usage in spatial descriptions. (Although ward is a district and not really a landmark.) The result shows that frequencies of case forms are closely related to the characteristics of landmarks. Our system extracts landmark information from spatial description and presents it in map style. The following subsections describe our system in detail. As a source of spatial descriptions, we have collected 2 million web pages using web crawling. The web pages are mainly those on Kyoto, the previous capital of Japan, which is the target area of our research. The web pages include online newspaper articles from Kyoto Shinbun, a newspaper in Kyoto, and websites under Kyoto-inet domain, a major ISP covering Kyoto area. Although a sentence or a group of sentences as a whole gives full information of spatial description, we limited our focus to cases of place names. Using morphological analyzer Chasen [9], sentences from 2,171,496 (about 2 million) web pages are separated into parts of speech. The list of place names used as candidates of landmarks was obtained from Zenrin Co., Ltd.’s map data. Zenrin’s map on Kyoto contains 7144 distinct ”markers”. Markers are significant geographical objects such as stations, temples, and universities. We considered markers as potential landmarks. 2 million web pages were scanned and numbers of case particles coming after markers were counted. Frequencies of different case particles after each marker were counted. Seven case particles Table 1 and a dependency marker ”ha” are used in our system. There are still other case particles and dependency markers in Japanese. However, we limited them to avoid system to be too complicated. In order to visualize landmark characteristics on a map, we used eight different icons. As mentioned in previous sections, icons are good way to include landmark characteristics into a map. Each of eight different cases was repre- sented by an icon with certain intuitive meanings attached to it. Data from Zenrin Co., Ltd. were used. We used resi- dence map and also area maps of scales of 10 thousandth, 250 thousandth, and 200 thousandth, on Kyoto City. Figure 4 and Figure 5 shows interface that presents landmarks of different characteristics with icons. Figure 4 shows landmarks with more than 20 appearances of ”de” case. Landmarks are indicated by human figure icons, since ”de” case is most likely be used as a location of human action. Famous shrines and universities are shown on the map, indicating some events are being held. Figure 5 shows landmarks with more than 5 appearances of ”kara” case. Footstep icons pointing downward, since ”kara” case indicates origin of motion indicate landmarks. Stations are landmarks that appear frequently in this figure. This can be explained that their characteristics as a starting point of motion is expressed as a high frequency of “kara” case, which has case meaning of “origin”. Graphical interface shows that frequency of cases corresponds with characteristics of landmarks to some extent. Characteristics of spatial descriptions in Japanese have been widely studied by Japanese linguists. Okutsu discussed relative nouns in detail [15]. Nakau described difference between case particles ”ni” and ”de” in details [13]. Comparison between case particles ”e” and ”ni” were discussed by Kunihiro [5]. National Research Center for Japanese Language (Kokuritsu Kokugo Kenkyuusho) has surveyed correspondence between case forms and case meanings. There are still other works discussing how languages are used in describing space [8][16]. The idea that maps should not be a simple picture of physical world but rather match cognitive understandings of geographical space comes from studies of cognitive ge- ography, or ”naive geography”. The basic principals were presented by Egenhofer et al. [2][10]. Concept of men- tal maps, which is a representation of human knowledge of geographical space, is described thoroughly by Nakamura [11]. Classic work by Lynch also suggests importance of landmarks in geographical cognition [6]. We have discussed how landmarks act as key concepts in presenting geographical information, structure of spatial descriptions in Japanese, and showed how case forms are related to characteristics of landmarks. Observations have revealed that frequencies of case forms are related to characteristics of landmarks. Future work will extend analysis onto other parts of speech in spatial descriptions, so that more information can be obtained. Relative nouns and spatial adjectives are the next target. Then cooccurences of landmarks will be ob- served, to extract network of landmarks. Validity of maps drawn must be checked. Experiments must be performed to confirm that those maps are useful in teaching directions or giving general overview of the region. It should be proven that landmark icons have really improved usefulness of map. This work has been supported in part by CREST of JST (Japan Science and Technology ...

Context 2

... to the result obtained, wards in Kyoto have weak bond as communities. This is one characteristic of geographical object ”ward”, and is expressed on its usage in spatial descriptions. (Although ward is a district and not really a landmark.) The result shows that frequencies of case forms are closely related to the characteristics of landmarks. Our system extracts landmark information from spatial description and presents it in map style. The following subsections describe our system in detail. As a source of spatial descriptions, we have collected 2 million web pages using web crawling. The web pages are mainly those on Kyoto, the previous capital of Japan, which is the target area of our research. The web pages include online newspaper articles from Kyoto Shinbun, a newspaper in Kyoto, and websites under Kyoto-inet domain, a major ISP covering Kyoto area. Although a sentence or a group of sentences as a whole gives full information of spatial description, we limited our focus to cases of place names. Using morphological analyzer Chasen [9], sentences from 2,171,496 (about 2 million) web pages are separated into parts of speech. The list of place names used as candidates of landmarks was obtained from Zenrin Co., Ltd.’s map data. Zenrin’s map on Kyoto contains 7144 distinct ”markers”. Markers are significant geographical objects such as stations, temples, and universities. We considered markers as potential landmarks. 2 million web pages were scanned and numbers of case particles coming after markers were counted. Frequencies of different case particles after each marker were counted. Seven case particles Table 1 and a dependency marker ”ha” are used in our system. There are still other case particles and dependency markers in Japanese. However, we limited them to avoid system to be too complicated. In order to visualize landmark characteristics on a map, we used eight different icons. As mentioned in previous sections, icons are good way to include landmark characteristics into a map. Each of eight different cases was repre- sented by an icon with certain intuitive meanings attached to it. Data from Zenrin Co., Ltd. were used. We used resi- dence map and also area maps of scales of 10 thousandth, 250 thousandth, and 200 thousandth, on Kyoto City. Figure 4 and Figure 5 shows interface that presents landmarks of different characteristics with icons. Figure 4 shows landmarks with more than 20 appearances of ”de” case. Landmarks are indicated by human figure icons, since ”de” case is most likely be used as a location of human action. Famous shrines and universities are shown on the map, indicating some events are being held. Figure 5 shows landmarks with more than 5 appearances of ”kara” case. Footstep icons pointing downward, since ”kara” case indicates origin of motion indicate landmarks. Stations are landmarks that appear frequently in this figure. This can be explained that their characteristics as a starting point of motion is expressed as a high frequency of “kara” case, which has case meaning of “origin”. Graphical interface shows that frequency of cases corresponds with characteristics of landmarks to some extent. Characteristics of spatial descriptions in Japanese have been widely studied by Japanese linguists. Okutsu discussed relative nouns in detail [15]. Nakau described difference between case particles ”ni” and ”de” in details [13]. Comparison between case particles ”e” and ”ni” were discussed by Kunihiro [5]. National Research Center for Japanese Language (Kokuritsu Kokugo Kenkyuusho) has surveyed correspondence between case forms and case meanings. There are still other works discussing how languages are used in describing space [8][16]. The idea that maps should not be a simple picture of physical world but rather match cognitive understandings of geographical space comes from studies of cognitive ge- ography, or ”naive geography”. The basic principals were presented by Egenhofer et al. [2][10]. Concept of men- tal maps, which is a representation of human knowledge of geographical space, is described thoroughly by Nakamura [11]. Classic work by Lynch also suggests importance of landmarks in geographical cognition [6]. We have discussed how landmarks act as key concepts in presenting geographical information, structure of spatial descriptions in Japanese, and showed how case forms are related to characteristics of landmarks. Observations have revealed that frequencies of case forms are related to characteristics of landmarks. Future work will extend analysis onto other parts of speech in spatial descriptions, so that more information can be obtained. Relative nouns and spatial adjectives are the next target. Then cooccurences of landmarks will be ob- served, to extract network of landmarks. Validity of maps drawn must be checked. Experiments must be performed to confirm that those maps are useful in teaching directions or giving general overview of the region. It should be proven that landmark icons have really improved usefulness of map. This work has been supported in part by CREST of JST (Japan Science and Technology ...