Guidelines for the Final Format of a Paper Submitted to SCANGIS 97

Verbal and Graphical Representation of Route Descriptions1

Verbal and Graphical Representation of Route Descriptions

Erik Stubkjær and Inger Lytje à

Dept. of Development and Planning, Aalborg University,

Fibigerstræde 11, DK-9220 Aalborg Øst

,

à Department of Communication, Aalborg University,

Langagervej 8, DK-9220 Aalborg Øst

,

Abstract: A natural language route description conveys information which can be provided alternatively by drawing the route on a topographic map. Relations between the verbal and the graphical modality of the route description are investigated on the basis of cognitive grammar.

A text corpus is analyzed as regards its phrases: Noun, Verbal, and Prepositional Phrases. The system Gratex which is developed according to Langacker's presentation of cognitive grammar was used for the analysis. The analysis revealed that most spatial information was carried by the Noun and Prepositional Phrases, while the Verbal Phrases carried little information. Structures which addresses spatial objects and relations are identified and discussed with a view to establish an operational relationship between the verbal and the graphical modality of route descriptions..

KEYWORDS: Cognitive grammar, route description, Danish text corpus, spatial objects.

1. Introduction

Multimedia products and equipment is popular among users, and vendors as well. The technology of integrating text and single images, music and speech, animations and video sequences is fascinating. The present paper addresses a fraction of this complex, but presents theory which may have relevance for the development of multimedia products.

The paper is part of an ongoing project, LinGIS. The project aims at eliciting conceptual structures within the domain of the land surveyor and geographical information systems. A text corpus which has been prepared by land surveying students makes the empirical basis of the study. The text corpus describes routes to the students' practice area, and topographical details, in Danish natural language. The route description conveys information which can be provided alternatively by drawing the route on a topographic map. This suggests an analysis of the relations between verbal and graphical descriptions of the same reality. The issue was raised if the outcome of a parsing of the verbal route description could be structured in such a way that a computer program could draw the route on a digital topographic map. Reference is made to a previous presentation which included a survey of research on route descriptions (Stubkjær, 1995), and to (Lytje & Stubkjær, 1997).

The following section outlines the theoretical basis of the present analysis. A part of the text corpus is analyzed as regards its phrases: Noun, Verbal, and Prepositional Phrases, and the outcome is reported in the three subsequent sections. The three sections include definitions of key terms. A discussion of the analysis summarizes the outcome.

2. Theoretical basis

The theoretical basis for the investigation is cognitive semantics, and cognitive grammar. The proponents of this research assumes a close relationship between syntax and semantics, and, furthermore, develop a cognitive semantics as opposed to a logical semantics. According to cognitive grammar linguistic expressions are interpreted as a coding of the meaning which is the purpose of the expression. Meaning is conceived as a conceptual structure, that is a psychological entity. Jackendoff calls these conceptual structures conceptual constituents (Jackendoff, 1983; 1990), while Langacker speaks of image schemes (Langacker, 1987; 1991). (The Danish translation of 'image schemes' is 'forestillingsskemaer', literally in English: 'imagination schemes'). A more or less formalized correspondence is claimed to exist between the linguistic expression and cognitive semantic structures. These conceptual structures are the means by which we organize our knowledge at a pre-linguistic, or, more definite, a non-linguistic level. This level is common to the visual, auditive (phonetic, musical) and motoric systems of man. For example, Ray Jackendoff asserts (Jackendoff, 1990: 23):

Each conceptual category supports the encoding of units not only on the basis of linguistic input, but also on the basis of the visual (or other sensory) environment.

This position constitutes a central, theoretical basis for the development of multi-media products.

3. Analysis of the Text Corpus

In order to identify the conceptual structures which mediate between the route description in verbal and graphical form we have analyzed the text corpus by means of the Gratex system (Lytje & Donner, 1996). The Gratex system can show the user the occurrence of word classes, phrases, and segments of the text. According to Langacker these grammatical categories correspond to semantic categories which in turn correspond to the conceptual structures mentioned above which makes the basis for drawing routes on a map.

3.1 Referents of the Noun Phrases

A general linguistic assumption is that noun phrases refer to members of the category thing. Noun phrases express the existence of things, and are thus a container of knowledge of the things of the universe of discourse. Furthermore, noun phrases express the relation of things relative to other things. In the present context thing may be terrain objects or places.

In this context, a place is related to the surface of the Earth, and has a certain extension. The extension is determined by cultural rather than legal or geodetic rules. Named town squares, for example 'Capitol', where social events take place are the prototype of place.

A route relates several places in sequential order. The term route belongs to natural language. As opposed to a track, a route is not visible in the terrain, it is an abstract entity which is defined by the route description. In mathematical graph theory the natural language term route is represented by a path. A path is made up of nodes connected by edges. The nodes correspond to the natural language's places. The mentioned mathematical formalism is part of the theoretical basis for cartographic databases.

A terrain object is the general term for objects which are immovable relative to the surface of the Earth, whether they are natural (stream, cliff, group of trees), or man-made (construction, ditch, road). In (Stubkjær, 1995) terrain objects were categorized as follows:

• forest, garden,

• hill, brook, watershed

• house, road

• place names,
• district names,
• (signed, postal) addresses,
• cadastral designations

The Gratex system makes no distinction between terrain object and the other, more abstract entities, as all the mentioned nouns are classified as physical objects. An earlier version of the Gratex system did provide such distinction, but experience showed that the lexicon entries of Gratex could not catch the differences. This is due to the fact that the question of whether a noun assumes the rôle of place, or terrain object is context dependent. To give an example: A cup in most contexts is a physical object, but it can likewise serve as the place where a fly is landing and moving around.

Analysis of the Noun Phrases

An analysis of the noun phrases of the corpus shows that they belong to one of two groups: A main group which comprises terrain objects and places, and a residual group. The phrases of the residual group relates terrain objects by means of the traversed route, as it is developed below.

The larger group of nominal phrases can be further subdivided into two groupings: The larger of these refer to terrain objects or places, for example

'byen', 'Frejlev', 'brugsforeningen', 'stort parcelhusområde', 'grusvej' , 'landbrugsejendom'
('the town', the place name 'Frejlev', 'the grocer', 'a large residential area', 'an unpaved road', 'a farm').

The smaller group of nominal phrases locates places relative to other places through a whole - part relation, that is without referring to the route. Interesting is that this whole - part relation is frequently marked by a genitive construction:

'områdets sydøstlige grænse' , områdets nordvestlige hjørne'
('the South-East boundary of the area', 'the North-West corner of the area').

The directions mentioned are the cardinal directions, but relative directions are used as well. An example is

'på venstre hånd' ('at the left hand/side').

Following Ruth Kaas one can interpret the phrase as synonymous with 'at the left of AGENT' (Kaas, 1989: 74). A deeper interpretation would be 'at the left side of the road'. This interpretation introduces the whole - part relation mentioned above: A road has sides, and a cul-de-sac has an end. Furthermore, if the cardinal directions are applied, the relative location is achieved without reference to the route ('along the southern side of the road'). If relative directions ('left', 'right' ) are used the parser program has to resolve the problem by keeping track of the heading direction relative to the cardinal directions.

The residual group is made up of phrases which connects places and/or terrain objects by referring to movement, for example

'forskellige tilkørselsveje', 'adgang', 'adgangsvej', 'en anden adgangsvej'
('different access roads', access', 'access road', 'another access road'.

A common trait of these expressions is that they refer to the (imaginary or real) movement of the reader through the terrain. Implicitly, they direct the route to follow. This interpretation is in line with Ruth Kaas who suggest that
'to give access to' in the lexicon is made synonymous with 'to lead to' (page 46)
'to have access to' in the lexicon is made synonymous with 'to arrive at' ( : 48), and that an 'access road' is interpreted as a 'unique, not context free conceptual constituent' (: 103).

Concluding, the analysis shows that generally nominal phrases articulates specific conceptual structures by asserting the existence of places and terrain objects. A minor part of the nominal phrases contribute by relating places to one another - a function which according to cognitive grammar (Langacker, 1986) is generally performed by verbal phrases and prepositional phrases, or both together.

3.2 Analysis of Verbal Phrases

The verbal phrases of a sentence typically express a dynamic relation between subject and object. The analysis reveals that the verb phrases present two kind of knowledge. One refer - as expected - to the dynamics of traversing the route. It will be demonstrated that this theme is expressed through a variety of perceptions of the traversing situation. The other kind of verb phrases refers a static world, namely the terrain objects and their spatial relations.

Traversing the Route

When a body moves through an environment, the movement can be described from the point of view of the stable environment, or relative to the center and direction of the moving body (Stubkjær, 1992, with further reference i.a. to 'Turtle geometry'). In the text corpus both perceptions appear. They are expressed by verbs of movement and verbs of action.

Verbs of movement are applied in expressions as

'kommer man', fortsætter man', 'svinger man'
('you arrive at', 'you continue', 'you turn').

These phrases describe the moving body which turns, continues, and arrives. Other expressions like

'ved at fortsætte', 'skal man svinge'
('by continuing', 'you have to turn')

use verbs of movement as well, but the construction of infinitive and modality introduces a condition:: if you really want to go there, or expresses politeness towards the reader.

An interesting expression is

'kommer der en vej' ('a road comes/appears').

Literally, the terrain object 'road' is moving. This is of course not the case, rather the perception from the point of view of the body is exaggerated: The percepting body, ignoring that it is moving, notes that the terrain object is moving into its perception field (range of vision).

Verbs of action are referring to the stable environment in a surprising way, as the agent of action is a terrain object. An example:

'grusvejen giver adgang til et gartneri'
('the unpaved road provides access to a market garden').

The present tense further emphasizes the active aspect. One interpretation is that the author in a metaphorical way makes the terrain objects into actors of a scene. Another interpretation is that the unpaved road and the market garden are next steps on the route to the destination. The latter interpretation filters what is relevant for retaining the meaning of the text and drawing the route. The meaning is to introduce a new terrain object, namely the market garden, and to relate it to a previously asserted object, namely the unpaved road.

Another example is:

'en grusvej med tilkørsel til Nibevej forsyner gartneriet og fortsætter i nord-syd retning ned gennem området og opdeler det i to halvdele'
('an unpaved road which has access to [the road named] Nibevej supplies the market garden and continues in North-South direction through the area and divides it into two halves')

Again, the two action verbs 'supplies' and 'divides' refer to terrain object. The unpaved road appears as an agent who organizes the spatial environment.

As a passing remark, mention is made of the fact that the author who introduced the metaphor of the acting terrain objects was a land surveying student. Land surveying is an inquiry into the position of terrain objects. The surveying process presupposes that the terrain objects are identified, that is mentally isolated from their context, like paper models of tables, chairs, and other furniture normally used for making lay-out plans of a home or an office. Furthermore, the land surveyor sets out buildings and other constructions according to a lay-out plan. A land surveyor thus does not regard the terrain objects as simply being integrated with the surface of the Earth. Rather, the terrain objects have a kind of individual existence.

Describing the Environment of the Route

Verbs generally describe action, but in the present corpus verbs frequently describe the existence or certain attributes of terrain objects.

The verb 'to be' often appears often in the text corpus. This is made visible by the Gratex system when querying for auxiliary verbs. Auxiliary verbs are used to make verbal phrases in the tense of the present or the past participle. Examples of this construction are:

'er beliggende', 'er fordelt', 'er der plantet'
('is lying/situated', 'is distributed', 'is planted')

Furthermore, the verb 'to be' is used to characterize terrain objects by describing their attributes:

'..er træerne lidt yngre og kun 4 - 5 m høje'
('are the trees slightly younger and only 4 - 5 meters tall')
'.. at den ikke er farbar' ( 'that it [the road] is not passable')
'.. er meget kuperet' ('is heavily undulating').

Finally, the verb is used in the plain form which refers to the existence of a thing. In the present corpus it means that a terrain object is asserted somewhere in the spatial environment of our concern. Examples are

'I Frejlev er der tre tilkørselsveje til området'
('From [the town] Frejlev there are three access roads to the area')
'.. mod nord er der ligeledes tre rækker træer'
('towards the North there are two rows of trees as well')
'.. i den østlige del af matrikel 2k er der en gruppe løvtræer'
('in the eastern part of the parcel 2k there is a clump of leaf-bearing trees')

When 'to be' is used to assert existence it usually appears in connection with a place marking. This is not surprising, as the corpus is concerned with the relative locations of places and terrain objects. Therefore, the meaning of existence is very close to the meaning of location, as expressed by the verb 'to lie'. The word 'lie', however, occurs only once in the discussed part of the text corpus, but the expression 'is lying' is used frequently. The construction 'is lying' applies the auxiliary verb as mentioned above.

Some of the existence expressions do not relate strictly to the route. Rather, they describe salient features of the environment of the route as a confirmation that you are still on the right track. The sentence

'Grøndalsvej er adgangsvej til et stort parcelhusområde'
([The road] 'Grøndalsvej is access road to an extensive area of housing estates')

has this function, as the route passes along the residential area, but not enters it.

Concluding this section it can be noted that verb phrases assert the existence of terrain objects, either by means of 'to be' or 'to lie' or synonyms. Terrain objects may be made agents by means of verbs of action, but in the present context the meaning is nothing but the existence and relation of terrain objects. Verbs of movement are used to direct the reader along the route, but mostly the direction giving is implicitly: A mere mentioning of roads and other features implies that this is the path to follow.

3.3 Analysis of Prepositional Phrases

The Gratex system colors the different phrase classes. If you select to have prepositional phrases highlighted the text is almost colored green. This means that prepositional phrases are used very frequently which is not surprising, as prepositional phrases are generally used for expressing spatial relations among objects.

The prepositional phrases appears in two groups: a static and a dynamic, both of which are further subdivided.

Directional Preposition Phrases

The prepositions 'til', 'ad', 'mod', 'gennem', and 'langs' ('to', 'by', 'towards', 'through', and 'along') are closely related to a movement. They frequently appears together with verbs of movement, for example:

'.. fortsætter ad ..' ('continues by')
'.. kommer til ..' ('arrives at')
'..svinger mod ..' ('turns towards')

The movement indicated by the verbs is qualified by the preposition phrase. In the Gratex system you can select compound expressions, for example a verb followed by a preposition, to confirm the statements.

A further subgroup is made of the preposition 'for' ('of') which generally appear with prefixed cardinal directions:

'.. syd-vest for byen Frejlev ..'
('South-West of the town Frejlev')

Finally, a group is made of prepositions which literally are temporal: 'før', 'efter' ('before', 'after'), but which can be spatially interpreted because of the movement down the route.

Relational Prepositional Phrases

The other main group of prepositions mark spatial relations without temporal implications. The larger of the subgroups indicates a whole - part relationship by means of the prepositions 'i', 'på', 'af' ('in', 'on', 'of'). Examples are:

'.. i den nordøstlige del ..' ('in the North-Eastern part')
'.. på venstre hånd ..' ('at the left hand')
'Den del af matriklen ..' ('the part of the parcel')
'.. østlig og vestlig side af den nord-syd gående vej'
('the East and West side of the North-South directed unpaved road')