Top  UNL Expression   Relations   Attributes  Universal Words  UNLKB  Knowledge Representation in UNL  Logical Expression in UNL  UNL System 

UNL 2005 Specifications

7 June 2005
Copyright © UNL Center of UNDL Foundation


Chapter 4  Universal Words

Universal Words are words that constitute the vocabulary of UNL. A Universal Word (UW) is not only a unit of the UNL syntactically and semantically for expressing a concept but also a basic element for constructing a UNL expression of a sentence or a compound concept. Such a Universal Word is represented as a node in the semantic network of UNL expression.

In this chapter, how a concept of a UW is defined, how a UW for such a concept is represented, what kinds of types of UWs exist in order to deal with various kinds of knowledge provided by natural languages, and finally, how a compound concept (hyper-node) is expressed in a UNL expression (semantic network) are described. This chapter consists of the following sections:

  1. Concept of UW

  2. Syntax of UW

  3. Types of UW

  4. Compound Concept (Scope)

4.1  Concept of UW

How to define a concept depends on how the concept is to be used in a knowledge representation (or processing) system because it may affect the efficiency or power of the system. In UNL, the function of a concept is focused on in defining the concept. The function of a concept means that what semantic relations the concept can have with other concepts, or with what concepts the concept can have the semantic relations, or what roles the concept can fill in satisfying various semantic relations. From this point of view, possible relations that a concept can have with other concepts are adopted in defining the concept. The possible relations imply also that the partner concepts of these relations must be satisfactory. These possible relations not only make a concept identifiable but also describe the behavior of the concept. This definition of a concept provides information on how the concept should be used in a UNL expression.

Based on this, concepts of UWs are designed and divided into categories of nominal concepts, verbal concept, adjective concepts and adverbial concepts. This section describes how the concepts of each category are defined using the relationship property among concepts.

4.1.1  Definition of a Nominal Concept

What roles can fill in relations with other concepts, for example whether a thing can initiate something happen or whether a thing will change or can be affected in an event, is the key factor in defining nominal concepts for knowledge representation. Nominal concepts of UNL are classified into various categories taking this factor into account. And the categories are defined in different levels in order to fit the needs of roles in different levels. This hierarchical structure builds up part of the UW system which is specified in chapter 5 UNLKB.

Nominal concepts are represented by 'nominal concept', which is followed by 'thing' immediately. Under 'thing', seven categories are provided according to types of concepts. In table 4.1, the second column shows the UWs representing the seven categories under 'thing', and the third and fourth columns show some examples of Master Definitions (means to define concept of UW, see UW manual for details) of UWs. In the examples of Master Definition (MD), light-colored parts show the parts that are not included in the UWs. As shown in table 4.1, UWs 'animal(icl>living thing)', 'group(icl>abstract thing)' and 'plant(icl>living thing)' are defined for representing "animal", "group" and "human" as 'living thing', 'abstract thing' and 'living thing' respectively. With the additional link described in each MD, these UWs are also linked under 'volitional thing'.

Table 4.1
thing abstract thing group(icl>abstract thing{,icl>volitional thing})
...
concrete thing living thing{(icl>concrete thing)}
...
animal(icl>living thing{,icl>volitional thing})
plant(icl>living thing{,icl>volitional thing})
...
functional thing @
place(icl>thing) @
pronominal thing @
time(icl>thing) @
volitional thing animal(icl>living thing{,icl>volitional thing})
group(icl>abstract thing{,icl>volitional thing})
human(icl>living thing{,icl>volitional thing})
...

Nominal concepts are defined to fit the following needs:

  1. To provide appropriate concepts for restricting other concepts. For example "to die" for the meaning "to stop living" is defined taking 'living thing' as 'obj' in the restriction of UW as in 'die(obj>living thing)'. Use of the restriction 'obj>living thing' differentiates this concept from other concepts of "die" like "to stop working".
  2. To allow lower concepts to inherit the properties of upper concepts.  For example, by defining the concept of an animal to be under 'living thing', relation of 'obj(die(obj>living thing,animal(icl>living thing))' can be true as 'die(obj>living thing)' takes 'living thing' as 'obj'.

4.1.2  Definition of a Verbal Concept

A verbal concept is defined by a set of relations that the concept can have with other concepts. In other words, a definition of a verbal concept is the semantic governing information over nominal concept regarding with what kind of concept to have what kind of relation.

Verbal concepts are divided into three categories represented by 'be', 'do' and 'occur'. Table 4.2 shows the full set of relations that each category can have. These relations are described in the topmost UWs of 'be', 'do' and 'occur'. UWs for verbal concepts are linked in each of these categories through 'icl' relation in the following manner:

  1. Every UW for verbal concept has to be linked under at least one other UW through 'icl' relation as 'icl(<uw to be defined>, <upper uw>)',
  2. <uw to be defined> inherits all properties (possible relations) from <upper uw>,
  3. if necessary, possible relations can be specialized in <uw to be defined> based on the possible relations in <upper uw>, and
  4. 'aoj' relation is indispensable for a UW under 'be', 'agt' relation is indispensable for a UW under 'do' and 'obj' is indispensable for a UW under 'occur'.
Table 4.2
verbal concept 'be' -- aoj ->
-- and ->
-- ben ->
-- cao ->
-- cnt ->
-- cob ->
-- con ->
-- coo ->
-- dur ->
-- int ->
-- man ->
-- obj ->
-- or ->
-- plc ->
-- plf ->
-- plt ->
-- rsn ->
-- tmf ->
-- tmt ->
-- icl ->
partner concepts
...
verbal concept 'do' -- agt ->
-- and ->
-- ben ->
-- cag ->
-- cnt ->
-- cob ->
-- con ->
-- coo ->
-- dur ->
-- gol ->
-- ins ->
-- int ->
-- man ->
-- met ->
-- obj ->
-- or ->
-- plc ->
-- plf ->
-- plt ->
-- ptn ->
-- pur ->
-- rsn ->
-- scn ->
-- seq ->
-- src ->
-- tim ->
-- tmf ->
-- tmt ->
-- icl ->
partner concepts
...
verbal concept 'occur' -- and ->
-- ben ->
-- cnt ->
-- cob ->
-- con ->
-- coo ->
-- dur ->
-- gol ->
-- int ->
-- man ->
-- obj ->
-- or ->
-- plc ->
-- plf ->
-- plt ->
-- rsn ->
-- scn ->
-- seq ->
-- src ->
-- tim ->
-- tmf ->
-- tmt ->
-- icl ->
partner concepts
...

Table 4.3 shows how the concept of "making ideas, feelings, etc. known to somebody" expressed by word "convey" is defined. In the left side, the path of master definitions (means to define concepts, see UW manual for details) through which UW 'convey(icl>communicate(agt>thing,gol>person,obj>thing))' is defined is shown. In the right side, result by each master definition is explained. UWs for concepts defined in the middle of the path like 'do(agt>thing,gol>thing,obj>thing)' and 'do(agt>thing,gol>volitional thing,obj>thing)' are considered useful also for defining other UWs. Otherwise they can be omitted and the path can be shortened.

Table 4.3

'do'

<= topmost UW of category 'do', possible relations are shown in tabel 4.2

do({icl>do,}agt>thing,gol>thing,obj>thing{,^ptn>thing,^src>thing})

<= UW 'do(agt>thing,gol>thing,obj>thing)' is defined under 'do', this UW does not take 'ptn' and 'src' relations that are defined in the upper UW 'do'

do({icl>do(}agt>thing,{gol>thing,obj>thing),}gol>volitional thing,obj>thing)

<= UW 'do(agt>thing,gol>volitional thing,obj>thing)' is defined under 'do(agt>thing,gol>thing,obj>thing), partner concept of 'gol' is restricted from 'thing' to 'volitional thing'

do({icl>do(}agt>thing,{gol>volitional thing,obj>thing),}gol>person{>human},obj>thing)

<= UW 'do(agt>thing,gol>person,obj>thing)' is defined under 'do(agt>thing,gol>volitional thing,obj>thing)', partner concept of 'gol' is restricted from 'volitional thing' to 'person{(icl>human)}'

communicate({icl>do(}agt>thing,gol>person,obj>thing{)})

<= UW 'communicate(agt>thing,gol>person,obj>thing)' is defined under 'do(agt>thing,gol>person,obj>thing)'

convey(icl>communicate(agt>thing,gol>person,obj>thing))

<= UW 'convey(icl>communicate(agt>thing,gol>person,obj>thing))' is defined under 'communicate(agt>thing,gol>person,obj>thing)'

4.1.3  Definition of an Adjective Concept

The same as a verbal concept, an adjective concept is also defined by a set of relations that the concept can have with other concepts. In the possible relations of definitions of adjective concepts, appropriate concepts are used as partner concepts in order to satisfy the relationship required by the adjective concepts. 

There are two categories in adjective concepts according to the features of predicativity and attributivity. Attributive concepts are divided into further two categories, one consists of the attributive concepts that express quantity, and the other one consists of the attributive concepts that qualify a concept.

These three categories are represented by 'uw(aoj>thing)', 'uw(mod<thing)' and 'uw(qua<thing)'. Table 4.4 shows the full set of relations that each category can have. These relations are described in the topmost UWs of 'uw(aoj>thing)', 'uw(mod<thing)' and 'uw(quq<thing)'. UWs for adjective concepts are linked in each of these categories through 'icl' relation in the following manner:

  1. Every UW for adjective concept has to be linked under at least one other UW through 'icl' relation as 'icl(<uw to be defined>, <upper uw>)',
  2. <uw to be defined> inherits all properties (possible relations) from <upper uw>,
  3. if necessary, possible relations can be specialized in <uw to be defined> based on the possible relations in <upper uw>, and
  4. 'aoj' relation is indispensable for a UW under 'uw(aoj>thing)', 'mod' relation is indispensable for a UW under 'uw(mod<thing)' and 'qua' is indispensable for a UW under 'uw(qua<thing)'.
Table 4.4
predicative concept
  'uw(aoj>thing)'
-- aoj ->
-- and ->
-- bas ->
-- ben ->
-- cao ->
-- cnt ->
-- cob ->
-- con ->
-- coo ->
-- dur ->
-- int ->
-- man ->
-- obj ->
-- or ->
-- plc ->
-- plf ->
-- plt ->
-- rsn ->
-- icl ->
partner concepts
...
attributive concept
  'uw(mod<thing)'


  'uw(qua<thing)'

-- and ->
-- bas ->
-- int ->
<- mod --
-- or ->
-- icl ->
partner concepts
...
-- and ->
-- bas ->
-- int ->
-- or ->
<- qua --
-- icl ->
partner concepts
...

For example, "emotional" is a predicative adjective and can be considered having different meanings when it describes a person and when it describes a thing. For example, as in "an emotional person", it means "a person who has the strong feeling"; whereas in "a emotional music", it means "a music which causes people to feel strong emotions". To distinguish these different meanings, UWs for expressing a person 'person(icl>human)' and a thing 'thing' are used in the relations of definitions of these two concepts respectively. UWs for these two concepts are 'emotional(aoj>person)' and 'emotional(aoj>thing)' as shown in the master definitions in table 4.5.

Another example is "essential". The different meanings for the predicative use and for the attributive use can be represented by UWs 'essential(aoj>thing)' and 'essential(mod<thing)', which are defined by the master definitions  shown in table 4.5.

Table 4.5
emotional({icl>uw(}aoj>thing{)})

<= UW 'emotional(aoj>thing)' is defined under 'uw(aoj>thing)' to describe a thing that causes people to feel strong emotions

emotional({icl>uw(}aoj>person{)})

<= UW 'emotional(aoj>person)' is defined under 'uw(aoj>thing)' to describe a person who has the emotional feeling

essential({icl>uw(}aoj>thing{)})

<= UW 'essential(aoj>thing)' is defined under 'uw(aoj>thing)' to describe a thing that is important

essential({icl>uw(}mod<thing{)})

<= UW 'essential(mod<thing)' is defined under 'uw(mod<thing)' to qualify a thing that is fundamental

4.1.4  Definition of an Adverbial Concept

Table 4.6 shows the full set of relations to define adverbial concepts. In the topmost UW 'how', 'bas' and 'obj' are not defined as they are not frequently required by adverbial concepts. The adverbial concepts that take 'obj' relation is defined by 'how({icl>how,}obj>thing)'. 

Table 4.6
adverbial concepts
under 'how'
-- and ->
-- bas ->
<- man --
-- obj ->
-- or ->
-- icl ->
partner concepts
...

Adverbial concepts are divided into the following categories:

  1. To modify all kind of concepts such as nominal concepts, verbal concepts, adjective concepts and adverbial concepts. Concepts that belonging to this category are defined directly under 'how',
  2. Only modify adjective concepts. Concepts that belonging to this category are defined under 'how(man<adjective concept)' which is under 'how',
  3. Only modify nominal concepts.  Concepts that belonging to this category are defined under 'how(man<thing)' which is under 'how', and
  4. Only modify verbal concepts. Concepts that belonging to this category are defined under 'how(man<verbial concept)' which is under 'how'.

In each of these categories, a sub category that taking 'obj' relation is defined as:

  1. 'how(obj>thing)' is defined under 'how',
  2. 'how(man<adjective concept,obj>thing)' is defined under 'how(man<adjective concept)',
  3. 'how(man<thing,obj>thing)' is defined under 'how(man<thing)', and
  4. 'how(man<verbal concept,obj>thing)' is defined under 'how(man<verbal concept)'.

Table 4.7 shows the structure of these concepts mentioned in 1. to 8. above. In this table, every UW is the lower concept of its left side concept.

Table 4.7
how how(obj>thing) @
how(man<adjective concept) how(man<adjective concept,obj>thing)
how(man<verbal concept) how(man<verbal concept,obj>thing)

Table 4.8 shows the examples MDs of adverbial concepts in the left column, and the explanations in the right column.

Table 4.8
always{(icl>how)}

<= UW 'always' is defined under 'how'

by means of(icl>how(obj>thing))

<= UW 'by means of(icl>how(obj>thing))' is defined under 'how(obj>thing)'. This concept takes a relation 'obj'.

4.2  Syntax of UW

A UW is made up of a character string (an English-language word) followed by a list of constraints. The following is the syntax of descriptive format of UWs:

 Table 4.9

<UW>

::= <headword> [<constraint list>]

<headword>

::= <character>c

<constraint list>

::= g(g <constraint> [ g,h <constraint>]c g)h  

<constraint>

::= <relation label> { g>h | g<h } <UW> [<constraint list>] |
     <relation label> { g>h | g<h } <UW> [<constraint list>] 
     [ { g>h | g<h } <UW> [<constraint list>] ] c  

<relation label>

::= gagth | andh | gaojh | gobjh | giclh | ...

<character>

::= gAh | ... | gZh | gah | ... | gzh | 0 | 1 | 2 | ... | 9 | g_h | h g | g#h | g!h | g$h | g%h | g=h | g^h | g~h | g|h | g@h | g+h | g-g | g<h | g>h | g?h  

Headword

The headword of a UW is an English expression (a word, a compound word, a phrase or a sentence) that is interpreted as a label for a set of concepts: the set made up of all concepts that may correspond to that in English. A Basic UW (without restriction) denotes this set. Each Restricted UW denotes a subset of this set that is defined by its constraint list. Extra UWs denote new sets of concepts that do not have English-language labels.

Thus, the headword serves to organize concepts and make it easier to remember which is which.

Constraint List

The constraint list restrict the concept of a UW to a subset or to a specific concept included within the Basic UW, thus the term gRestricted UWsh. For example, the Basic UW gdrinkh, without a constraint list, includes the concepts of gputting liquids in the mouthh, gliquids that are put in the mouthh, gliquids with alcoholh, gabsorbh, etc. The Restricted UW edrink(agt>thing,obj>liquid)f denotes the subset of concept gputting liquids in the mouthh, which corresponds to verbs such as gdrinkh, ggulph, gchugh and gslurph in English.

A restriction (constraint) of a UW is made up of a pair of a relation and a pre-defined UW (or part expression of a pre-defined UW) that has the relation with this UW. If more than one restrictions are necessary, a comma g,h should be used between restrictions.  A Restricted UW is defined through a Master Definition. In a Master Definition, full expression of a pre-defined UW must be described in the restriction; whereas as for a UW, if and only if the uniqueness can be kept, part of a pre-defined UW (i.e. the headword or plus part restriction of the pre-defined UW) can be used in the restriction.

Relation Label

Relation labels used in the constraint list must be defined in the UNL specifications and should be sorted in alphabetical order if more than one constraints are used.

4.3  Types of UW

A UW is a character string and most of the UWs are basically made up of an English expression with restrictions. A UW can express various levelsf concepts depending on the restrictions and can be used to express a more specific or particular concept or an instance by giving attributes and IDs or restrictions from other UNL expressions. The UWs are divided into four types:

go
take
house
state

state(icl>express(agt>thing,gol>person,obj>thing))
state(icl>country)
state(icl>region)
state(icl>abstract thing)
state(icl>government)
 

ikebana(icl>flower arrangement)
samba(icl>dance)
soufflé(icl>food)
 

1234

4.3.1  Basic UWs

Basic UWs are character strings that correspond to English words. Such a basic UW denotes all the concepts that may correspond to those in English.  However a basic UW is not used if the English expression is ambiguous.  Such a basic UW is usually used as the headwords of Restricted UWs for its various specific concepts.  A basic UW is used if the English expression has no ambiguity.

4.3.2  Restricted UWs

Restricted UWs are by far the most important. A Restricted UW is made up of a headword (English expression) with restrictions. It is necessary when the English expression of headword has broader sense (more meanings) than the concept aimed to define. The restrictions restrict the range of the concept that an English expression represents. Each Restricted UW made from an English expression represents a more specific or particular concept, or a subset of the concepts of the English expression.

For example, following are the Restricted UWs made from the English word gstateh:

state((icl>express(agt>thing,gol>person,obj>thing)) is a more specific concept that denotes an action in which humans express something.
state(icl>country) is a more specific sense of gstateh that denotes a country.
state(icl>region) is a more specific sense of gstateh that denotes a region of a country.
state(icl>abstract thing) is a more specific sense of gstateh that denotes a kind of condition that persons or things are in. This UW is defined as a more general concept that can be referred to when defining other synonymous UWs, such as gsituationh or gconditionh.
state(icl>government) is a more specific sense of gstateh that denotes a kind of government.

The information in parentheses is the constraint list and it describes some conceptual restrictions; this is why they are called Restricted UWs. Informally, the restrictions mean grestrict your attention to this particular sense of the wordh.

4.3.3  Extra UWs

Extra UWs denote concepts that are not found in English and therefore have to be introduced as extra categories. Foreign-language words are used as headwords using English (Alphabetical) characters.

For example, following are the examples of Extra UWs:

ikebana(icl>flower arrangement) is ga kind of flower arrangementh for the meaning of gsomething you do with flowersh,
samba(icl>dance) is ga kind of danceh, and
soufflé(icl>food) is ga kind of foodh.

To the extent that these concepts exist for English speakers, they are expressed with foreign-language loanwords and do not always appear in English dictionaries. So they simply have to be added to be able to use these specific concepts in the UNL system. The restrictions give the idea of what kind of concept is associated with these Extra UWs and the constraints provide the binary relations between this concept and other, more general, concepts already defined. Needless to say, an Extra UW is also defined through a Master Definition, and a pre-defined UW or its part expression must be used in the restrictions of an Extra UW.

4.3.4  Temporary UWs

A number or an address of email that has to be used as it is not necessary to define. They can appear in a UNL document and are treated as temporary UWs.

4.4  Compound concept : Scopes

A compound concept is a set of binary relations that are grouped together to express a complex concept. A sentence itself is considered as a compound concept. Compound concepts denote complex concepts that are to be interpreted/understood as a whole so that one can talk about their parts all at the same time. A compound concept is expressed by a scope in UNL expressions. A scope makes it possible when a compound concept is necessary to be connected with other UWs.

Consider the following example:
[Women who wear big hats in movie theaters] should be asked [to leave].
The part of the sentence within square brackets is what should be asked. Only when they are grouped together and considered as a whole unit can the correct interpretation be obtained.

Attributes can be attached to them to express negation, speaker attitudes, etc., which are usually interpreted as modifying the main UW (attached with @entry) and its coordinate UWs within the scope.

4.4.1  The Way to Define a Scope

A scope is defined by placing a Scope-ID immediately after the Relation Label in all of the binary relations that are to be grouped together. Thus, in the example below, g:01h indicates all of the elements (relations) that are to be grouped together to define the scope numbered by "01".

agt:01(wear(aoj>thing,obj>hat), woman(icl>person).@pl)
obj:01(wear(aoj>thing,obj>hat), hat(icl>wear))
aoj:01(big(aoj>thing), hat(icl>wear))
plc:01(wear(aoj>thing,obj>hat), theater(icl>facilities))
mod:01(theater(icl>facilities), movie(icl>art))
agt:01(leave(agt>thing,obj>place).@entry, woman(icl>person).@pl)

After this group has been defined, wherever the Scope-ID is, for instance g01h in the above example, it can be used to cite the scope. The way to cite a scope is explained in the next section.

In a scope, one entry node marked by @entry is necessary.

4.4.2  The Way to Cite a Scope

Once defined, a scope can be cited or referred to by simply using the Scope-ID as an UW.  The method is to indicate the Scope-ID following a colon g:h.  The reference to a scope is called a Scope Node.  The Scope Node has the following syntax:

Table 4.10

<Scope Node>

::= g:h <Scope-ID> [ <Attribute List> ]

<Scope-ID>

::= two digits of a number h01h – g99h, except g00h

<Attribute List>

::= { g.h <Attribute> } c

<Attribute>

::= g@entryh | g@mayh | g@pasth | ...

To complete the UNL expression of g[Women who wear big hats in movie theaters] should be asked [to leave]h, the following are necessary:

obj(ask(agt>thing,gol>person,obj>uw).@should.@entry, :01)
gol(ask(agt>thing,gol>person,obj>uw).@should.@entry, woman(icl>person).@pl.@topic)

eobj(ask(agt>thing,gol>person,obj>uw).@should.@entry, :01)f shows that Scope 01 is the obj of gaskh.
e:01f shows the scope node. It is interpreted as the whole set of binary relations defined above. It means that g:01h should be understood as comprising all of these binary relations.  Scopes can be cited within other Scopes.


Top  UNL Expression   Relations   Attributes  Universal Words  UNLKB  Knowledge Representation in UNL  Logical Expression in UNL  UNL System