The
Universal Networking Language (UNL)
Specifications
Version 3
Edition 2
UNDL Foundation
1 July 2003
The UNL is an acronym for “Universal Networking Language”. It is a computer
language that enables computers to process information and knowledge across the
language barriers. It is an artificial language that replicates, in the cyber
world, the functions of natural languages in human communication. As a result,
it enables people to express all knowledge conveyed by natural languages. It
also enables computer to intercommunicate, thus providing people with a
linguistic infrastructure for distributing, receiving and understanding
multilingual information.
The UNL expresses information or knowledge in the form of semantic
network with hyper-node. Different from natural languages, UNL expressions are
unambiguous. In the UNL semantic network, nodes represent concepts, and arcs
represent relations between concepts. Concepts can be annotated.
Although the UNL is a language for
computers, it has all the components of a natural language. It is composed of
words expressing concepts called “Universal
words”, also referred to as UWs which are inter-linked with other UWs to form
sentences. These links, known as “relations”, specify role of each word in a sentence.
The subjective meaning intended by the speaker can be expressed through
“attributes”. In addition, the “Knowledge Base (UNLKB)” is provided to define
semantics of UWs. The UNLKB defines every possible relations between concepts including
hierarchical relations and inference mechanism based on inclusion relations
between concepts. Thus, the UNLKB provides semantic background of the UNL to
make sure the meaning of the UNL expressions is unambiguous.
Binary relations are the building blocks of UNL expressions. They are made up of a relation and two UWs. This section deals with the definition and interpretation of the binary relations of the UNL expression.
There are two forms for expressing the UNL expressions, one is the table form and the other is the list form. The table form of a UNL expression is more readable than the list form, but the list form of a UNL expression is more compact than the table form. Here, only the table form is explained and the list form is shown in Chapter 5.
Any component, such as a word, phrase or title and, of course, a sentence of a natural language can be represented with UNL expressions. A UNL expression therefore consists of a UW or a (set of) binary relation(s). In UNL documents, a UNL expression for a sentence is enclosed by the tags {unl} and {/unl} inside [S] and [/S]. If a UNL expression consists of a UW, this UW should be enclosed further by the tags [W] and [/W]. If necessary, the whole sentence can also be expressed as a scope. In this case, the Compound UW-ID of the scope should be enclosed by [W] and [/W].
Thus, the UNL expression of a sentence is the following:
{unl}
<Binary Relation>
...
{/unl}
or,
{unl}
[W]
<UW><Attribute List>
[/W]
{/unl}
or,
{unl}
[W]
”:”<Compound UW-ID><Attribute List>
[/W]
<Binary Relation>
...
{/unl}
A binary relation is made up as follows:
|
<Binary Relation> |
::= <Relation Label> [“:”<Compound UW-ID>] “(“ {{ <UW1> [":" <UW-ID1>]} | { “:” <Compound UW-ID1> }}[<Attribute List>] “,” {{ <UW2> [":" <UW-ID2>]} | { “:” <Compound UW-ID2> }}[<Attribute List>] “)” |
|
<Relation Label> |
::= a relation, see “Chapter 2: Relations” |
|
<UW> |
::= an UW, see “Chapter 3: Universal Words” |
|
<Attribute List> |
::= { “.” <Attribute> } … |
|
<Attribute> |
::= an attribute, see “Chapter 4: Attributes” |
|
<UW-ID> |
::= two characters of ‘0’ – ‘9’ and ‘A’ – ‘Z’ |
|
<Compound UW-ID> |
::= two-digit decimal number (00 – 99) 00 is used for representing the main sentence, which can be omitted. |
Compound UW-IDs are strings of two digits used to identify each instance specified by Compound UWs. Compound UWs are groups of binary relations (so-called “Scope-Nodes”) that can be referred to as a UW.
For instance, the following shows an example of a UNL expression of the sentence “I can hear a dog barking outside”.
{unl}
aoj(hear(icl>perceive(agt>thing,obj>thing)).@entry.@ability, I)
obj(hear(icl>perceive(agt>thing,obj>thing)).@entry.@ability, :01)
agt:01(bark(agt>dog).@entry, dog(icl>mammal))
plc:01(bark(agt>dog).@entry, outside(icl>place))
{/unl}
In the above UNL expressions, “aoj”, “agt” and “obj” are the relation labels, “hear(icl>perceive(agt>thing,obj>thing))”, “I”, “bark(agt>dog)”, “dog(icl>mammal)” and “outside(icl>place)” are the UWs, and “:01”, which appears three times in the example, shows the Compound UW-ID. The Compound UW-ID appears in the position of a UW, the so-called “scope-node”, and is used to cite or refer to a Compound UW previously defined. Binary relations indicated by the Compound UW-ID define the contents of the scope. A scope-node always begin with “:” followed by the two digits of a Compound UW-ID.
UW-IDs are
omitted from the above UNL expression. When a UW is unique in a UNL expression,
the UW-ID can be omitted.
The UW-ID is used to indicate some referential information, for example that there are two or more different occurrences of the same concept (they are not co-referent). Normally, if the same UW occurs more than once, it is in all cases understood to refer to the same entity or occurrence. For example, if one man greeted another man, the same UW would be used twice -- “man(icl>male person)” and it is possible to distinguish one from the other with the UW-IDs:
man(icl>male person):01 for the first and
man(icl>male person):02 for the other, to make it clear that the first man did not greet himself.
This section deals with the definition and interpretation of the relation labels of the UNL.
The relations between UWs in binary relations have different labels according to the different roles they play. These Relation-Labels are listed and defined below.
A relation label is represented as strings of 3 characters or less.
There are many factors to be considered in choosing an inventory of relations. The principles for choosing relations are as follows.
Principle-1 Necessary Condition
When
an UW has relations between more than two other UWs, each relation label should
be set so as to be able to identify each relation on the premise that there is
enough knowledge about the concept of each UW expressed.
Principle-2 Sufficient Condition
When
there are relations between UWs, each relation label should be set so as to be
able to understand the role of each UW only by referring to the relation label.
The following are the relations defined according to the above principles.
Agt defines a thing that initiates an action.
agt (do, thing)
agt [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
An agent is defined as the relation between:
UW1 - do, and
UW2 - a thing
where:
· UW2 initiates UW1, or
· UW2 is thought of as having a direct role in making UW1 happen.
|
agt(break(agt>thing,obj>thing), John(icl>person)) agt(translate(agt>thing,gol>language,obj>information,src>language), computer(icl>machine)) agt(run(icl>act(agt>volitional thing)), car(icl>vehicle)) agt(break(agt>thing,obj>thing), explosion(icl>event)) |
John breaks … computer translates … car runs … explosion breaks … |
An agent is different from cag in that an agent initiates the action, whereas a co-agent initiates a different, accompanied action.
An agent is different from ptn in that an agent is the focused initiator of the action, whereas a partner is a non-focused initiator.
An agent is different from con in that an agent is the focused initiator of the action, whereas a condition is an indirect, usually unfocussed, influence on the action.
And defines a conjunctive relation between concepts.
and (uw, uw)
and [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A conjunction is defined as the relation between:
UW1 – a concept, and
UW2 – another concept,
where:
· The UWs are different, and
· UW1 and UW2 are seen as grouped together, and
· what is said of UW1 is also said of UW2.
|
and(quickly, easily) and(sing(agt>person), dance(agt>person)) and(Mary(icl>person), John(icl>person)) |
… easily and
quickly … singing and dancing … John and Mary |
A conjunction is different from or in that with and things are grouped together to say the same thing about both of them, whereas with or we separate them to indicate that what is true about one is not true about the other.
A conjunction is different from cag in that when the agents are conjoined, both initiate an explicit event, whereas with cag, the co-agent initiates an implicit event.
A conjunction is different from ptn in that when the agents and partners are conjoined, both are in focus, whereas with ptn, the partner is not in focus (as compared to the agent).
A conjunction is different from coo and seq in meaning, although in many cases the same expressions can be used for both. A conjunction only means that terms are grouped together; no information about time is implied. Coo, on the other hand, means that the terms are in the same time, whether they are considered to be grouped together or not. In turn, seq means that the terms are ordered in time, one after the other.
Aoj defines a thing that is in a state or has an attribute.
aoj (uw(aoj>thing), thing)
aoj (thing, thing)
aoj (be, thing)
aoj [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A thing with an attribute is defined as the relation between:
UW1 – a state or a thing which represents a state, and
UW2 – a thing,
where:
· UW1 is an attribute or state of UW2, or
· UW1 is a state associated with UW2.
|
aoj(red(aoj>thing), leaf(pof>plant)) aoj(available, information) aoj(nice, ski(agt>person)) aoj(teacher(icl>occupation), John(icl>person)) aoj(have(aoj>thing,obj>thing), I) obj(have(aoj>thing,obj>thing), pen(icl>writing instrument)) aoj(know(aoj>thing,obj>thing), John(icl>person)) aoj:01(difficult(aoj>thing,obj>thing), it) aoj(:01, John(icl>person)) |
... leaf is red. This information is available for … Skiiing is nice. John is a teacher. I have a pen. John knows … It is difficult for John. |
A thing with an attribute is different from mod in that mod gives some restriction, whereas aoj gives a state or characteristic.
A thing with an attribute is different from ben in that a beneficiary is quite independent from a focused event or state. This event or state can be considered as exerting a good or bad influence, whereas aoj has a closer relation and can be considered as describing a state or characteristic.
A thing with an attribute is different from obj in that obj defines a thing which is directly affected by an action or phenomenon, whereas, aoj defines a thing in a state.
Bas defines a thing used as the basis (standard) for expressing a degree.
bas (degree, thing)
bas [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A basis is defined as the relation between:
UW1 – a degree, and
UW2 – a thing,
where:
· UW1 is a degree expressing similarity or difference, such as “more”, “most”, “less”, “same”, “similar”, “as much as”, “at least” etc., and
· UW2 is something used as the basis for evaluating the characteristics or quantity of some other (focused) thing.
|
bas(more(aoj>thing), 7) bas(more(icl>how), Jack(icl>person)) bas(same(icl>how), girl(icl>female person).@pl) bas(at least, :01) qua:01(dollar(icl>money).@pl, 500) man(beautiful, more(icl>how)) bas(more(icl>how), rose(icl>flower)) aoj(:01, John(icl>person)) man:01(quiet(aoj>thing), more(icl>how)) bas:01(more(icl>how), shy(aoj>thing,mod<thing)) |
Ten is three more than seven. Betty weighs more than Jack (does). We treat boys
exactly the same as girls It’ll cost at least 500 dollars. A tulip is more beautiful than a rose John is more quiet than shy. |
Ben defines an indirectly related beneficiary or victim of an event or state.
ben (occur, thing)
ben (do, thing)
ben (uw(aoj>thing), thing)
ben [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A beneficiary is defined as the relation between:
UW1 – an event or state, and
UW2 – a thing,
where:
· UW2 is thought of as being indirectly affected by UW1, as the beneficiary or victim.
|
ben(give(agt>thing,gol>thing,obj>thing)), Mary(icl>person)) ben(good(aoj>thing,mod<thing)), John(icl>person)) |
To give … for Mary. It is good for John to … |
A beneficiary is different from aoj in that aoj has a close relation and can be considered as describing a state characteristic, whereas a beneficiary is quite independent from a focused event or state, but this event or state can be considered as exerting a good or bad influence.
Cag defines a thing not in focus that initiates an implicit event that is done in parallel.
cag (do, thing)
cag [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A co-agent is defined as the relation between:
UW1 - an action, and
UW2 - a thing
where:
· There is an implicit action that is independent of, but accompanies, UW1, and
· UW2 is thought of as initiating the implicit action, and
· UW2 and the implicit action are seen as not being in focus (as compared to the agent’s action).
|
cag(walk(icl>do), John(icl>person)) cag(live(icl>do), aunt(icl>person)) |
To walk with John To live with … aunt |
A co-agent is different from agt in that differing independent actions occur for an agent and a co-agent. Moreover, an agent and its action are in focus, while a co-agent and its action are not in focus.
A co-agent is different from the ptn in that the co-agent initiates an action that is independent of an agent’s action, whereas a partner initiates the same action together with an agent.
A co-agent is different from con in that a co-agent initiates a non-focused action, whereas a condition is an indirect influence on the focused action.
Cao defines a thing not in focus that is in a parallel state.
cao (uw(aoj>thing), thing)
cao (thing, thing)
cao [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A co-thing with an attribute is defined as the relation between:
UW1 – a state or a thing which represents a state, and
UW2 – a thing,
where:
· There is an implicit state that is independent of, but accompanies, UW1, and
· UW2 is in an implicit state, or
· UW2 is associated with an implicit state.
|
cao(exist(icl>be), you) |
… be with you |
A co-thing with an attribute is different from aoj in that there is a different, independent state for the thing with an attribute and a co-thing with an attribute, respectively.
Cnt defines an equivalent concept.
cnt (thing, thing)
cnt [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A content is defined as the relation between:
UW1 – a thing, and
UW2 – a thing,
where:
· UW2 is a content or explanation of UW1.
|
cnt(UNL(icl>Universal Networking Language), Universal Networking Language) cnt(Internet(icl>communication network), amalgamation(icl>harmony)) cnt(language generator, deconverter.@double_quote) |
UNL, Universal
Networking Language The Internet: an
amalgamation a language generator “deconverter”… |
Cob defines a thing that is directly affected by an implicit event done in parallel or an implicit state in parallel.
cob (occur, thing)
cob (do, thing)
cob (uw(aoj>thing,obj>thing), thing)
cob [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A “co-object” is defined as the relation between:
UW1 – an event or state, and
UW2 – a thing,
where:
· UW2 is thought of as directly affected by an implicit event done in parallel or an implicit state in parallel.
|
cob(die(obj>living thing), Mary(icl>person)) obj(injure(icl>hurt(agt>thing,obj>living thing), John(icl>person)) cob(injure(icl>hurt(agt>thing,obj>living thing), friend(icl>comrade).@pl ) |
… dead with Mary John was injured in the accident with his friends |
A co-object is different from obj in that the obj is in focus, whereas cob is related to a second, non-focused implicit event or state.
Con defines a non-focused event or state that conditions a focused event or state.
con (occur, occur)
con (occur, do)
con (occur, uw(aoj>thing))
con (do, occur)
con (do, do)
con (do, uw(aoj>thing))
con (uw(aoj>thing), occur)
con (uw(aoj>thing), do)
con (uw(aoj>thing), uw(aoj>thing))
con [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A condition is defined as the relation between:
UW1 – a focused event or state, and
UW2 – a conditioning event or state,
where:
· UW1 and UW2 are different and
· UW2 is thought of as having an indirect or external role in making UW1 happen, that is, as some conditioning or inhibiting factor (real or hypothesized) that influences whether or when UW1 can happen.
|
aoj:01(tired(aoj>thing,mod<thing), you) con(go(icl>move(agt>thing,gol>place,src>place)), :01) |
If you are tired, we will go straight home. |
Coo defines a co-occurrent event or state for a focused event or state.
coo (occur, occur)
coo (occur, do)
coo (occur, uw(aoj>thing))
coo (do, occur)
coo (do, do)
coo (do, uw(aoj>thing))
coo (uw(aoj>thing), occur)
coo (uw(aoj>thing), do)
coo (uw(aoj>thing), uw(aoj>thing))
coo [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A co-occurrence is defined as the relation between:
UW1 – a focused event or state, and
UW2 – a co-occurrent event or state,
where:
· UW1 and UW2 are different, and
· UW1 occurs or is true at the same time as UW2.
|
coo(run(icl>act(agt>volitional thing)), cry(icl>weep(agt>volitional thing)) coo(red(aoj>thing,mod<thing), hot(aoj>thing,mod<thing)) |
… was crying
while running … is red while … is hot |
A co-occurrence is different from seq in that seq describes events or states that do not occur at the same time, but one after the other, whereas coo describes events that occur simultaneously.
A co-occurrence is different from tim in that coo relates the times of events or states with other events or states, whereas tim relates events or states directly with points or intervals of time.
Dur defines a period of time during which an event occurs or a state exists.
dur (occur, period)
dur (occur, event)
dur (occur, state)
dur (occur, occur)
dur (occur, do)
dur (occur, uw(aoj>thing))
dur (do, period)
dur (do, event)
dur (do, state)
dur (do, occur)
dur (do, do)
dur (do, uw(aoj>thing))
dur (uw(aoj>thing), period)
dur (uw(aoj>thing), event)
dur (uw(aoj>thing), state)
dur (uw(aoj>thing), occur)
dur (uw(aoj>thing), do)
dur (uw(aoj>thing), uw(aoj>thing))
dur [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A duration is defined as the relation between:
UW1 – an event or a state, and
UW2 – a period during which the event or state continues.
|
dur(work(agt>person), hour(icl>period)) qua(hour(icl>period), 9) dur(talk(icl>express(agt>thing,gol>person,obj>thing), meeting(icl>event) dur(come(icl>move(agt>thing,gol>place,src>place), absence(icl>state)) |
… work nine hours (a day) … talk … during meeting … come during (my) absence |
Fmt defines a range between two things.
fmt (thing, thing)
fmt [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A range (from-to) is defined as the relation between:
UW1 – a range-initial thing, and
UW2 – a range-final thing,
where:
· The UWs are different, and
· UW2 describes the beginning of a range and UW1 describes the end.
|
fmt(a(icl>letter), z(icl>letter)) fmt(Osaka(icl>city), New York(icl>city)) fmt(Monday(icl>day), Friday(icl>day)) |
the alphabets from a to z the distance from
Osaka to New York the weekdays from Monday to Friday |
A range is different from src and gol in that for src and gol the initial and final states of certain obj are characterized with respect to some event, whereas fmt makes a similar characterization but without linking the endpoints of a range to some event.
A range is different from plf and plt or tmf and tmt in that fmt defines endpoints of a range without reference to any sort of event, whereas plf, plt, tmf and tmt delimit events.
Frm defines an origin of a thing.
frm (thing, thing)
frm [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
An origin is defined as the relation between:
UW1 – a thing, and
UW2 – an origin of the thing,
where:
· UW2 describes the origin such as the original position of UW1.
|
frm(visitor(icl>person), Japan(icl>country)) |
a visitor from Japan |
Gol defines a final state of object or a thing finally associated with the object of an event.
gol (occur(gol>thing), thing)
gol (do(gol>thing), thing)
gol [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A final state is defined as the relation between:
UW1 – an event, and
UW2 – a state or thing,
where:
· UW2 is the specific state describing the obj (of UW1) at the end of UW1, or
· UW2 is a thing that is associated with the obj (of UW1) and the end of UW1.
|
gol(change(gol>thing,obj>thing,src>thing), red(aoj>thing,mod<thing)) gol(deposit(icl>save(agt>thing,obj>thing)), account(icl>record)) |
the lights changed
from green to red millions were deposited in a Swiss bank account |
A final state is different from tmf and plf in that gol describes qualitative characteristics and not time or place.
A final state is different from src in that gol describes the characteristics of the obj at the final state of the event.
Ins defines an instrument to carry out an event.
ins (do, concrete thing)
ins [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
An instrument is defined as the relation between:
UW1 – an event, and
UW2 – a concrete thing,
where:
· UW2 specifies the concrete thing that is used in order to make UW1 happen.
|
ins(look(agt>thing,obj>thing), telescope(icl>optical instrument) ins(write(icl>express(agt>thing,obj>thing)), pencil(icl>stationery)) ins(cut(agt>thing,obj>thing,opl>thing), scissors(icl>cutley)) |
look at stars through [with] a telescope write [draw]
with a pencil He cut the string with a pair of scissors |
An instrument is different from man in that man describes an event as a whole, whereas ins characterizes one of the components of the event: the use of the instrument.
An instrument is different from met in that met is used for abstract things (abstract means or methods), whereas “ins” is used for concrete things.
Man defines a way to carry out an event or the characteristics of a state.
man (occur, how)
man (do, how)
man (uw(aoj>thing), how)
man [“:”<Compound UW-ID>] “(“ {<UW1>|“:”<Compound UW-ID>} “,” {<UW2>|”:”<Compound UW-ID>} “)”
A “manner” is defined as the relation between:
UW1 – an event or state, and
UW2 – a manner,
where:
· The UWs are different, and
· UW1 is done or exists in a way characterized by UW2.
|
man(move(agt<thing,gol>place,src>place), quickly) man(visit(agt>thing,obj>thing)), often) man(beautiful, very(icl>how)) |
move quickly I often visit him. it is very beautiful. |
A manner is different from ins or met in that met describes how an event is carried out in terms of the instruments or component steps of the event, whereas man describes other quantitative or qualitative characteristics of the event as a whole.
Met defines a means to carry out an event.
met (do, abstr