The Universal Networking Language beyond Machine Translation1 Hiroshi Uchida, Meiying Zhu undl foundation September 26, 2001

1. Introduction

The Internet has emerged as the global information infrastructure, revolutionizing access to information, as well as the speed by which it is transmitted and received. With the technology of electronic mail, for example, people may communicate rapidly over long distances. Not all users, however, can use their own language for communication.

The Universal Networking Language (UNL) is an artificial language in the form of semantic network for computers to express and exchange every kind of information.

Since the advent of computers, researchers around the world have worked towards developing a system that would overcome language barriers.

While lots of different systems have been developed by various organizations, each has their special representation of a given language. This results in incompatibilities between systems. Then, it is impossible to break language barriers in all over the world, even if we get together all the results in one system.

Against this backdrop, the concept of UNL as a common language for all computer systems was born.

With the approach of UNL, the results of the past research and development can be applied to the present development, and make the infrastructure of future research and development.

2. UNL

The UNL can express knowledge like a natural language. The UNL can express contents like a natural language.

The situation is same as machine translation. For example, if we put all the result of research and development of machine translation, we cannot realize multilingual machine translation systems which can break language barriers.

The UNL expresses information classifying objectivity and subjectivity. Objectivity is expressed using UWs and relations. Subjectivity is expressed using attributes by attaching them to UWs.

{org:es}

{/org}

tim(begin(icl>do).@entry.@past, long ago(icl>ago))

mod(city(icl>region).@def, Babylon(icl>city))

plc(begin(icl>do).@entry.@past, city (icl>region).@def)

agt(begin(icl>do).@entry.@past, people(icl>person).@def)

obj(begin(icl>do).@entry.@past, build(icl>do))

agt(build(icl>do), people.@def)

obj(build(icl>do), tower(icl>building).@indef)

aoj(huge(icl>big), tower(icl>building).@indef)

aoj(seem(icl>be).@past, tower(icl>building).@indef)

obj(seem(icl>be).@past, reach(icl>come).@begin.@soon)

obj(reach(icl>come).@begin-soon, tower(icl>building).@indef)

gol(reach(icl>come).@begin-soon, heaven(icl>region).@def.@pl)

{/unl}

3. Universal Words

A Universal Word represents simple or compound concepts. UWs are made up of a character string (an English-language word) followed by a list of constraints. There are three kinds of UWs. Basic UWs, Restricted UWs and Extra UWs.

	::= []
	::= …
	::= “(“ [ “,” ]… “)”
	::= { “>” | “<” } [] | { “>” | “<” } [] [ { “>” | “<” } [] ] …
	::= “agt” | and” | “aoj” | “obj” | “icl” | ...
	::= “A” | ... | “Z” | “a” | ... | “z” | 0 | 1 | 2 | ... | 9 | “_” | ” “ | “#” | “!” | “$” | “%” | “=” | “^” | “~” | “|” | “@” | “+” | “-“ | “<” | “>” | “?” | “ ”

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

The Basic UW “drink”, with no Constraint List, includes the concepts of “putting liquids in the mouth”, “liquids that are put in the mouth”, “liquids with alcohol”, “absorb” and others.

The Restricted UW “drink(icl>do(obj>liquid))” denotes the subset of these concepts that includes “putting liquids in the mouth”, which in turn corresponds to verbs such as “drink”, “gulp”, “chug” and “slurp” in English.
Consider again the examples of Restricted UWs given above:

state(icl>do(obj>thing) is more specific concept

(arbitrarily associated with the English word “state”) that denotes situations

in which humans produce some information, or state something.

state(icl>nation) is more specific sense of “state” that denotes a nation.

state(icl>situation) is more specific sense of “state” that denotes a kind of situation.

state(icl>government) is more specific sense of “state” that denotes a kind of government.
The information in parentheses is the Constraint List and it describes some conceptual restrictions, that is why these are called Restricted UWs. Informally, the restrictions mean “restrict your attention to this particular sense of the word”. Thus, the focus is clearly the idea and not the specific English word.

It often turns out that for a given language there is a wide variety of different words for these concepts and not, coincidentally, all the same word, as in English.

Notice that by organizing these senses around the English words, we can simplify the task of making a new UW/Specific Language dictionary: we can use a bilingual English/Specific Language dictionary and proceed from there, specifying the number different concepts necessary for each English word.

This of course does not mean that we’re translating English words; we’re just using the English dictionary to remind us of the concepts that we will want to deal with and thus to organize work more efficiently.

samba(icl>dance) “a kind of dance”

soufflé(icl>food, pof>egg) “a kind of food made with eggs”

murano(icl>glass, aoj>colorful) “a kind of colorful glass”

4. Relations

Binary relations are the building blocks of UNL sentences. They are made up of a relation and two UWs. 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.

When a UW has relations between more than two other UWs, each relation label should be set as to be able to identify each relation on the premise that we have enough knowledge about a concept of each UW express.

When there are relations between UWs, each relation label, we should be set as to be able to understand each role of each UW only by referring a relation label.

5. Attributes

Attributes of UWs are used to describe subjectivity of sentences. They show what is said from the speaker’s point of view: how the speaker views what is said. This includes phenomena technically called “speech acts”, “propositional attitudes”, “truth values”, etc. Conceptual relations and UWs are used to describe objectivity of sentences. Attributed of UWs enrich this description with more information about how the speaker views these states-of-affairs and his attitudes toward them.

Time with respect to the speaker

Where does the speaker situate his description in time, taking his moment of speaking as a point of reference? A time before he spoke? After? At approximately the same time? This is the information that defines “narrative time” as past, present or future. These Attributes are attached to the main predicate.

Although in many languages this information is signaled by tense markings on verbs, the concept is not tense, but “time with respect to the speaker”. The clearest example is the simple present tense in English, which is not interpreted as present time, but as “independently of specific times”.

Consider the example: The earth is round.

This sentence is true in the past, in the present and in the future, independently of speaker time, so although the tense is “present” it is not interpreted as present time.

@past	happened in the past ex) He went there yesterday. ex) It was snowing yesterday
@present	happening at present ex) It’s raining hard.
@future	will happen in future ex) He will arrive tomorrow

Speaker’s view of Aspect

A speaker can emphasize or focus on a part of an event or treat it as a whole unit. This is closely linked to how the speaker places the event in time. These Attributes are attached to the main predicate.

He can focus on the beginning of the event, looking forward to it (@begin), or backward to it (@begin).

He can also focus on the end of the event, looking forward to it (@end) or backward to it from nearby (@end) or from farther away (@complete).

Degree of forwardness or backwardness (@soon, @just).

He can focus on the middle of the event (@progress).

The speaker can choose to focus on the lasting effects or final state of the event (@state) or on the event as a repeating unit (@repeat).

The feeling of incompleteness or not yet happen of an event with respect to the speaker (@yet).

@begin	beginning of an event or a state ex) It began to work again.	work.@begin.@past
@complete	finishing/completion of a (whole) event. ex) I've looked through the script	look.@entry.@complete
@continue	continuation of an event ex) He went on talking.	talk.@continue.@past
@custom	customary or repetitious action ex) I used to visit [I would often go] there when I was a boy	visit.@custom.@past
@end	end/termination of an event or a state ex) I have done it.	do.@end.@present
@experience	Experience ex) Have you ever visited Japan? ex) I have been there.	visit.@experience.@interrogation visit.@exterience
@experience	Experience ex) Have you ever visited Japan? ex) I have been there.	visit.@experience.@interrogation visit.@exterience
@progress	an event is in progress ex) I am working now.	work.@progress.@present
@repeat	repetition of an event ex) He is jumping.	jump.@entry.@present.@repeat
@state	final state or the existence of the object on which an action has been taken ex) It is broken.	break.@state

The following attributes are used to modify the attributes above.

@just	ex) He has just come.	come.@end.@just
@soon	ex) The train is about to leave.	leave.@begin.@soon
@yet	feeling of not yet begin or end/complete ex) I have not yet done it.	do.@complete.@not.@yet

Speaker's view of Reference

Whether an expression refers to a single individual, a small group or a whole set is often not clear. The expression “the lion” is not sufficiently explicit for us to know whether the speaker means “one particular lion” or “all lions”. Consider the following examples:

The lion is a feline mammal.

The lion is eating an anti-lope.

In the first example, it seems reasonable to suppose that the speaker understood “the lion” as “all lions”, whereas in the second example as “one particular lion”.

The following Attributes are used to make explicit what the speaker’s view of reference seems to be.

@generic	generic concept
@def	already referred
@indef	non-specific class
@not	complement set
@ordinal	ordinal number

These Attributes are usually attached to UWs that denote things.

Speaker’s Focus

The speaker can choose to focus or emphasize the parts of a sentence to show how important he thinks they are in the situation described. This is often related to sentence structure.

@emphasis	Emphasis ex) “I do like it”
@entry	Entry point or main UW of whole UNL expressions or in a hyper (scope) node
@qfocus	The focused UW of a question
@theme	Instantiates an object from different class
@title	Title
@topic	The topic UW of a sentence

One UW marked with "@entry" is essential to each UNL expression or in a Compound UW.

Speaker’s attitudes

The speaker can also express, directly or indirectly, what his attitudes or emotions are toward what is being said or who it is being said to. This includes respect and politeness toward the listener and surprise toward what is being said.

@confirmation	Confirmation ex) “You won't say that, will you?” ex) “sou desu ne?” (In Japanese)
@exclamation	Feeling of exclamation ex) “kirei na!”(“How beautiful (it is)!”In Japanese) ex) “Oh!, look out!”, “Ow!”
@imperative	Imperative ex) “Get up!” ex) “You will please leave the room.”
@interrogative	Interrogation ex) “Who is it?”
@invitation	Inducement to do something ex) “Will / Won’t you have some tea?” ex) “Let’s go, shall we?”
@polite	Polite feeling ex) “Could you (please)...” ex) “If you could … I would …”
@request	Request ex) “please don’t forget (@request)”
@respect	Respectful feeling ex) “o taku(@respect)”(“(your) house” in Japanese) ex) “Good morning(@respect), sir.”
@vocative	Vocative Ex) “Boys(@vocative), be ambitious!”

The variety of possibilities reflects degrees of belief, emphasis, and the extent to which what is said should be interpreted as a suggestion or order, as well as many other social factors such as the relative status of the speakers.

Speaker's view point
The following labels are used to clarify the speaker’s viewpoint information.

@ability	ability, capability of doing things ex) He can speak English but he can't write it very well.
@admire	express speaker's admiration
@although	ex) Quit smoking but he still smoke
@apodosis-real	apodosis: reality in the first person ex) We should (would) love to go abroad if we had the chance
@apodosis-unreal	apodosis: A supposed result from a supposition contrary to reality ex) If we had more money, we could buy a car.
@apodosis-cond	apodosis: A supposed result from an assumed condition ex) He would smoke too much if I did not stop him
@ask-back	ask back
@conclusion	ex) He is her husband ; she is his wife.
＠doubt	have doubt
@expectation	expectation to other's ex) He'll help you if you ask him. Will you have another cup of coffee? Will you (please, kindly, etc.) open the window? Would you excuse me ?
@grant	to give consent to do ex) Can I smoke in here? Could I smoke in here ? You may borrow my car if you like.
@grant-not	to not give consent to ex ) You {mustn't/are not allowed to/may not} borrow my car.
@induce	induce to do
@inevitability	supposition that something is inevitable ex) They should be home by now. The game will (must / should) be finished by now. Oil will float (floats) on water. He'll (always) talk for hours if you give him the chance. There must be a mistake. Mustn't there be another reason for his behavior? They ought to be here by now.
@insistence	strong will to do ex) You shall do as I say. He shall be punished. It's your own fault; you would take the baby with you.
@intention	will, intention to do ex) He shall get this money. You shall do exactly as you wish. I'll write as soon I as can.. We won't stay longer than two hours. He will do it, whatever you say. He will keep interrupting me.
@may	supposition of actual possibility ex ) We could go to the concert. The road may be blocked. We might go to the concert. What you say might be true.
@obligation	to oblige someone ex) The vendor shall maintain the equipment in good repair.
@obligation-not	forbid to do ex) You must be back by 10 o'clock. Yesterday you had to be back by 10 o'clock. Yesterday you said you must {had to} be back by 10 o'clock. You {needn't/don't have to/are not obliged to} be back by 10 o'clock.
@possibility	assume reasonable possibility ex ) Anybody can make mistakes. The road can be blocked. The road could be blocked.
@probability	assume probability ex) That would be his mother.
@regret	feel sorry
@should	to feel duty ex) You should do as he says.
@unexpected- presumption	presumption contrary to a wish or expectation ex) It is odd that you should say this to me. I am sorry that this should have happened.
@unexpected- consequence	consequence contrary to a wish or expectation ex) I made a draft, but it still needs another work.
@will	will to do ex) I shall not be long. We shall let you know our decision. We shall overcome.

Convention

Typical UNL structures can be expressed by attribute, to avoid the complexity of enconverting and deconverting. These attributes do not express speaker’s information.

@pl	Plural
@angle_bracket	< > is used
@double_parenthesis	(( )) is used
@double_quotation	“ ” is used
@parenthesis	( ) is used
@single_quotation	‘ ’ is used
@square_bracket	[ ] is used

6. Knowledge Base

The UNL Knowledge Base defines possible binary relations between UWs (Universal Words). The knowledge base is a set of knowledge base entries. The format of knowledge base entries is as follows.

::= "="
::= "(" "," ")"
::= "0" | "1" | ... | "255"
When the degree of certainty is "0", it means the relation between two UWs is false. When the degree of certainty is more than "1", it means the relation between two UWs is true, and the bigger the number is, the more the credibility is.
The UW system has been introduced to:

1. 
   generate a word when a concept is not included in a language;
   
2. 
   reduce the number of knowledge base entries which can be deductively inferred.
   

For this purpose the "icl" relation was introduced to make it possible to inherit properties from upper UW's. And each UW is categorized according to the role of concept to other concepts.

For example a knowledge "A dog can eat foods." is expressed in the following manner.
icl(dog(icl>animal), animal(icl>living thing))=1

agt (eat(icl>do(obj>thing), animal(icl>living thing))=1

obj(eat(icl>do(obj>thing), food(icl>functional thing))=1

7. The UNL System

The UNL system consists of the UNL, Language Servers and basic tools.

The UNL is consists of Universal Words(UWs), relations, attributes and knowledge base.

A language server is consists of a deconverter and an enconverter. A deconverter is a language generation system from the UNL and consists of deconversion software, generation rules and dictionaries for a language. An enconverter is a UNL generation system from a language and consists of enconversion software, analysis rules and dictionaries for a language.

Basic tools are UNL viewer to see UNL documents in their languages and UNL editor to make UNL document using their languages.

Let’s explain how UNL can be used through the Internet. Any person with access to the Internet will be able to “enconvert” text written in their own language into UNL text. And likewise, any UNL text can be “deconverted” into a variety of native languages.

To illustrate the dual processes of “enconversion” and “deconversion,” let’s look at a home page developed in Arabic. Through UNL, we will be able to read this page in Spanish.

The processes of “enconversion” and “deconversion” are provided by a Language Server which resides in the network of the Internet. The “enconverter” and “deconverter” are responsible for converting a particular language into UNL, and vice versa. The “enconverter” “enconverts” a language into UNL, while the “deconverter” “deconverts” UNL into a native language.

In this example, the Arabic Language Server and the Spanish Language Server provide the conversion service.

When home pages are developed in Arabic, the UNL Editor recognizes the contents as Arabic and sends a request to the Arabic Language Server to “enconvert” the text. Once the Arabic text is “enconverted” to UNL, the Arabic Language Server sends the results back to the UNL Editor.

Home page designers can now embed UNL into their pages.

When we read this page in Spanish, the UNL Viewer recognizes the contents as UNL and sends a request to the Spanish Language Server to “deconvert” the text. Once UNL is “deconverted” to Spanish, the Spanish Language Server sends the results back to the UNL Editor.

As you see, the text – once converted to UNL – may be converted to many different languages. For example, home pages can be designed in one’s native language and then “enconverted” to UNL before being uploaded. Once a home page is expressed in UNL, it can be read in a variety of languages.

8. Deconverter

A "deconverter" is a software that automatically deconverts UNL into native languages. It is important to achieve a high quality and correct results. It is also important that the basic architecture of the "deconverter" is widely shared throughout the world, in order to treat all languages with the same quality and precision standards. Technology developed for a language can be applied to other languages as long as the architecture is shared.

A software for deconversion called "DeCo" which constitutes a deconverter together with a word dictionary, co-occurrence dictionary and conversion rules for a language. This "DeCo" is language independent software that is applicable for any languages. The DeCO does syntactic and morphological generation ast the same time, and does knowledge base-based word generation. It also does co-occurrence-based word selection.
A "Deconverter", which generates natural language from UNL, plays a core role in the UNL system. It is very significant that "deconverter" is capable of expressing UNL information with very high accuracy.

9. Enconverter

An "enconverter" is a software that automatically or interactively enconverts natural languages text into UNL. A software for enconversion called "EnCo" which constitutes an enconverter together with a word dictionary, UNL knowledge base and conversion rules for a language. This "EnCo" is language independent software, then it is applicable for any languages. The EnCo performs morphological, syntactic and semantic analysis synchronously. It also does knowledge base and co-occurrence based disambiguation.

An "enconverter", as it generates UNL from natural languages, enables people to make UNL documents without any knowledge about UNL. It means that users of the UNL system do not need learn UNL. This makes UNL quite different from Esperanto, for instance.

10. Dictionary

A word dictionary stores information for a language. It stores information concerning what kinds of UWs(concepts) words of the language expresses and where those words can be used. A word dictionary stores the following items:

1. 
   Universal words for identifying concepts
   
2. 
   Word headings for words that can express concepts
   
3. 
   Information on the syntactical behavior of words
   

A word dictionary provides information for computers to understand natural language, and express information in natural language. A dictionary entry consists of a correspondence between a concept and a word, and information concerning syntax properties of a word when that correspondence was established.

The following shows the text format of word entries:
::= ";"
::= "[" "]"
::= "{" "}"
::= """ """
::= "("{ "," }... ")"
::= "<" "," ","
">:"
::= "A" | "B" | ... | "Z"
::=

::=

11. Conclusion

The UNL system can be used in many ways in various locations around the world. For example, it is easy for us to imagine UNL being used in fields as diverse as e-commerce, medicine, social welfare, business, libraries, and entertainment. In addition, UNL can expand the possibilities of other technologies, such as voice recognition and voice synthesis software, thereby enabling virtual communication. For example, many universities have virtual university projects and UNL could become an important technology of these. In short, we are confident that the applications of UNL will improve both access to knowledge and support distance learning throughout the world.

With the Internet serving as the global information infrastructure, UNL is the medium that facilitates communication within this infrastructure. In this way, UNL can connect – and even improve – all kinds of human activities.
As a more conrete applications of the UNL System, following application systems are considered.
Multilingual information Service

Information retrieval system

UNL based search engine

Machine translation system

Expert system

UNL should be developed by all the people in the world. Universal words necessary for each language are different from each language. The UNL is a kind of language for computer that everybody will participate to create and bring up.

1 This paper is submitted to the “International Symposium on Language in Cyberspace” held at 26 - 27 September 2001, Seoul of Korea, organized by The Korean National Commission for UNESCO and The Cyber Communication Academic Society in Korea, and sponsored by Institute of Information Technology Assessment, Rep. of Korea and UNESCO.

© Copyright UNL Centre / UNDL Foundation. All rights reserved.

Directory: publications
publications -> Acm word Template for sig site
publications ->  Preparation of Papers for ieee transactions on medical imaging
publications -> Adjih, C., Georgiadis, L., Jacquet, P., & Szpankowski, W. (2006). Multicast tree structure and the power law
publications -> Swiss Federal Institute of Technology (eth) Zurich Computer Engineering and Networks Laboratory
publications -> Quantitative skills
publications -> Multi-core cpu and gpu implementation of Discrete Periodic Radon Transform and Its Inverse
publications -> List of Publications Department of Mechanical Engineering ucek, jntu kakinada
publications -> 1. 2 Authority 1 3 Planning Area 1
publications -> Sa michelson, 2011: Impact of Sea-Spray on the Atmospheric Surface Layer. Bound. Layer Meteor., 140 ( 3 ), 361-381, doi: 10. 1007/s10546-011-9617-1, issn: Jun-14, ids: 807TW, sep 2011 Bao, jw, cw fairall, sa michelson

Download 110.74 Kb.

Share with your friends: