CATAPHORIC LEXICALISATION

Robert Michael Foster

Research Institute in Information and Language Processing, Wolverhampton University

Wulfruna Street, Wolverhampton, WV1 1LY, U.K.

Keywords: Controlled Language, Lexicalization.

Abstract: A traditional lexicalization analysis of a word looks backwards in time, describing each change in the

word’s form and usage patterns from actuation (creation) to the present day. I suggest that this traditional

view of lexicalization can be labelled Anaphoric Lexicalization to reflect its focus upon what has passed. A

corresponding forward-looking process can then be envisaged called Cataphoric Lexicalization. Applying

Cataphoric Lexicalization to an existing phrase from a sub-language generates a series of possible lexemes

that might represent the target phrase in the future. The method is illustrated using a domain specific

example. The conclusion suggests that rigorous application of Cataphoric Lexicalization by a community

would in time result in a naturally evolved controlled lexicon.

1 INTRODUCTION

Usually when a new concept or method is

introduced into a company or community, a phrase

incorporating words that are already familiar, is used

to describe the new concept or method. The

literature (Brinton and Traugott, 2005) gives an

extensive survey of theories about Lexicalization

and Language change, which suggest that

identifiable processes are at work as language has

changed in past centuries. I propose that a company

might derive significant benefits by deliberately

applying the identified processes of language change

as their corporate sub language (Grishman and

Kitteridge 1986) accommodates new concepts and

techniques.

This paper presents a new methodology, called

Cataphoric Lexicalisation, for applying this idea. A

company who apply this method as an integral part

of their processes for writing rules and policies,

controls the actuation of new words (Weinreich

1968) as their corporate sub-language evolves

(Kirby 2007). Such control will eventually result in

the natural evolution of a controlled (Goyvaerts

1996) version of their corporate sub-language.

The UK company featured in this study applies a

system of Authorisation codes which when printed

upon a signed certificate, signify achievement of a

specified level of responsibility. The creation of new

Authorisation codes is a significant process in the

overall development of the community’s sub-

language. This paper illustrates the application of the

new methodology by describing the creation of a

new Authorisation code.

Section 2 describes the context of the study.

Section 3 describes how Cataphoric Lexicalization is

different from traditional (Anaphoric)

Lexicalisation. Section 4 describes application of the

method and section 5 lists the generated lexemes

selected by domain experts. Section 6 has

conclusions and descriptions of future work,

including plans for a software implementation.

2 CONTEXT

2.1 Authorisation Codes

The UK Company featured in this study operates a

system of authorisation certificates issued to staff

following successful completion of training and

assessment courses. The system of certificates and

associated reminder reports is managed using an in-

house computer system. Several new authorisation

codes are added to the system each year and some of

these have become assimilated as new words into

the lexicon of the company's sub-language

(Grishman and Kitteridge 1986).

Tight formatting restrictions are applied by the

computer system to the textual composition of the

526

Michael Foster R. (2010).

CATAPHORIC LEXICALISATION.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Information Systems Analysis and Speciﬁcation, pages

526-529

DOI: 10.5220/0002980005260529

 SciTePress

entry into the on-screen ‘authorisation code’ field in

the computer system. Each code is restricted to

lexemes consisting of a maximum of 7 Upper case

or 9 mixed case characters. These restrictions guard

against clipping of the codes when they are printed.

If these restrictions were not applied then there is a

risk of ambiguous specifications of permitted duties

appearing on authorisation certificates and/or lists

that remind managers of the need for refresher

training or re-authorisation interviews.

2.2 Mechanical Tree Harvesting

Power lines and other vulnerable components of a

national electricity distribution network can be

damaged during unusual weather conditions, when

trees grow towards the lines and breach specified

clearance distances. The featured company has a

progressive policy for removing such vegetation,

called Mechanical Tree Harvesting, which specifies

the safe technique for removal and subsequent

disposal of the vegetation as fuel for a carbon neutral

biomass plant.

Staff who carry out this activity have received

training in accordance with guidelines from the

Forestry commission to ensure the work is carried

out correctly. The company intends to issue these

staff with a signed certificate so they can prove that

they have been authorised to carry out this work.

This paper describes how Cataphoric Lexicalization

was used to create the Authorisation code.

3 CATAPHORIC

LEXICALIZATION

3.1 Theory

Many Lexicalization Theories appear in the

literature seeking to analyse and explain language

change. A thorough review, which also lists the

processes applied within the method, is provided in

the publication ‘Lexicalization and Language

Change’ (Brinton and Traugott, 2005).

Subsequent studies (including Okazaki et al.

2006, Gries, 2006, Baker and Brew 2008, and Cook

and Stevenson 2009) have applied these language

change theories in the analysis of historical

documents and managed corpora. In this paper these

applications of Lexicalisation theory are labelled

‘Anaphoric’ to reflect their backwards references to

Language change events that occurred before the

application of Lexicalisation theory.

In this first ever application of a formal method

of language change to the featured domain, the aim

is to accelerate naturally occurring language change

processes. The language change event will happen

after the application of Lexicalisation theory. Since

the process involves looking forwards at possible

lexemes that might reasonably represent the Target

Noun Phrase (TNP), this application of

Lexicalisation theories is given the label Cataphoric

Lexicalisation.

3.2 Process

The steps of the Cataphoric |Lexicalisation process

are as follows:

Step 1: Define the Target Noun Phrase (TNP).

Make a clear statement of the Noun Phrase for

which a new lexeme is required and identify other

Noun Phrases previously used to refer to the

concept.

Step 2: Define Corpus Boundary Criteria. The

corpus that defines the sub language within which

the new lexeme will be used must be clearly defined.

Step 3: Rank Language Change Processes. Apply

Anaphoric (traditional) Lexicalisation processes to

the defined corpus, including previous versions of

the included documents, in order to identify previous

language change events within the corpus. Then

rank the language change events in order of

frequency of occurrence. The selection of available

processes quoted here is taken from the book

‘Lexicalization and Language change’ (Brinton and

Traugott, 2005), however Cataphoric Lexicalization

is not restricted to using only these processes.

(a) Compounding

(b) Derivation

(d) Clipping / Elipsis

(e) Blending

(f) Back Formation

(g) Initialism (acronym)

(h) Coinage (root creation)

One ranking method might be to count every

occurrence of every lexeme that has resulted from

each language change process and use this simple

frequency count as the ranking criterion. An

alternative approach might be to count just once

each different lexeme formed from the application of

each language change process. Cataphoric

Lexicalization makes no restriction upon ranking of

the language change processes. The possibility exists

CATAPHORIC LEXICALISATION

527

for the method to be refined by introducing more

responsive feature calculations (Cook and Stevenson

2009). See ‘future work’.

Step 4: Apply the Identified Language Change

Processes to the TNP. Perhaps the language change

processes are applied to the TNP one at a time, but

there will be occasions when a combination of

processes is more appropriate. This is where the

ranking of observed processes within the corpus can

save time.

Step 5: Apply Domain Specific Evaluation

Processes and Choose the most Appropriate

Lexeme. Such evaluation processes will depend

upon the domain but they are likely to involve

domain experts.

4 EXPERIMENT

An illustrative, manually implemented experiment

has been included here to demonstrate Cataphoric

Lexicalization in action. The aim is to generate a

range of possible alternative signifiers for a new

authorisation code that represents the trained skills

for the activity of Mechanical Tree Harvesting.

Domain specialists will then make a selection from

the available alternatives.

Step 1: Define the TNP. The definitions identified

during initial discussions with domain experts are

listed below:

Mechanical Tree Harvesting (p):

Present participle for Process

Mechanical Tree Harvester (n):

Sense1: Operator authorised to apply the process

Sense2: Machine used to carry out the process.

The phrase ‘Tree Trimming’ was identified as

referring to a similar concept to the TNP.

Step 2: Define Corpus Boundary Criteria. The

target community is the group of people who will

interface with the UK company in relation to

protection of assets through tree trimming. Therefore

the corpus is compiled from all documents in the

company policy library that contain the signifier

‘Tree Trimming’.

Step 3: Rank Language Change Processes. No

Anaphoric Lexicalization analysis results exist for

the defined corpus and resource allocations do not

permit a full lexicalization survey, therefore a

manual ranking process is used in this case. Analysis

of the existing bank of Authorisation codes revealed

a clear preference for Initialism word creation in this

domain.

Step 4: Apply the Identified Language Change

Processes to the TNP. Some examples that were

generated using the listed language change processes

are provided below:

(a) Compounding: eg TreeHarvesting

(b) Derivation: eg Trestology …. Harvestism

(d) Clipping / Elipsis:eg MechVesting; MTHarv

(e) Blending: eg. MchTHrv… MTreHrv….

(f) Back Formation: eg Mecher…Trester

(g) Initialism (acronym): eg MTHV; MTH

(h) Coinage (creation): eg. TreVest; Tresting

Step 5: Apply Domain Specific Evaluation

Processes and choose the most Appropriate

Lexeme. This step involves domain experts

assigning a usability score to the candidate lexemes.

The results are given in the next section.

5 RESULTS

The candidate Authorisation codes that were given a

usability score of five or greater are quoted below

along with their score:

Candidate Code Usability Score

MTHarv 9

MTHV 8

RMTrestO 5

The above results highlight an observation

identified during step 3. It is clear that members of

this community have a preference for the initialism

(creation of acronyms) language change process.

However it must also be noted that an even higher

score than for the acronym MTHV was awarded to a

combination of initialism with head noun clipping.

In accordance with the recursive possibilities offered

by Cataphoric Lexicalisation, this observation will

be incorporated into future applications of the

process to this domain.

6 CONCLUSIONS AND FUTURE

WORK

In this paper the Cataphoric Lexicalization approach

to language change has been described and then

applied manually. The result is a list of candidate

lexeme signifiers for the Target Noun Phrase (TNP).

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

528

A pragmatic, domain specific evaluation of the

output from the system was then applied, providing

a more refined language change process ranking for

the featured domain in addition to a lexeme that

yielded the highest usability score.

A survey of available software that might be

adapted to implement Cataphoric Lexicalisation had

to include the default development platform for the

featured UK company which is Microsoft Access

(MSAccess). The Python based Natural Language

ToolKit (Loper and Bird 2002) was also examined

along with various commercially available software

packages, including Random Word Generator

(Wittmeyer 2009) from Gammadyne software.

Following this review of the available software

options, the decision has been taken to establish the

specified TNP without using a special tool, and then

to approach the corpus analysis and language change

process identification (Steps 2 and 3) using the

Natural Language ToolKit (Loper and Bird 2002).

Users will then be able to view outcomes from

various string concatenation functions of step 4 and

subsequent evaluation step 5 using an interface that

will be produced using MSAccess. The hope is that

the development effort will be less than might be

imagined thanks to the clearly bounded policy

document library, protected by a well organized

change management system.

In line with observations included earlier, the

rankings of language change processes in

implementations of step 3 will be refined after each

application of the method. Consideration will also be

given to including selected Grammaticalisation

theories into step 4 in order to further widen the

range of generated candidate lexemes. Perhaps

future versions of the system will also generate

suggested ranking of generated signifiers in Step 5

by making comparisons with signifiers of the same

category within the corpus or with previous

selections made by domain experts.

REFERENCES

Baker, K. and Brew, C. 2008.”Statistical identification of

English loanwords in Korean using automatically

generated training data.” In Proceedings of the Sixth

International Language Resources and Evaluation

Conference (LREC’08), pages 1159–1163, Marrakech.

Brinton, L. L., Traugott, C.G, 2005 “Lexicalisation and

Language Change” Cambridge University Press

Cook P. , and Stevenson S. 2009 "Automatically

Identifying the Source Words of Lexical Blends in

English" Computational Linguistics Volume 36,

Number 1

Goyvaerts, P. 1996 “Controlled English, curse or

blessing? A user perspective’ Proceedings of the 1st

International workshop on Controlled Language

Applications (CLAW ’96) (Leuven). 137-42

Gries, S. Th. 2006. “Cognitive determinants of subtractive

word-formation processes: A corpus-based

perspective.” Cognitive Linguistics, 17(4):535–558.

Grishman, R. and Kitteridge R. (eds) 1986 “Analys-ing

Language in restricted domains: Sublanguage

description and processing” Hillsdale NJ: Lawrence

Erlbaum Associates

Kirby, S., 2007. “The Evolution of Meaning-space

Structure through Iterated Learning” In Lyon, C.,

Nehaniv, C., and Cangelosi, A., editors, Emergence of

Communication and Language, pages 253–268.

Springer Verlag.

Loper E.and Bird S., 2002. “NLTK: The Natural

Language Toolkit.” In Proceedings of the ACL

Workshop on Effective Tools and Methodologies for

Teaching Natural Language Processing and

Computational Linguistics, pages 62–69. Somerset,

NJ: Association for Computational Linguistics.

http://arXiv.org/abs/cs/0205028.

Okazaki, N. and Ananiadou S., 2006. “A term recognition

approach to acronym recognition.” In Proceedings of

the 21st International Conference on Computational

Linguistics and the 44

Annual Meeting of the

Association for Computational Linguistics (Coling-

ACL 2006), pages 643–650, Sydney.

Weinreich, U., Labov, W., and Herzog, M.I., 1968

“Empirical foundations for a theory of language

change.” In W.P. Lehman and Yakov Malkiel, (eds.)

Directions for Historical Linguistics 97-195 Austin

and London: University of texas Press.

rights reserved. Company Registration Number:

04192923, VAT Number: GB 893 5571 78

http://www.gammadyne.com/rndword.htm

CATAPHORIC LEXICALISATION

529