<!DOCTYPE article PUBLIC "-//NLM//DTD JATS (Z39.96) Journal Archiving and Interchange DTD v1.0 20120330//EN" "JATS-archivearticle1.dtd">
<article xmlns:xlink="http://www.w3.org/1999/xlink">
  <front>
    <journal-meta />
    <article-meta>
      <title-group>
        <article-title>Towards Processable Afrikaans</article-title>
      </title-group>
      <contrib-group>
        <contrib contrib-type="author">
          <string-name>Laurette Pretorius</string-name>
          <email>LPretorius@csir.co.za</email>
          <xref ref-type="aff" rid="aff1">1</xref>
        </contrib>
        <contrib contrib-type="author">
          <string-name>Rolf Schwitter</string-name>
          <email>Rolf.Schwitter@mq.edu.au</email>
          <xref ref-type="aff" rid="aff0">0</xref>
        </contrib>
        <aff id="aff0">
          <label>0</label>
          <institution>Centre for Language Technology, Macquarie University</institution>
          ,
          <addr-line>Sydney 2109 NSW</addr-line>
          ,
          <country country="AU">Australia</country>
        </aff>
        <aff id="aff1">
          <label>1</label>
          <institution>School of Computing, University of South Africa and Meraka Institute</institution>
          ,
          <addr-line>CSIR, Pretoria</addr-line>
          ,
          <country country="ZA">South Africa</country>
        </aff>
      </contrib-group>
      <abstract>
        <p>In this paper we discuss a number of structural problems that we face when we design a machine-oriented controlled natural language for Afrikaans taking the underlying principles of Attempto Controlled English (ACE) and Processable English (PENG) as a starting point. Machine-oriented controlled natural languages are well-de ned subsets of natural languages that can be translated unambiguously into a formal language (and sometimes vice versa) and be used for automated reasoning. Over the last decade, a number of machine-oriented controlled natural languages have been designed and used for speci cation purposes, knowledge acquisition and knowledge representation, and as interface languages to the Semantic Web { among them Attempto Controlled English (ACE) [6], Boeing's Computer-Processable Language (CPL) [2], Common Logic Controlled English (CLCE) [9], and Processable English (PENG) [7]. All these controlled natural languages are based on English, but little work has be done in this domain for other languages. In this paper, we look at Afrikaans and work towards a machine-oriented controlled natural language for Afrikaans, based on PENG. Afrikaans is an Indo-European language derived from 17th century Dutch and mainly spoken in South Africa and Namibia. It is estimated that Afrikaans is the home language of 6.45 million people and the second or third language for 6.75 million people.3 In the following, we will discuss how a subset of ACE and PENG is related to Afrikaans and identify a number of structural problems that we face when we intend to translate Afrikaans automatically into a rst-order notation. In addition to English and Afrikaans, South Africa has nine other o cial languages that belong to the Bantu language family, and di er signi cantly from English and Afrikaans. Investigating the development of controlled natural languages for the South African Bantu languages constitutes a future challenge.</p>
      </abstract>
    </article-meta>
  </front>
  <body>
    <sec id="sec-1">
      <title>1 Introduction</title>
    </sec>
    <sec id="sec-2">
      <title>ACE and PENG</title>
      <p>ACE and PENG are machine-oriented controlled natural languages that can be
used as speci cation and knowledge representation languages. They are tractable
3 http://www.southafrica.info/about/people/language.htm
subsets of English that can be unambiguously translated by a machine into
rstorder logic. A subset of the grammar of these two languages covers declarative
sentences and questions. Their vocabularies consist of prede ned function words,
some prede ned phrases, and content words. In a nutshell, an ACE or PENG
text consists of a sequence of declarative sentences that are linked by approved
forms of intra- and intersentential anaphoric expressions.
2.1</p>
      <sec id="sec-2-1">
        <title>Simple Sentences</title>
        <p>
          Simple sentences in ACE and PENG have the following general structure:
subject + verb + [ complements ] + { adjuncts }
Every sentence has a subject and a verb. Complements are required for transitive
verbs and ditransitive verbs but adjuncts are always optional. For example, the
following sentence consists of a subject that has the form of a noun phrase
a customer, a transitive verb inserts that takes the noun phrase a card as a
complement, and an adjunct in form of a prepositional phrase into the slot:
1. A customer inserts a card into the slot.4
Not only the verb can be described in more detail as illustrated by the
prepositional modi er in (
          <xref ref-type="bibr" rid="ref1">1</xref>
          ) but also the nouns. For example in (
          <xref ref-type="bibr" rid="ref2">2</xref>
          ):
        </p>
        <p>2. John's customer who is new inserts a valid card of Mary into a slot A.
the noun customer is further described with the help of a possessive noun and
a relative clause, the noun card via an adjective and an of-prepositional phrase,
and nally the noun slot via a name as apposition.
2.2</p>
      </sec>
      <sec id="sec-2-2">
        <title>Complex/Composite Sentences</title>
        <p>
          Complex sentences (or composite sentences in \ACE-speak") are built
recursively from simpler sentences through coordination (3 and 4), subordination (
          <xref ref-type="bibr" rid="ref5">5</xref>
          ),
quanti cation (
          <xref ref-type="bibr" rid="ref6">6</xref>
          ), and negation (
          <xref ref-type="bibr" rid="ref7 ref8 ref9">7-9</xref>
          ):
3. A customer inserts a card and enters a code.
4. A customer inserts a card or enters a code.
5. If a card is valid then a customer inserts it.
6. Every customer inserts a card.
7. A customer does not insert a card.
8. No customer inserts a card.
        </p>
        <p>9. It is not the case that a customer inserts a card.</p>
        <p>Note that this is only a small subset of complex sentences that are approved by
these two controlled natural languages (for details see [6]).
4 The examples have been taken from the speci cation of an ATM [5].
2.3
ACE and PENG support yes/no-questions (10) and wh-questions (11):
10. Does the customer insert the card into the slot?
11. Where does the customer insert the card?
Yes/no-questions can be used to investigate whether or not a speci c situation
is true, and wh-questions interrogate di erent aspects of a situation.
3</p>
      </sec>
    </sec>
    <sec id="sec-3">
      <title>Afrikaans Speci cs</title>
      <p>Afrikaans and English have co-existed in South Africa for many years and the
main di erences between the two languages are well documented (see, for
example, [1]). We focus on the following ve main aspects of Afrikaans in which the
language di ers from English and consequently from ACE and PENG.
3.1</p>
      <sec id="sec-3-1">
        <title>Particle Verbs</title>
        <p>Particle verbs (phrasal verbs) (Afrikaans: partikelwerkwoorde) are common in
Afrikaans, as also in English. The verb `indruk' (insert) is one example. In
Afrikaans, the verb part of the particle verb behaves like an ordinary central
verb in terms of placement in the sentence. The complement/particle has
varying positions according to the construction [3, p.104]. For example, in main and
simple sentences without an auxiliary verb such as in (13):
12. A customer inserts a card.</p>
        <p>13. 'n Klant druk 'n kaart in.
the short complement/particle stands at the end of the sentences but in
subordinate clauses and after auxiliary verbs it is pre xed to the verb.
3.2</p>
      </sec>
      <sec id="sec-3-2">
        <title>The Preterite or Past Tense</title>
        <p>The preterite of the majority of Afrikaans verbs is formed by the auxiliary verb
`het' + the past participle of the verb. Moreover, in the presence of complements
and/or adjuncts the auxiliary verb `het' is separated from the past participle,
with the past participle occurring at the end of the sentence. For example, for
the verb `indruk' (insert), we have (15):
14. A customer inserted a card.</p>
        <p>15. 'n Klant het 'n kaart ingedruk.
3.3</p>
      </sec>
      <sec id="sec-3-3">
        <title>Subordinate Clauses</title>
        <p>Similar to English simple sentences in Afrikaans also follow the general structure:
subject + verb + complements + adjuncts (S+V+C+A)
However, in subordinate clauses the structure is usually S+C+A+V. So
composite sentences that consist of a main sentence and a subordinate clause such as
in (17) have the following structure [1, pp.46-47] where ( ) indicates optionality:
S + V(+C + A) + S(+C + A) + V
| m{azin } | subor{dzinate }
16. John's customer who is new inserts a valid card of Mary into a slot A.
17. John se klant wat nuut is druk 'n geldige kaart van Mary in 'n gleuf A in.
In composite sentences that consist of a conditional subordinate clause followed
by the main sentence such as in (19):
18. If a card is valid then a customer inserts it.</p>
        <p>19. As 'n kaart geldig is dan druk 'n klant dit in.
the main sentence has a V+S+O structure [3, p.222].
3.4</p>
      </sec>
      <sec id="sec-3-4">
        <title>Negation</title>
        <p>One of the most complex issues in Afrikaans is its double negation [1, pp.54-58].
Although double negation is the `rule' there are constructions in which not all
negations are doubled. As a guideline, negation in the main sentence is usually
doubled, even when they are composed, for example in (21):
20. It is not the case that a customer inserts a card.
21. Dit is nie die geval dat 'n klant 'n kaart indruk nie.
22. It is not the case that a customer does not insert a card.</p>
        <p>23. Dit is nie die geval dat 'n klant nie 'n kaart indruk nie.</p>
        <p>However, in subordinate clauses that form part of a main sentence that already
has negation such as in (23), negation is usually not doubled.
3.5</p>
      </sec>
      <sec id="sec-3-5">
        <title>Questions</title>
        <p>Both yes/no-questions and wh-questions in Afrikaans are obtained from
declarative sentences by means of inversion [4, p.186], for example:
24. Does a customer insert a card?
25. Druk 'n klant 'n kaart in?
26. What does a customer insert?
27. Wat druk 'n klant in?
28. Where does a customer insert a card?
29. Waar druk 'n klant 'n kaart in?
In particular, if a declarative sentence has the structure S+V+(C+A) then an
associated yes/no-question will have the structure V+S+(C+A) while an
associated wh-question has either a WH+V+S(+A) or a WH+V(+S)+A structure.</p>
      </sec>
    </sec>
    <sec id="sec-4">
      <title>First experiment</title>
      <p>The experiment was done with PENG Light [8], for which a systematically
compiled suite of approximately two hundred test sentences had been developed. As
a rst step we translated selected test sentences into Afrikaans to ascertain to
what extent PENG Light is suitable for these sentences. The PENG Light
lexicon was extended to include the Afrikaans words. We focussed on particle verbs,
the past tense and double negation, which all constitute forms of long-distance
dependencies. In all cases the Afrikaans constructions could be accommodated
by modi cations to the PENG grammar. It remains to consider to subordinate
clauses, yes/no questions and wh-questions, and to run PENG Light on the full,
translated suite of test sentences.
5</p>
    </sec>
    <sec id="sec-5">
      <title>Conclusions</title>
      <p>
        We compared a common subset of ACE and PENG with Afrikaans and identi ed
ve structural aspects in which Afrikaans di ers from these two machine-oriented
controlled natural languages: (
        <xref ref-type="bibr" rid="ref1">1</xref>
        ) particle verbs are frequent in Afrikaans and
their treatment requires a mechanism that deals with long distance dependencies
if these verbs occur in a main sentence; (
        <xref ref-type="bibr" rid="ref2">2</xref>
        ) the past tense in Afrikaans is mainly
formed by the auxiliary verb `het' + the past participle of the verb, with the
past participle occurring at the end of the sentence; (
        <xref ref-type="bibr" rid="ref3">3</xref>
        ) verbs in subordinate
clauses occupy the nal position; (
        <xref ref-type="bibr" rid="ref4">4</xref>
        ) double negation appears in main sentences
but not necessarily in subordinate clauses that depend on the main sentence;
(
        <xref ref-type="bibr" rid="ref5">5</xref>
        ) questions are built via inversion and their forms are in general simpler than
in English. PENG Light was successfully modi ed and applied to Afrikaans
sentences with particle verbs, the past tense and double negation, and will now
be extended to cater for subordinate clauses, yes/no questions and wh-questions.
      </p>
    </sec>
  </body>
  <back>
    <ref-list>
      <ref id="ref1">
        <mixed-citation>
          1.
          <string-name>
            <surname>Carstens</surname>
            ,
            <given-names>W.A.M.</given-names>
          </string-name>
          :
          <string-name>
            <surname>Norme vir Afrikaans. J.L. Van Schaik</surname>
          </string-name>
          (
          <year>1994</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref2">
        <mixed-citation>
          2.
          <string-name>
            <surname>Clark</surname>
            ,
            <given-names>P.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Harrison</surname>
            ,
            <given-names>P.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Jenkins</surname>
            ,
            <given-names>T.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Thompson</surname>
            ,
            <given-names>T.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Wojcik</surname>
          </string-name>
          , R.:
          <article-title>Acquiring and Using World Knowledge Using a Restricted Subset of English</article-title>
          .
          <source>In: Proceedings of FLAIRS05</source>
          , pp.
          <volume>506</volume>
          {
          <issue>511</issue>
          (
          <year>2005</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref3">
        <mixed-citation>
          3.
          <string-name>
            <surname>De Villiers</surname>
            ,
            <given-names>M.</given-names>
          </string-name>
          :
          <article-title>Afrikaans Grammatika vir Volwassenes</article-title>
          . Nasou
          <string-name>
            <surname>Beperk</surname>
          </string-name>
          (
          <year>1983</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref4">
        <mixed-citation>
          4.
          <string-name>
            <given-names>Du</given-names>
            <surname>Toit</surname>
          </string-name>
          ,
          <string-name>
            <surname>P.J.:</surname>
          </string-name>
          <article-title>Taalleer vir Onderwyser en Student</article-title>
          .
          <source>Academica</source>
          (
          <year>1982</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref5">
        <mixed-citation>
          5.
          <string-name>
            <surname>Fuchs</surname>
            ,
            <given-names>N.E.</given-names>
          </string-name>
          , Ho er,
          <string-name>
            <given-names>S.</given-names>
            ,
            <surname>Kaljurand</surname>
          </string-name>
          ,
          <string-name>
            <given-names>K.</given-names>
            ,
            <surname>Rinaldi</surname>
          </string-name>
          ,
          <string-name>
            <given-names>F.</given-names>
            ,
            <surname>Schneider</surname>
          </string-name>
          ,
          <string-name>
            <surname>G.</surname>
          </string-name>
          :
          <article-title>Attempto Controlled English: A Knowledge Representation Language Readable by Humans and Machines</article-title>
          .
          <source>In: Reasoning Web: First International Summer School</source>
          <year>2005</year>
          , LNCS 3564, pp.
          <volume>213</volume>
          {
          <issue>250</issue>
          (
          <year>2005</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref6">
        <mixed-citation>
          6.
          <string-name>
            <surname>Fuchs</surname>
            ,
            <given-names>N.E.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Kaljurand</surname>
            ,
            <given-names>K.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Kuhn</surname>
            ,
            <given-names>T.</given-names>
          </string-name>
          :
          <article-title>Attempto Controlled English for Knowledge Representation</article-title>
          .
          <source>In: Reasoning Web, LNCS 5224</source>
          , pp.
          <volume>10</volume>
          {
          <issue>124</issue>
          (
          <year>2008</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref7">
        <mixed-citation>
          7.
          <string-name>
            <surname>Schwitter</surname>
            ,
            <given-names>R.</given-names>
          </string-name>
          ,
          <string-name>
            <surname>Tilbrook</surname>
            ,
            <given-names>M.</given-names>
          </string-name>
          :
          <article-title>Meaningful Web Annotations for Humans and Machines using Controlled Natural Language</article-title>
          .
          <source>In: Expert Systems</source>
          ,
          <volume>25</volume>
          (
          <issue>3</issue>
          ), pp.
          <volume>253</volume>
          {
          <issue>267</issue>
          (
          <year>2008</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref8">
        <mixed-citation>
          8.
          <string-name>
            <surname>Schwitter</surname>
          </string-name>
          , R.:
          <article-title>Working for Two: A Bidirectional Grammar for a Controlled Natural Language</article-title>
          .
          <source>In:Proceedings of AI</source>
          <year>2008</year>
          , LNAI 5360, pp.
          <fpage>168</fpage>
          -
          <lpage>179</lpage>
          (
          <year>2008</year>
          )
        </mixed-citation>
      </ref>
      <ref id="ref9">
        <mixed-citation>
          9.
          <string-name>
            <surname>Sowa</surname>
            ,
            <given-names>J.F.</given-names>
          </string-name>
          :
          <article-title>Common Logic Controlled English</article-title>
          .
          <source>Draft, 24 February</source>
          <year>2004</year>
          , http://www.jfsowa.com/clce/specs.htm (
          <year>2004</year>
          )
        </mixed-citation>
      </ref>
    </ref-list>
  </back>
</article>