sign in sign up
<<
Home , Algerian Arabic, Shadda, Tunisian Arabic

A Conventional Orthography for Algerian

Houda Saadane 1 and Nizar Habash 2

(1) Univ. Grenoble Alpes, LIDILEM, Grenoble, GEOLSemantics & Consulting, , France [email protected] (2) New York University Abu Dhabi, United Arab Emirates [email protected]

ing some specific features. In addition to MSA Abstract and DA, foreign languages, particularly French and English have been increasingly part of the is an Arabic spoken Arabic spoken in daily basis. in characterized by the absence of writing resources and standardization, hence With the emergence of Internet and social media, it is considered as an under-resourced lan- ALG (and other DAs) have become the language guage. It differs from Modern Standard Ara- of informal online communication, for instance bic on all levels of linguistic representation, from and to lexicon emails, blogs, discussion forums, SMS, etc. Most and . In this paper, we present a con- Arabic natural language processing (NLP) tools ventional orthography for Algerian Arabic, and resources were developed specially to treat following a previous effort on developing a MSA. Corresponding tools processing ALG are conventional orthography for Dialectal Ara- not as mature and sophisticated as those for bic (or CODA), demonstrated for Egyptian MSA. This is due to the recent involvement of and . We explain the design works on ALG dialect and the limit quantity of principles of Algerian CODA and provide a results and resources generated till today. To ad- detailed description of its guidelines. dress this problem, some solutions propose to apply NLP tools designed for MSA directly to 1 Introduction ALG. This proposition is interesting but yields to The Arabic language today is characterized by a significantly low performance. This is why it is complex state of polyglossia. Modern Standard necessary to develop solutions and build re- Arabic (MSA) is the official variety of Arabic sources for ALG treatment. used primarily in written literal contexts. There is also a large number of whose dominant In this paper, we present a basic layout of ALG features are noticeable to Arab-speaking people. processing which is necessary to build efficient The Arabic dialects differ from Modern Standard NLP tools and applications. This layout is a de- Arabic (MSA) on all levels of linguistic repre- sign of standard common convention orthogra- sentation, from phonology and morphology to phy dedicated to ALG dialect. The proposed lexicon and syntax. MSA is classified as a high standard is an extension of that proposed in the variety as is contains lot of normalization and work of Habash et al, (2012a) who proposed a standardization. It is generally considered as a Conventional Orthography for Dialectal Arabic prestigious, valued and ; hence (CODA). CODA is designed in order to develop it is used for training (media and education). Ar- computational models of Arabic dialects and abic Dialects (DA) are considered a low variety provided a detailed description of its guidelines which includes languages with less normaliza- as applied to (EGY). tion and standardization. These languages are used in daily life, interviews and for informal In this paper, we present a conventional orthog- conversations. Algerian Arabic (henceforth, raphy for Algerian Arabic. The paper is orga- ALG) is one of the Western group of Arabic dia- nized as follows. Section 2, discusses related lects spoken in Algeria. ALG differs from other works. In Section 3, we present an historical Arabic dialects, neighboring or far ones by hav- overview of ALG. In Section 4, we highlight the

69 Proceedings of the Second Workshop on Arabic Natural Language Processing, pages 69–79, Beijing, China, July 26-31, 2015. c 2014 Association for Computational Linguistics linguistic differences between ALG and the lan- which allows restoring the to dialects guages MSA, EGY and TUN in order to moti- corpora. And finally, they propose results on ma- vate some of our ALG CODA decisions. In Sec- chine translation between MSA and Algerian tion 5, we present ALG CODA guidelines. dialects.

2 Related works In the same way, (Harrat et al., 2015) present an Arabic multi-dialect study including dialects Studying and processing dialects is an interesting from both the and the Middle-east that recent research area which took progressively a they compare to the big attention, especially with the explosion of (MSA). Three dialects from Maghreb are con- internet public communications. Hence, there is cerned by this study: two from Algeria : Anna- actually a big interest to develop new tools to ba's dialect (ANB), the language spoken in the process an exploit the huge quantities of re- east of Algeria, on 's dialect (ALG), the sources established using dialects (oral commu- language used in the capital of Algeria, and one nications, web, social networks, etc.). However, from , on 's dialect (TUN) spoken in Arabic dialects are languages without standardi- the south of Tunisia and two dialects from Mid- zation or normalization, these why much efforts dle-east (Syria and Palestine). The resources are necessary to modernize Arabic orthography which have been built from scratch have lead to and develop orthographies for Arabic dialects. a collection of a multi-dialect parallel resource.

Maamouri et al. (2004) have developed a set of Furthermore, (Zribi, et al., 2014) extend the rules for Levantine dialects. These rules define CODA guidelines to take into account to Tunisi- the conversational transcription an dialect and (Jarrar, et al., 2014) have adapted guidelines and annotation conventions. Habsh et it to the Palestinian dialect. In addition, authors al.(2012a) have proposed a conventional orthog- of Egyptian and Tunisian CODA encourage the raphy for Egyptian dialectal (CODA). This work adaptation of CODA to other Arabic dialects in is inspired by the Linguistic Data Consortium order to create linguistic resources. Following (LDC) guidelines for transcribing. However, this council, we extend in this paper CODA CODA is intended for general purpose writing guidelines to ALG. allowing many abstracts from these variations, whereas the LDC guideline are dedicated for 3 Algerian Arabic: Historical Overview transcription, and thus focus more on phonologi- cal variations in sub-dialects. A proposition for Arabic speakers have Arabic dialects or vernacu- transcription Algerain dialect are developed in lar as their mother tongues. These dialects can be (Harrat et., 2014) where a set of rules for tran- stratified in two big families of dialects: the scription Algerain dialect are defined and a Western group (the Maghreb) or North African grapheme-to-phoneme converter for this dialect group and the Eastern group (the Mashriq). Alge- was presented. Grapheme-to-Phoneme (G2P) rian dialect, noted ALG, is one of the Western conversion or is the pro- group which is spoken in Algeria. This dialect is jazaAyriy or اي daArja ħ1 or دار cess which converts a written form of a word to also called dziyriy simply meaning "Algerian". These دزي -its pronunciation form; hence this technique fo cuses only on phonological variations. variations do not create generally barriers to un- derstand the dialect. In addition to ALG, the Al- To remedy the lack of building resources and gerian’s population speaks also Berber but with tools dedicated to the treatment of ALG issue, different ratios: ALG is used by 70 to 80% of the (Harrat et al., 2014) built parallel corpora for Al- population however; the Berber language is the gerian dialects, because their ultimate purpose is mother tongue of 25% to 30% of population. to achieve a Machine Translation (MT) for Mod- Berber is used mainly in center of Algeria (Al- ern Standard Arabic (MSA) and Algerian dia- giers and Kabylie), East of Algeria (Béjaia and lects (AD), in both directions. They also propose Sétif), in Aures (chaoui), the Mzab (north of the language tools to process these dialects. First, they developed a morphological analysis model 1 of dialects by adapting BAMA, a well-known Arabic is presented in the Habash-Soudi- Buckwalter (HSB) scheme (Habash et al., 2007). MSA analyzer. Then they propose a Phonological transcriptions will be presented between /…/ diacritization system, based on a MT process but we will use the HSB forms when possible to minimize confusion from different symbol sets.

70 Sahara) and it is used by the Twaregue based in acterized by increased French influence and the south of the Sahara (Hoggar mountains). Even if introduction of some other languages like Italian ALG is spoken by Algeria’s population, estimat- and Spanish due to migratory flow from Europe ed to 40 million of persons, it is characterized by (Ibrahimi, 2006). The influence of these lan- variation of this same dialect according to geo- guages on ALG realizes in frequent code- graphic location of ALG’s speakers. switching without any phonology adaptation in daily conversations, particularly from French, This dialect cannot be presented as homogeneous e.g., “ lycée ”, “salon”, “ quartier ”, “ normal ”, etc. linguistic system but it has many varieties. Ac- cording to (Derradji et al., 2002) we distinguish 4 Comparison among Algerian, Egyp- four varieties for ALG as follow: I) the Oranais: tian, Tunisian and Standard Arabic is the variety spoken in the Western of Algeria, precisely from Moroccan frontiers to the limit of There are many differences among ALG, EGY, Ténès, ii) Algérois: this variety covers the central TUN and MSA regarding many levels: phono- zones of Algeria to Béjaia and it is widely logical, morphological and orthographic. In this spread, iii) Rural: the speakers of this variety are section we present some of these differences that located in the East of Algeria like Constantine, are important and determinant of the distinction Annaba or Sétif, and iv) Sahara: is the dialect of between these Arabic flavors. We refer the read- the south of Algeria population. ALG is also the er to (Habash, 2010) for further elements and language used in press, television, social com- discussions. munication, internet exchanges, SMS, etc. Only 4.1 Phonological Variations in official communications, both reading and writing ones, where ALG is not used. We give in the following list the major phono- logical differences between ALG and both MSA Furthermore, we note that ALG is enriched by and EGY: q/ is one/ (ق ) the languages of the groups colonized or man- The consonant equivalent the MSA aged the Algerian population during the history of the sounds that deserve special attention. This of the country. Among these group’s languages sound has many varieties of pronunciation in we can cite: Turkish, Spanish, Italian and more Algerian Arabic dialects that we can find in the recently French. This enrichment, materialized different regions, cities and localities of Algeria. by the presence of foreign words in the dialect, Hence, the pronunciation of "q" can be realized has contributed to create many varieties of ALG as q, g, ʔ, or k. -q]: like Moroccan and Tuni] "ق" from one region to another one, with a quite • uvular stop complex linguistic situation resulting from this sian dialects, this pronunciation is present in language mixture. Indeed, this language mixture ALG in different localities as in some urban has been studied by many socio-linguistic like cities like Algiers or Constantine. g]: this sound is also used] "ڨ" Morsly, 1986; Ibrahimi, 1997; Benrabah, 1999; • palatal sound) Arezki, 2008). They described the linguistic in both Moroccan and Tunisian dialects in ad- landscape of Algeria as ' ' or dition to the ALG one. In Algeria, this sound “poly-glossic” where multiple languages and is used in some cities like Annaba and Sétif, language varieties coexist. In other words, the in addition to the dialects where this ALG is a suitable example of a complex socio- sound is widely employed. linguistic situation (Morsly, 1986). • [ʔ]: this sound is used in city in the same manner we find it in the Historically, Berber was the native language of Egyptian dialect. the population of the Maghreb in general and • k postpalatal: this sound is a particularity of Algeria in particular before the Islamic conquest, the ALG dialect that we do not find it in the which introduced Arabic in all aspects of life. other north African dialects. This sound is Centuries of various foreign powers introduced used in the rural localities and some cities like vocabulary from Turkish, Spanish and finally , , Msirda and Trara. (and most dominantly today) French. French colonization tried to impose the We note that in the case of dialects not using as the only way of communication during its 132 glottal stop consonant, there are some exceptions year control of Algeria. This situation caused a where the pronunciation is the same way regard- significant decline in the Arabic language, char- less of the dialect. This is the case of the word

71 bagra ħ‘ cow’ which is pronounced in the The pronunciation of the glottal stop phoneme ة same way using the palatal sound bagra . that appears in many MSA words in ALG dialect has different forms: -j/ has • The glottal stop becomes longue: this pronun/ (ج ) The pronunciation of the consonant also different from specific for a location or a ciation is also present in TUN and EGY dia- س : group of speakers in the north of . It is lects. We can give as example the words Di ŷb ذ ,'faAs ‘pickaxe س /pronounced [dj] in Algiers and most of central faÂs /fa’s/ → /fa:s diyb ‘wolf’, and ذ /ndjaH ‘success’, but /Di’b/ → /Di:b ح Algeria as in the word d/ con- mu ŵmin /mu’men/ → /mumin/ muwmin/ (د ) j/ precedes a/ (ج ) when the consonant sonant it will be pronounced with the ‘beliver’. [j] like in the word jdid ‘new’. In Egypt this • The glottal stop disappears: it consists on consonant is pronounced as /g/. For Tunisian, simply removing the glottal when pronouncing is the word. This form is also used in TUN and 'ج ' ,Tlemcenian and east Algerian speakers realized as /j/ or /z/ when the word contains the EGY dialects. For instance, let us take the fol- /:zarqaA’ /zarqa:’/ → /zarqa زرء :z/ like in the words lowing word/ (ز ) s/ or/ (س ) consonant .’zarqA ‘blue زر zebs ; and ز ğibs or djibs ‘plaster’ become ςzuwz. • The glottal stop is replaced by a semi- وز ςadjuwz ‘old women’ become ز /w/ or /y/: this pronunciation is found in ALG γ/ is assimilated in dif- and TUN dialects and not in EGY one. It is / (غ ) The MSA consonant أَ ﱠ َ ferent manner according to some categories of used for instance in the case of the words أ , wuk~al ّو → ’speakers. In the eastern Algerian Sahara, like /Âak~al /‘to give eating q/, /Âa ms/ ‘yesterday’ → yaAmas / (ق ) M'sila and BouSaâda, / γ/ is assimilated to for instance, the words γaAliy ‘expensive’ • The glottal stop is replaced by the letter /l/: sγayra ħ ‘small’, are pronounced re- This form is also used uniquely in the ALG ة and spectively / qaAliy /, and / sqayra /. Sometimes, it and TUN dialects unlike the EGY one. Let us x/, like Tunisian and eastern take the following examples of using of this / (خ ) is assimilated to أرض ,/Âaf ςa/ ‘snake’ → /laf ςa/ أ :Algeria speakers, e.g., the word ‘washed’ is form larD /. We note that/ ض → ’pronounced /xssel/ or / γssel/. /ÂaarD / ‘earth the given examples are also exceptions where -θ/ can be we use the same form for both definite and in / (ث ) The interdental MSA consonant .t/, in both ALG and EGY definite/ (ت ) pronounced as :/garlic’ is • The glottal stop is replaced by the letter /h م dialects like for the word ' uwm θ tuwm/. But it is also pro- opposite to the EGY dialect, the ALG and/ م pronounced as nounced / θ/ in some urban Algerian dialects as in TUN ones use this form to pronounce in some أَ ﱠ f/ like in nomadic dia- cases the glottal stop, like in the words/ (ف ) ,uwm θ م the word ﱠھ → ’lects of Mostaganem where for instance the word Âaj~aAla ħ /Âajja:la/ ‘widow Âam~aAlaA ّأ ,/θaAniy ‘also’ is pronounced faAniy ; or hajjaAla ħ /hajja:la ham~aAlaA ّھ → ’s/ in some cases in EGY dialect, for exam- /Âamma:laA / ‘however/ (س) ple, the word θaAbit ‘fixe’ is pronounced /hamma:laA/. saabit. Another MSA interdental consonant ﺳ has also special pronunciations; it is the conso- Unlike the Egyptian dialect, the Algerian dialect .ð/. In the EGY dialect, it can be pro- elides many short vowels in unstressed contexts/ (ذ ) nant ðhab ‘gold’ This feature characterizes also the other Maghreb ذھ d/, like the word / (د ) nounced :z/ for instance the dialects. This is the case of the following words / (ز ) dhab , or دھ pronounced -clever’ is realized zakiy. However, in MSA jamal ‘Camel’ (and EGY / gamal /) be‘ ذ word -ð/ has one of comes ALG / jmal/. In addition, this feature in/ (ذ ) the ALG dialect, the consonant d/. troduce an interesting element to distinguish the/ (د ) ð/ or/ (ذ ) :the following pronunciations -arm’ can be pro- Maghreb dialects from the EGY one, this ele‘ ذراع For instance the word nounced ðraA ς or draA ς. Moreover, in some re- ment is the presence of a succession of two con- gions in Algeria, like Mostaganem, this conso- sonants at the beginning of the word which in- troduces a specific particularity in the ذھ v/, like for the word / (ڢ ) nant is realized as ðhab 'gold’ pronounced vhab . scheme ‘ fςal’ in ALG instead of ‘ fa ςal’ in EGY, like in the verb MSA /qatal / ‘ killed’ (and EGY / ’atal /) becomes ALG / qtal /.

72 yuškur-uwA . To enforce ُْ ُُوا :The MSA ay and aw are generally have the form reduced uniformly to /i:/ and /u:/ . For example, this statement we refer to (Souag, 2005) work let us take the words: /HayT / ‘wall’ becomes where they defend that: “ As is common in Alge- lawn / ‘color’ becomes ALG ria, when normal short vowel would lead/ ن ,/ ALG / Hi:T /lu:n/. We note that this particularity is found in to another short vowel being in an open , the younger generation speakers; however, older we have slight lengthening on the first member ’yaD ṛab 'he hits ب :speakers still retain them in some words and con- so as to change the stress rukba ر ,”they hit" ا stills pronounced → yaD~arbuwA د texts, for instance the word ruk~ubtiy ‘my knee’ ; this ر → ’ςawd/ ‘horse’ by some old speakers. 'knee/ need not occur, however, if the con- Another feature of ALG dialect, shared with the sonant to be geminated is one of the sonorants r, TUN one, is the pronunciation of the MSA /a:/: ṛ, l, n, although for younger speakers it often in some words it is realized as /e:/ and in others does. I have the impression that these compensa- jam:al/ tory geminates are not held as long as normal/ َ َ ْل remains /a:/. For example, the word ‘beauty’ with this signification is pronounced geminates; this needs further investigation.” with /a:/ but it is realized with /e:/ in the word -jme:l/ meaning ‘camels’. Otherwise, ALG dialect uses, like the other Ara/ ْ َ ْل bic dialects, only the suffix /yn/ to form the 4.2 Morphological Variations regular plural. However, the ALG elides the ALG dialect has also some morphological as- short vowels in plural forms like in the following pects that are different from that of the MSA, and examples: ْ َ ْ ُ mulHad 'unbeliever', in the plural ,'muhandis 'engineer َُ ْ ِ ْس , closer to that of Maghreb dialects. These aspects form ِ ْ ْ ُ mulHdiyn muhandsiyn . But in some dialects, like َُ ْ ْ ِﺳ ْ .consist essentially on a simplification of some pl inflexions and inclusion of new clitics as follow: the EGY one, they don’t elide the short vowel, muhandis َُ ْ ِ ْس for instance the plural of muhandisiyn . But for َُ ْ ِ ِﺳ ْ As regards the inflexion, in ALG dialect, like 'engineer' in EGY is other Arabic ones, the casual endings in nouns some exception, like for the active and mood are lost. We note that the indica- [1A2i3] → [1A23-iyn] (Gadalla, 2000), this eli- tive mood is the one which is used as default un- sion is maintained whatever the dialect like for َ ْ ِ ْ → 'like the other moods that are not used. Moreover, the word ْ ِ َ SaAyim 'fasting the dual and the feminine plural disappeared; SaAymiyn . they are assimilated to the masculine in the plural –/ šakartun ∼a Cohen (1912) describes the emphatic suffix َ َ ُْ ﱠ form. For example, the word ‘they (fem.pl.) thank’ is normalized in the ALG tiyk/ as a characteristics of the Muslim Algiers škar-tuwA ‘they thank’. In addi- dialect that is used to express adverbs ending ْ َ ُْا dialect in tion, the first and the second person of the singu- with –a like in for the words gana ‘also’ -za ςma ‘suppos ز , lar form are conjugated in the same way in the which becomes ganaAtiyk . šakartu ‘I edly’ which becomes za ςmaAtiyk َ َ ْ ُت dialect, e.g., in MSA we say šakarta ‘you thanks’, these two َ َ ْ َت thank’ and Aista12a3] which exists in] اﺳ forms are normalized in ALG dialect in the fol- For the form -škart ‘I/you thank’. the different dialects, the ALG introduces in ad ْ َ ْ ْت :lowing unique form This simplification can lead to some ambiguities dition a new variant of this form. This variant is ssa-12a3] and it is used essentially by the] ﺳ .in ALG speakers of the west of Algeria (Marçais, 1902). Aistaklaf اِ ْﺳَ َْ ْ The ALG dialect modifies the interne form of the For example, let us take the verb ssaklaf ﱠﺳ َْ ْ verbs when it does their flexion in imperfective ‘take care of’ can be also used like . saklaf َﺳ َْ ْ form. It introduces a gemination in the first radi- or cal letter and moving to this radical the vowel of Another feature of the ALG dialect is the inser- the second one. This modification is applied only tion of vowel /i:/ between the stem and the con- in the plural form and the 2nd person of feminine sonantal suffixes of the perfect form of the pri- singular. For example, in ALG the verb ‘to mary geminate verb, e.g in MSA the verb šad ∼a/šadadtu 'he/I pulled' becomes in ّ / دت ُْ ُ ْ thank’ in 3rd person of masculine singular is yu-škur (he is thanking) and in 3rd person of ALG / ّ šad ∼/šad ∼iyt . This feature is also .yuš~ukr-uwA present in the other Arabic dialects ﱡُ ُْوا :masculine plural we have (they are thanking) but in EGY the same case

73 The passive in uses vowel For these dialects words can be spelled phono- changes and not verb derivation but in ALG as in logically or etymologically using their corre- many Arabic dialects, the passive form is ob- sponding MSA form. This fact creates some in- tained by prefixing the verb with one the follow- consistency among dialect writers. For example, ing elements: the corresponding word to ‘gold’ can be written ðhab . In addition, in some cases ذھ dhab or دھ t- / tt-, for example : tabnaý ‘it was • built’, ttarfad ‘it was lifted’ the phonology or underlying morphology is re- • n-, for instance : nftah 'it opened' flected by some regular phonological assimila- Tuwmuwbiyl ‘cars’ is ط .tn- / or /nt/, e.g., ntkal 'was edible', tion writing, e.g/ • إﺳ , Tuwnuwbiyl ط tnaqtal 'to be killed'. We note that this last also written as إﺳ element is specific for the ALG dialect. AismaA ςiyl , ‘Ismaël’ is also written as AismaA ςiyn , min ba ςd ‘after’ is also written The ALG dialect uses the particle «n» for the as mim ba ςd. Furthermore, these different first person of singular like the other Maghreb spelling can conduce to some semantic confu- šarbuwA ا dialects. This particle is generally absent from sion, like for šrbw may be the Mashreq dialects like EGY one. In those dia- ‘they drank’ or šarbuh ‘he drank it’. Finally, lects the «n» is substituted by the «a» like shown the shortened long vowels, can be spelled long or šAfw+hA/ šfw+hA ھ/ ھ ,in the following example: /naktab/ ‘I write’ short, for instance they saw her, and majaAbaš ‘he didn’t‘ ا in ALG while the equivalent of it in EGY is /Aaktib /. bring’ mAjaAbaš .

4.4 Lexical Variations Like several dialects (EGY and TUN), ALG in- clude the clictics, that are reduced forms of the As presented in Section 3, the Algerian dialect, ,like other Arabic dialects, has been influenced ه + MSA words, e.g., the proclitic ha + which strictly precedes with the definite ar- over centuries, by other languages like Berber, Al + is related to the MSA demonstrative Turkish, Italian, Spanish and French. Table 1 ال + ticle haðihi , e.g.; (MSA → shows some examples of borrowed words 2 in ھه haðaA and ھا pronouns .haðihi AldunyaA → haAldinyaA ALG ھه ا (ALG 'this life'. Words Translation Transliteration Origin a tortoise Fakruwn ون ςa+ a ,ع + Several dialects include the proclitic reduced form of the preposition /ςalaý / Moustache šliA γam Berber 'on/upon/about/to’. For example, (MSA → ALG) a throat Qarjuwma ħ Socks tqaAšiyr ة → /ςalaý AlTaAwila ħ/ او a drunkard sukaArjiy Turkish ﺳر -ςaAlmaAyda ħ 'on the table'. The same interpreta Feast Zarda ħ زردة م + f a and+ ف+ tion is valid for the proclitics Party fiyšTa ħ m+; which are the reduced form of the preposi- Italian Foul Zabla ħ ز tions fiy 'in' and min 'from' respectively. Money Suwrdiy ردي -Also, several dialect include the non-MSA nega a week siymaAna ħ ﺳ š. For example Spanish+ + ش + tion circum-clitic + mA Snickers Spardiyna ħ ﺳد mA qriyteš ‘I haven’t read’. a school Sukwiyla ħ ُﺳ

Table TaAbla ħ ط French Phone Tiyliyfuwn ن Furthermore, ALG almost lost all of the nominal

Nurse Farmliy زوج dual forms, which are replaced with the word zudwj /zu:dj/ 'two' with the plural form, e.g., Table 1: The origin and the meaning of some bor- .zuwdj rowed words used in ALG زوج → MSA → ALG) kitaAbayn) ktub 'two books' 4.3 Orthographic Variations 5 Algerian Arabic CODA Guidelines The orthographic variation in writing of Arabic In this section we present a mapping of the dialects words is due to two reasons: i) the non- CODA convention for the Algerian dialect. The existence of an orthographic standard for Arabic CODA convention is presented and its goals and dialects because these varieties are not codified and normalized, and ii) the phonological differ- ences between MSA and Algerian dialect (ALG). 2 We refer to (Guella, 2011) for more examples.

74 principals are described in details in (Habash et the map of the Algerian dialect (ALG) to CODA al, 2012a). An example of Algerian CODA is by summarizing the specific CODA guidelines presented in Table 5. for ALG. Firstly we chose a variant of the ALG which is the one used in the media as default. 5.1 CODA Guiding Principles This variant represents the dialect of the capital We summarize the main CODA design elements city Algiers and follows the same orthographic (Habash et al., 2012a, Eskander et al., 2013): rules as MSA by taking into accounts all the fol- lowing exceptions and extensions. • CODA is an internally consistent and coher- 5.3 Phonological Extensions ent convention for writing Dialectal Arabic. • CODA is created for computational purposes. Long Vowels In ALG CODA the long vowel • CODA uses the . /e:/, which do not exist in MSA, will be written • CODA is intended as a unified framework for as ay or iA depending on its MSA cognate: ay or writing all Arabic dialects. aA , respectively. In MSA orthography, the se- • CODA aims to strike an optimal balance be- quence iA is not possible, hence using words tween maintaining a level of dialectal unique- with aA MSA cognates can be a good solution ness and establishing conventions based on for ALG. This orientation is suitable since the MSA-DA similarities. basic non-diacritical form of the word is pre- daAr /da:r/ ‘turn’ and دار ,CODA is designed respecting many principles: served, for instance diAr /de:r/ ‘do’. This extension is present also in 1. CODA is an Ad Hoc convention which uses Tunisian CODA unlike the Egyptian one. only the Arabic script characters including the used for writing MSA. Vowel Shortening Like the EGY and TUN 2. CODA is consistent as it associates to each CODA, the ALG long vowels are written in long DA word a unique orthographic form that form. In some cases, which are shortened in cer- represents its phonology and morphology. tain cases such as when adding affixes and clitics ش ,CODA uses and extends the basic MSA or- even if it is writing long. For example .3 thographic decisions (rules, exceptions and mA jAb+hA+š ‘he did not forghets for her’ and tquwl lhm /tqullhum/ ‘you tell them’ (not ل -ad hoc choices), e.g., using for pho nological gemination or spelling the definite tqulhm ). This vowel shorting can be also morphemically. considered in words with two long vowels. Pho- 4. CODA generally preserves the phonological nologically, in DA, even if the two long vowels form of dialectal words given the unique are written, only one is allowed in a word, in phonological rules of each dialect (e.g., other terms, it should be only one stressed sylla- vowel shortening), and the limitations of ble in each phonological word. For instance, Arabic script (e.g., using a and a SaAymiyn ‘fasting’ (not Saymiyn ). glide consonant to write a long vowel). 5.4 Phono-Lexical Exceptions 5. CODA preserves DA morphology and syn- q/ is used to/ (ق ) tax. The Algerian "qaf" The letter 6. CODA is easy to learn and write. represent the four following : /q/, /g/ 7. The CODA principles are the same for all (like TUN), /k/ and (') (like EGY). The table 2 the dialects, however each dialect will have gives some examples of exceptional pronuncia- its proper CODA map. This unique map re- tion for /g/. spects the phonology and the morphology of CODA Pronunciation English the considered dialect. baqra ħ /bagra/ Cow ة -CODA is not a purely phonological repre .8 sentation. Text in CODA can be read per- qaAnaAtiyk /ga :na :ti :k/ so … qiAwriy /ge:wriy/ foreign وري fectly in dialect given the specific dialect

and its CODA map. Table 2: ALG exceptional pronunciation examples

5.2 Algerian CODA As we said above, CODA principles are applica- Consonant with Multiple Pronunciations ble for all dialects but with a specific map for In ALG we use the MSA forms to write conso- each dialect. Hence, in this section we present nants with multiple pronunciations. The used MSA form has to be closer to the considerate

75 consonant if it has a corresponding MSA cog- N of Number Construct The ALG CODA adds nate. We give in Table 3 some examples. Like the phoneme /n/ after some numerals in construct -sTaAšn TaAbla ħ ‘16 ta ﺳ ط ,.TUN CODA, the ALG one has more variations cases, e.g than the ones addressed in EGY CODA as for bles’ whereas the number 16 is pronounced alone sTaAš . This exception is valid for Number ﺳش the former the efforts were focused on Cairene Arabic. Hence, ALG seems to have more MSA- Construct forms with number between 11 and 19 like pronunciations where MSA spelling is simp- preceding a noun in the singular. This property is ly the same as ALG. also valid in TUN CODA.

Hamza Spelling Hamzated MSA cognate may 5.5 Morphological Extensions not be spelled in ALG CODA in a way corre- Attached clitics ALG dialect, as many other dia- sponding to the MSA cognate. In other words, lects, uses almost all the attached clitics in MSA, Al +, the future particle ال+ the glottal stop will be spelled phonologically. the definite article Ha + (expressed in east of Algeria ح + This feature is also present in EGY and TUN proclitic و CODA. However, when is pronounced in like Annaba city), the coordinating š. In + +ش ALG, we apply the same MSA spelling rules. + w+, the negation particle enclitic Furthermore, the glottal stop phoneme, appearing addition ALG uses the new attached clitics re- ه+ ,+m م+ ,+ς ع+ ,.in many MSA words, has disappeared in ALG, duced forms of the MSA, e.g f+. The following table illustrates some ف+ ,+fAs 'pickaxe' (not like MSA h س :like in the words examples of these clitics where we consider the ذ Diyb 'wolf' (not like MSA ذ ,( faÂs س wikliynaAhaAlkum ‘and we have وھ Di ŷr). In addition, words starting with Hamzated word ’eaten your food ارض ,Alif are not seen in ALG CODA, e.g .( larD ض AlAarD /larD/ ‘earth’ (not Enclitics Suffixes Stem Proclitics

و ھ ل CODA Pronunciations English kum l haA naA kliy wi ςjuwz /ςadju:z /, /ςzu:z/ old women ز وھ ςju:z/ Table 4: Tokenization of the word/ wikliynaAhaAlkum θaAniy /fa:niy/, / θa:niy/ Also

Sadr /sadr/, /Sadr/ Chest ر Separated Clitics The spelling rule for the indi- qahwa ħ /qahwa/, /gahwa/, Coffee ة

/kahwa/, /’ahwa/ rect enclitics and the negation proclitic γsal /γsal/,/xsal/ he washed mA is preserved in the ALG CODA map. This map puts a separation using a space between the γaliy /γaa: li/, /qaa:li/ Expensive negation particle and the indirect object, e.g., mA jAb lkumš /ma+jab+lkum+š/ 'he did ب faAsda ħ /fa:zda/, /fa:sda/ Corrupt ﺳة .'ðhab /ðhab/, /dhab/ Gold not give/com you ذھ /vhab/ hbaT /hbaT/, /HbaT/ he descended 5.6 Lexical Exceptions ھ

Table3: examples of multiple pronunciations in The ALG CODA, like the TUN and EGY ones, ALG. contains a list of Algerian dialect words that have a specific ad hoc spelling. This specific spelling Definite Article If the word contains the article may be inconsistent with the map of CODA in- we must distinguish between the sun and troduced above and can be spelled commonly in ,( ال) Al the moon letters. In the case of the sun letters, the different ways. These exceptions include for in- "L" is silent and the letter that follows is doubled stance: ھذو haðuwk (not ھوك gemination) in pronunciation and in writing, • The ) hakðaA ‘like this’ (not ھا ,’AnnhAr ). Con- haðuka ħ) ‘that اّر AlnnhAr 'day' (not اّر ,.e.g ھا hakdaA or ھا haAkðaA , or ھا -versely, with the moon letters, the ‘A’ is not pro nounced, the "L" of the article is pronounced and haAkdaA ) the letter that follows is not doubled, neither in • The preposition 'I know' is expressed with the Alqmar phrase َِ َْ ςlaý baAliy (not ا ,.pronunciation nor in writing, e.g ‘the moon’ (not lqmar ) (Saadane and ςambaAliy, or ςan baAliy , or Semmar, 2012; Biadsy et al., 2009). ςlabaAliy )

76 ا ر ا وا ا اأة . إء ﷲ ع اء ا Raw Text راھ إء ﷲ أم ﺳه و . إء ﷲ و ب ، ب وا وودھ . و ع ع ا وا ه أة ف اا وش راھ ف . . mrHbA bkm fy plAtw HS ħ brnAmj AlxT lHmr lnhAr Alywm ħ wlly ytzAmn m ς ςyd Almr ħ. ǍnšA' Allh gA ς AlnsA' Aly rAhm yšwfw fynA ǍnšA' Allh ÂyAm sςydh wjmyl ħ fHyAthm. ǍnšA' Allh ythnAw b mAlyhm, b wAldyhm wwlAdhm. qbl mnrwhw llmwDw ς ntA ς Alywm ħ wAly xSSnAh llmrÂħ f AljzAyr wkyfAš rAhy ςAyšħ xlwnA nrhbw bAlDywf tς lbrnAmj . ا ا ر ام وا ا ااة . ا ﷲ ع ا ا راھ CODA ا ا ﷲ ام ﺳة و . ا ﷲ وا ، ا وودھ . وا ع ع ام وا ه اة ا وش راھ ا ف ع ا . . mrHbA bkm fy blAtw HS ħ brnAmj AlxT AlHmr lnhAr Alywm wAlly ytzAmn m ς ςyd AlmrA ħ, AnšA Allh qA ς AlnsA Aly rAhm yšwfwA fynA AnšA Allh AyAm s ςyd ħ wjmyl ħ fHyAthm. AnšA Allh ythnAwA bmAlyhm, bwAldyhm wwlAdhm. qbl mA nrwhwA llmwDw ς tA ς Alywm wAlly xSSnAh llmrA ħ fAljzAyr wkfAš rAhy ςAyšħ xlwnA nrhbwA bAlDywf tA ς AlbrnAmj. English Hello everyone, in « The Red Line » daily show, which coincides with the Women's Day. God willing, for all the women who watch this show, they may have happy and beautiful days in their lives. God willing, and they will rejoice in their families, parents and children. Before addressing the topic of the day, where we focus on women in Algeria and how they are living, let's welcome to our program's guests.

Table 5: An example sentence in ALG

( za ςma ز za ςma ħ (not ز The adverbs • ‘supposedly’, Durka ħ (not Durka ) Acknowledgment ‘now’, gaAna ħ (not gaAna ) ‘also’ The first author was supported by the DGCIS (Ministry of Industry) and DGA (Ministry of In addition, in influence and integration of for- Defense): RAPID Project 'ORELO', referenced eign words from other languages, like French, by N°142906001. The second author was sup- Berber or Italian, have emerged new phonemes ported by DARPA Contract No. HR0011-12-C- like /g/, /p/ or /v/. These phonemes are used to 0014. Any opinions, findings and conclusions or express sounds that do not exist in MSA, but in recommendations expressed in this paper are CODA we will use the following Arabic charac- those of the authors and do not necessarily reflect ters: /q/, /b/ and /f/ to express respectively g, p the views of DARPA. We would like to thank jaAfiAl ‘detergent’, Bilel Gueni, Emad Mohamed and Djamel ل ,and v. For example .Belarbi for helpful feedback ون ,’kaAvi ‘stupid’, puwpiya ħ ‘doll qiyduwn ‘handlebar’.

6 Conclusions and Future Work We presented in this paper a set of guidelines towards a conventional orthography for Algerian Arabic. We discussed the various challenges of working with Algerian Arabic and how we ad- dress them. In the future, we plan to use the de- veloped guidelines to annotated collections of Algerian Arabic texts, in a first step towards de- veloping resources and tools for Algerian Arabic processing.

77 References on Human Language Technologies, Graeme Hirst, editor. Morgan & Claypool Publishers. Abdenour Arezki. 2008 . Le rôle et la place du français dans le système éducatif algérien . Re- Nizar Habash, Mona Diab and Owen Rambow. vue du Réseau des Observatoires du Français 2012a. Conventional Orthography for Dialectal Contemporain en Afrique, (23), 21-31. Arabic . In: Proceedings of the Language Re-

sources and Evaluation Conference (LREC), Is- Mohamed Benrabah. 1999. Langue et pouvoir en tanbul. Algérie: Histoire d'un traumatisme linguistique .

Seguier Editions. Nizar Habash, Ramy Eskander and Abdelati

Hawwari. 2012b . A Morphological Analyzer for Fadi Biadsy, Nizar Habash and Julia Hirschberg. Egyptian Arabic . In the Proceedings of the 2009, Improving the Arabic Pronunciation Dic- Workshop on Computational Research in Pho- tionary for Phone and Word Recognition with netics, Phonology, and Morphology Linguistically-Based Pronunciation Rules , The (SIGMORPHON) in the North American chapter 2009 Annual Conference of the North American of the Association for Computational Linguistics Chapter of the ACL, pages 397–405, Boulder, (NAACL), Montreal, Canada. Colorado.

Salima Harrat, Karima Meftouh, Mourad Abbas Marcel Cohen. 1912. Le parler arabe des Juifs and Kamel Smaïli. 2014. Grapheme To Phoneme d’Alger . Champion :Paris. Conversion-An Arabic Dialect Case . In Spoken Language Technologies for Under-resourced Yacine Derradji, Valéry Debov, Ambroise Quef- Languages. félec, Dalila S. Dekdouk and Yasmina C. Ben- chefra. 2002. Le français en Algérie : lexique et Salima Harrat, Karima Meftouh, Mourad Abbas dynamique des langues , Ed. Duclot, AUF, 2002, and Kamel Smaili. 2014 . Building Resources for 590 p. Algerian Arabic Dialects . Corpus (sentences), 4000(6415), 2415. Ramy Eskander, Nizar Habash, Owen Rambow and Nadi Tomeh. 2013. Processing Spontaneous Salima Harrat, Karima Meftouh, Mourad Abbas, Orthography. In Proceedings of Conference of Salma Jamoussi, Motaz Saad, and Kamel Smaili. the North American Association for Computa- 2015. Cross-Dialectal Arabic Processing. In tional Linguistics (NAACL), Atlanta, Georgia. Computational Linguistics and Intelligent Text Processing (pp. 620-632). Springer International Charles A. Ferguson. 1959. . Word- Publishing. Journal of the International Linguistic Associa- tion, 1959, vol. 15, no 2, p. 325-340. Khawla T. Ibrahimi. 1997. Les Algériens et leur (s) langue (s): éléments pour une approche so- Hassan A. Gadalla. 2000. Comparative Mor- ciolinguistique de la société algérienne . Éds. El phology of Standard and Egyptian Arabic (Vol. Hikma. 5). Lincom Europa. Khawla T. Ibrahimi, K. 2006. L’Algérie: coexis- Noureddine Guella. 2011. Emprunts lexicaux tence et concurrence des langues . L’Année du dans des dialectes arabes algériens . Synergies Maghreb, (I), 207-218. Monde arabe, 8, 81-88.

Mustafa Jarrar, Nizar Habash, Diyam Akra and Nizar Habash, Abdelhadi Soudi and Tim Buck- Nasser Zalmout. 2014. Building a Corpus for walter. 2007. On Arabic Transliteration . Book : a Preliminary Study . ANLP Chapter. In Arabic Computational Morphology: 2014 , 18. Knowledge-based and Empirical Methods . Edi- tors Antal van den Bosch and Abdelhadi Soudi. Mohamed Maamouri, Tim Buckwalter and Christopher Cieri. 2004. Dialectal Arabic tele- Nizar Habash. 2010. Introduction to Arabic Nat- phone speech corpus: Principles, tool design, ural Language Processing . Synthesis Lectures and transcription conventions . In NEMLAR In-

78 ternational Conference on Arabic Language Re- sources and Tools, Cairo (pp. 22-23).

William Marçais. 1902. Le dialecte arabe parlé à Tlemcen: grammaire, textes et glossaire (Vol. 26). E. Leroux.

Philippe Marçais. 1956. Le parler arabe de Djid- jelli: Nord constantinois, Algérie (Vol. 16). Li- brairie d'Amérique et d'Orient Adrien- Maisonneuve.

Dalila Morsly. 1986. Multilingualism in Algeria . The Fergusonian Impact: In Honor of Charles A. Ferguson on the Occasion of His, 65.

Houda Saadane, Aurélie Rossi, Christian Fluhr and Mathieu Guidère. 2012. Transcription of Arabic names into . In Sciences of Electron- ics, Technologies of Information and Telecom- munications (SETIT), 2012 6th International Conference on (pp. 857-866). IEEE.

Houda Saadane and Nasredine Semma. 2013. Transcription des noms arabes en écriture latine . Revue RIST| Vol, 20(2), 57.

Lameen Souag. 2005 . Notes on the Algerian Ar- abic dialect of Dellys . Estudios de dialectología norteafricana y andalusí, 9, 1-30.

Ines Zribi, Rahma Boujelbane, Abir Masmoudi, Mariem Ellouze, Lamia Belguith, and Nizar Ha- bash. 2014. A Conventional Orthography for Tunisian Arabic. In Proceedings of the Language Resources and Evaluation Conference (LREC), Reykjavik, Iceland.

79