On the Pronunciation Dictionaries of Contemporary German: Principles of Selection and Lemmatization of Lexical Material

Abstract

1. Introduction

In view of the lexical richness and dynamics of natural languages — in this case German, a language with one of the richest vocabularies in the world — the authors of dictionaries are repeatedly confronted with the question of selecting the lexical material to be included. In the case of a pronunciation dictionary, the difficulty is even greater than usual. In comparison to a regular dictionary, it is expected that many more foreign words and, above all, many native and foreign names which are usually missing in other types of dictionaries will be considered in such a reference book:

‘The set of words and phrases given in a pronouncing dictionary is likely to differ from that in a general dictionary, since editors are inclined to feel that it is important to include a higher proportion of entries whose pronunciation is not predictable. Rare or unusual words might be included because of their difficulty or exotic nature, rather than omitted because of their infrequency or obscurity, especially in dictionaries aimed at native speakers. Loanwords and phrases borrowed from other languages feature prominently.

Pronouncing dictionaries typically include more place names, proper names, and trade names than general dictionaries do. […] Names with their origins in foreign or ancient languages are also usually well represented. Again, they merit inclusion because of the likelihood that their pronunciations will be unusual or unexpected, or that they will be a matter of debate.’ (Sangster 2015: 299)

This paper constitutes an attempt at a critical analysis of the principles of selecting the lexical material in four modern pronunciation dictionaries that were published within the last four decades and which are still used nowadays, yet to varying degrees. These are (for complete bibliographic information see References):

• Großes Wörterbuch der Deutschen Aussprache of 1982 (GWDA),

• Das Aussprachewörterbuch by DUDEN, fourth edition of 2000 (DU2000),

• Deutsches Aussprachewörterbuch of 2009 (DAWB),

• Das Aussprachewörterbuch by DUDEN, seventh edition of 2015 (DU2015).

The inclusion of the four dictionaries, among them the two older ones: GWDA and DU2000, was appropriately justified in Nycz and Tęcza (2019, in print).

In its most important aspects, the following analysis makes a reference to the outcome of our previous research of three pronunciation dictionaries (excluding the most recent one) as presented in Tęcza and Nycz (2016) and, to a lesser extent, to the comparative description of the latest two dictionaries provided in Tęcza (2018).

2. Lemma selection in diverse vocabulary groups

In general, the starting point for compiling the word list of a dictionary of any type is a ‘lemma candidate list’, which is obtained from various sources (more on this in Section 4). ‘In the end, the lexicographer must decide which words or which lexicographical items should be estimated as important for a dictionary. […] The selection of the lemmata can be based on corpus evaluations (frequency data), but can also be orientated at the lemma lists of other dictionaries.’ (Schierholz 2015: 332; see also Section 4). In the selection process, the dictionary type must be taken into account and the focus is on the needs and expectations of the dictionary user: ‘You need a clear understanding of who will use the dictionary, what they will use it for, and what kinds of skill they will bring to the task.’ (Atkins and Rundell 2008: 28; on user profiling and user research see ibid. 28–32).

In principle, all the reference works to be analysed below address the needs of both native and non-native speakers (cf. GWDA: 5, DU2000: 5, DAWB 4^th cover page, DU2015: 9–10).¹ This fact inevitably has an influence on the composition of their word lists:

‘Native speakers look to pronunciation dictionaries to find authoritative information on unfamiliar or controversial words, to improve their own pronunciation, to educate or advise others, or to settle arguments. […]

Non-native speakers use general pronouncing dictionaries in broadly the same way as native speakers; to support their learning and to increase the accuracy of their pronunciation of the relevant language. Depending on their level of fluency and awareness of the language’s pronunciation patterns, a non-native speaker might wish to look up a wider range of words than a native speaker, who might focus more on the complicated, controversial, or unusual words. […]

Striking a balance when it comes to entry selection, to serve the needs of both groups of users, is a particular challenge for editors of major pronouncing dictionaries. The material chosen must be broad and complex enough to interest the expert native speaker with a large vocabulary, without being so arcane as to alienate other users.’ (Sangster 2015: 294–295)

Out of the examined dictionaries, DAWB is the one that most fully defines its selection criteria. In Chapter Three of the introductory matter entitled ‘Struktur und Auswahl des Wortschatzes’, the dictionary user is familiarized with the following:

‘Die Beschreibung der bundesdeutschen Standardaussprache in diesem Buch bezieht sich auf Wörter und Namen, die in der öffentlichen Sprechkommunikation gebraucht werden und hauptsächlich nicht zu den Fachsprachen, beispielsweise der Medizin, Informatik oder Physik, gehören. Sie sind deutscher oder fremder Herkunft und können einerseits als Simplizia […] und andererseits als Komposita […], als Bindestrichwörter und als Verbindungen eines Grundwortes mit Präfixen oder Suffixen auftreten. […] Hinzu kommen feste Wortgruppen sowie einfache und verschieden zusammengebildete Namen. Mit den rund 150.000 Eintragungen im Wörterverzeichnis wird versucht, dieser Differenziertheit und Vielfalt gerecht zu werden.’ (DAWB: 17)

However, it seems obvious that the entire remaining lexicon of the standard language, even after the exclusion of the strictly specific terminology, cannot be included (cf., for instance, Landau 2003: 95). Thus, a further selection is required according to certain criteria:

‘Da es aber unmöglich ist, alle in der standardgemäßen öffentlichen Kommunikation verwendeten Ausdrücke und Benennungen in ein solches Buch aufzunehmen, musste bei der Auswahl mit Blick auf die Benutzerinteressen differenziert werden. Dabei wurde […] auch die Fähigkeit in Rechnung gestellt, aus angebotenen Beispielen oder Mustern […] Rückschlüsse auf nicht aufgenommene Bezeichnungen zu ziehen. Die wichtigsten Auswahlkriterien waren wechselweise Aktualität bzw. Gebrauchshäufigkeit und orthoepischer Anspruch. Die systematische Darbietung der verschiedenen Gruppen von Wörtern und Namen war dabei nachrangig. Vollständigkeit ließ sich bei keiner Gruppe erreichen, jedoch gab es Stufungen. […] Relative Vollständigkeit konnte nur bei den Simplizia deutscher Herkunft und solchen, die als ältere Entlehnungen einzustufen sind, angestrebt werden.’ (DAWB: 20–21, italics by the authors)

These guidelines can be similarly applied to the three other dictionaries: Non-simplicia, which are extraordinarily numerous and very frequently used in public communication, were recorded mainly when their phonetic realization differs from the pronunciation of their constituents in free use, for example, as a result of changed accent structures or sound assimilations at the constituent boundary (cf. GWDA: 146, DU2000: 20, DAWB: 278 and DU2015: 14; see also the argumentation in Hirschfeld and Stock 2006a: 97–98, and Hirschfeld and Stock 2006b: 50–51). A clue that can be found explicitly in the DUDEN editions may also be applied to all four dictionaries, namely that all words have been recorded for which more than one pronunciation variant seems to be appropriate (DU2015: 14; see also DU2000: 18).

However, the authors of DAWB also declare, in the case of particular simplicia, to have striven to exemplify the most important types of the respective derivatives and compounds (cf. 21); and the authors of DU2015 claim to have included some very common compounds, even though they do not show any peculiarities in pronunciation (cf. 14). On the other hand, DU2000 is the only one among the four dictionaries to include a tabular list of transcribed German affixes in its introductory matter (see 25). It is very useful for quick consultation, and can be significantly helpful if a derivative is missing in the dictionary. This compilation, however, was not transferred to the latest edition DU2015. There is in fact a section dealing with the phonetic realization of affixes (‘Nebenton und Präfixe/Suffixe’, cf. 72f.) in DU2015 ’s chapter on ‘Variation in der Standardaussprache’, but since it mainly discusses some problematic aspects and the geographical differentiation of the pronunciation of affixes (or the vowels contained therein), it can hardly be used for quick consultation.

Considering the total volume of each dictionary, measured by the total number of entries, it would be reasonable to expect that the recording frequency of complex words be the highest in DAWB and the lowest in GWDA. Random samples seem to confirm this assumption at least as far as DAWB is concerned (see Table 1).

Despite the small number of randomly chosen first constituents included in Table 1, it may well be assumed that in the two DUDEN dictionaries disproportionately fewer complex words in relation to the total number of keywords are recorded: GWDA, whose total number of entries is less than half that of either of the DUDEN editions, contains a similar number of complex word lemmas, and DAWB, whose total amount of entries is only about 15% larger, several times as many. In his overall very critical review of DAWB, Kleiner (2011: 91) considers the increased inclusion of complex words in the index as one of only two advantages of this dictionary (the other one being its description of the national standard pronunciations in Austria and in Switzerland).

Furthermore, in Table 1, three differences between DU2000 and DU2015 occur: In two cases an entry has been added to DU2015 (Jugendliche, Waschschüssel), and in one case an entry (Kurzstreckler) still present in DU2000 has been removed (see Table 1).

A particular group of word formation constructions are the prefixed verbs because they are usually widely represented in various dictionaries, even in those with a smaller number of entries. The two DUDEN editions proceed untypically in this respect insofar as they mostly do not record prefixed verbs:

‘Präfigierte und zusammengesetzte Verben wurden im Allgemeinen nur dann aufgenommen, wenn der zweite Teil des Verbs allein nicht vorkommt. So wurden z.B. die Verben befleißen und ausmergeln aufgenommen, weil *fleißen und *mergeln allein nicht existieren. Dagegen wurden etwa die Verben beachten und aussäen nicht berücksichtigt, weil die einfachen Verben achten und säen verzeichnet sind.’ (DU2000: 18)

This means that both DU2000 and DU2015 basically omit among others those cases in which prefixation affects the pronunciation, for example by voicing assimilation or gemination of consonants at the syllable boundary (abdecken, abbilden, aussäen etc.). In contrast, in GWDA and especially in DAWB there are many prefixed verbs (see Table 2).

An important type of derivation: the female noun forms denoting professions or persons derived in a regular manner from the corresponding masculine forms, are treated somewhat inconsistently in the dictionaries under analysis. In DU2000 and in DAWB, there is a statement concerning this matter, namely that the authors have basically omitted such derivatives (see DU2000: 21 and DAWB: 278), irrespective of the accent shift that occurs in some of them, for instance nouns ending with -or-in (such as Senator [zeˈna:to:^ɐ], but: Senatorin [zenaˈto:rɪn]).

However, as the examples in Table 3 indicate, the authors of the two dictionaries have not consistently implemented the directive they themselves formulated. Table 3 also shows that DU2015, whose authors fail to comment on their approach in more detail,² differs from its predecessor DU2000 probably just by a single entry, but a very relevant one, namely Bundeskanzlerin, as far as feminine forms of professional titles and persons are concerned. In the other dictionaries, only the masculine form Kanzler, in DAWB additionally Bundeskanzler, can be found. Notably, DU2000 does not list any compounds with Bundes-, which is a very characteristic constituent of a number of public institutions’ names in Germany. In DU2015 this deficiency is made up to a certain extent. Along with Bundeskanzler and Bundeskanzlerin, it includes Bundeskanzleramt, Bundesgericht, Bundesliga, Bundesregierung, and Bundestag. However, DAWB lists more than thirty compounds of this type, including such as Bundestagsneuwahl or Bundeswehrreform, the use of which is mainly limited to certain phases of political life. The East German GWDA has also recorded at least one entry including the constituent in question, namely the official state name Bundesrepublik Deutschland.

Getting back to the feminine derivatives, GWDA contains a relatively significant number of them in the word list, which might have been selected according to the principle of the frequency of use (see Table 3).

A similar gradation among the dictionaries in question regarding the number of lemmas recorded, as previously diagnosed for complex words and illustrated with examples in Table 1 (the highest number of entries in DAWB and the lowest in GWDA), could also apply to multiword expressions. Their selection criteria are explicitly mentioned only in DAWB; according to the authors, the selection is largely confined to idiomatic phrases (p. 21), and only those of them that are highly frequent are taken into account (p. 22).

Three types of phrases are mainly represented: 1) idiomatic phrases from Latin, Greek, French, etc. which are often used as quotations, 2) phraseological terms, and 3) multipart names (see DAWB: 21). It is said in another place (p. 278) that phrases (in particular, multipart names and phrases of Latin origin used as quotations) are only included as examples. Multipart expressions and multipart names are also mentioned at one point in the introductory matter of GWDA, but only in the context of the principles of alphabetic arrangement of entries. Multiword expressions are not mentioned in any way in the two DUDEN editions, which, however, does not mean that they are in principle absent from the dictionaries (see Table 4).

Lexemes of foreign origin, including a number of names, are of particular importance in a pronunciation dictionary.³ There are overlaps with the two vocabulary classes already mentioned: complex words and above all, as can be concluded from Table 4, multiword expressions. An explicit explanation of the selection criteria can be found also in this case only in DAWB, where at first a distinction is made between ‘older’ and ‘younger borrowings’ (a division that obviously corresponds to the traditional linguistic dichotomy of loanword and foreign word). As in the case of native simplicia, the authors sought ‘relative completeness’ (p. 21) when selecting (simple) older borrowings that are largely assimilated and whose phoneme-grapheme relations therefore correspond to the German ones.

On the other hand, younger borrowings — countless weakly assimilated names and terms which have been taken up to a great extent in the recent past, and for which there is usually no Germanized term established by use (see DAWB: 20) — represent a much bigger challenge for an orthoepic reference book:

‘Sie fallen häufig schon in der Schreibung durch fremde Buchstaben, fremde Buchsta-benkombinationen oder Buchstaben/Buchstabenkombinationen in ungewohnten Positionen auf. Ihre Phonem-Graphem-Beziehungen werden den deutschen Regeln zwar ebenfalls angepasst […], die grafische Eindeutschung geht jedoch nicht so weit wie bei den älteren Entlehnungen. Außerdem treten hier deutlich mehr fremde Phoneme und deren Allophone auf und die fremden Silbifizierungsregeln bleiben mehr oder weniger modifiziert in Kraft.’ (p. 20)

Another difficulty lies in high dynamics of borrowing and unstable use of newer foreign word material:

‘Sehr viele der hier anzuführenden Wörter und Namen sind nur vorübergehend im Gebrauch. Meist unvorhersehbar gewinnen sie an Aktualität, werden dann in der Sprech- und Schreib-kommunikation kurzzeitig mit hoher Frequenz realisiert und verlieren schnell wieder an Gebrauchshäufigkeit. Ein Wörterbuch kann einem solch schnellen Wechsel nicht folgen – es muss sich auf andere Weise dem Problem stellen. Es kann das für die Herkunftssprache geltende Prinzip der Eindeutschung am Verhältnis von fremder und eingedeutschter Aussprache behandeln und dieses Prinzip an häufig gebrauchten Übernahmen aus der betreffenden Sprache demonstrieren.’ (ibid.)

In fact, the method described above is used in all four dictionaries: they all include quite informative sections discussing this problem in the introductory matter entitled ‘Zur Aussprache von Wörtern aus fremden Sprachen’ (GWDA), ‘Zur Aussprache fremder Sprachen’ (DU2000 and DU2015), and ‘Eindeutschung von Namen und Wörtern aus anderen Sprachen’ (DAWB). Whether the percentage of recent borrowings in the entire lists of entries in the four dictionaries is comparable, could be reliably ascertained only through a detailed count. Of course, there are differences between them that obviously result from the different age of each dictionary (see compilation in Table 5).

None of the dictionaries contain some most recent borrowings that have already become established in German and are also used in everyday language, such as E-Paper, Geoblocking, Smartwatch, Speed-Dating, Wearable(s), and others.

Nowadays, high dynamics of borrowing foreign words into German, especially English terms from the field of new technologies, means that ‘closed’ dictionaries — whether printed or digitized — will be able to keep up with the development only to a limited extent.⁴ Therefore, a transition to ‘open’ online versions of pronunciation dictionaries seems to be just a matter of time (see also Section 4).

3. Selection and treatment of inflected forms

Modern German is still an inflected language, and compared to English it has many more different morphological word forms. Hence, in the process of compiling a pronunciation dictionary, the question repeatedly arises which of the inflected forms should be included and in which manner.⁵

For the inflected word classes, naturally the base (citation) form is always recorded at first. By this is meant: the nominative singular of nouns (except for pluralia tantum), the infinitive present active of verbs, and the bare, endingless form of adjectives and adverbs. However, the words in question often have inflected forms whose pronunciation cannot be easily deduced from the respective base form, or even suppletive forms constructed from completely different morphemes as the base form. Special solutions are required to record such divergent inflected forms in a pronunciation dictionary.

The first field where the problem arises is the plural form of the noun. As an overview of the dictionaries shows, all of them contain those plural forms that exhibit accent shift in relation to the singular form (Traktor – Traktoren), as well as foreign plurals. GWDA, DU2000, and DU2015 seem to record all plurals with umlauts — including, of course, those in which the umlaut, like in Koch — Köche, additionally causes the [x]-[ç] alternation (due to the fact that in German, the grapheme sequence <ch> is basically pronounced as a velar fricative [x] after back vowels like [o:], and as a palatal fricative [ç] after front vowels like [ø:]). Also fully listed are plurals without umlauts but with alternation or voicing of the stem-final consonant after adding an ending (Zeisig – Zeisige, Sieb – Siebe, Kind – Kinder). In contrast, DAWB proves quite inconsistent in these cases: While such forms as Äpfel, Wände, Brüder, or Mütter have been included, other forms, for example Öfen, Nächte, Töchter, or Väter, are missing. Furthermore, Siebe and Berge are present, but Diebe and Tage are not, and the plural of the Könige-Pfennige-Zeisige type seems to be completely absent. Yet it is not always possible to find related forms in the vicinity of a particular singular entry (as e.g. in the case of nächtelang or Töchterchen), from which the pronunciation of the respective plural forms could be reliably concluded.

Only in GWDA many native plural forms ending in -(e)n have been recorded, and not only those in which an assimilation of [n] takes place at the stem-final plosive due to schwa omission (Ecke – Ecken, Rübe – Rüben), but also some relatively unproblematic cases where [n] just becomes syllabic (Wade – Waden).

From a technical point of view, every dictionary deals differently with the plural specification. In GWDA, the plural form, preceded by the abbreviation Pl., is written in the alphabetic script directly next to the respective entry in singular and is emphasized in bold, like the entry in singular. This is followed by both singular and plural transcriptions. If there are two different pronunciation variants of the plural, the first transcription is usually fully specified and the second one occurs reduced to the deviating sequence, typically the ending. In DU2000 and DU2015, the plural forms are also specified within the corresponding entries in singular. However, the main entry in singular appears first followed by its transcription, and then the plural form, both in alphabetical spelling and in transcription, is provided. In the case of very short words, the full plural is given. For longer forms, only graphemic and phonetic sequences differing from the singular are typically recorded.

In cases when the plural form should be alphabetically placed in the dictionary at a different position than the base form, DU2000, in contrast both to GWDA and its successor DU2015, often uses additional references (e.g. ‘Häuser vgl. Haus’; ‘Modi vgl. Modus’; ‘Visa vgl. Visum’, etc.).

In DU2015, the references of the previous edition seem to have been largely removed. To take up the examples just mentioned: the lemmas Häuser and Visa are no longer present in DU2015, the lemma Modi is still there — but no longer as a reference, but rather as a separate entry with its own transcription. The motivation for this inconsistent approach cannot be easily deduced.

In DAWB, however, all plural forms represent completely independent entries, without any references to the singular. As a result, some plurals need to be searched at a different position than the base form. An overview can be found in Table 6. Because of the equal treatment of the cited plural forms in the two DUDEN dictionaries, the examples given here are presented only once.⁶

In addition, the two DUDEN dictionaries indicate the pronunciation of strong masculine and neuter genitive forms — however, only in the cases where the inflectional morpheme coding genitive affects the voicing of the stem-final consonant (though by far not consistently), as well as for a number of foreign words. Thus, in both DU2000 and DU2015, the appropriate indication is given in the case of Betrag and Betrug, but not in the case of Betrieb or Bewerb (although it is, in turn, present in Erwerb); it can be found in the entries Bund, Kind, and Land, but not in Hund, Wind, or Geld. It is rather unlikely that a uniform selection principle was pursued here.

Another part of speech in which the inflection often has a strong sound altering effect is the verb with its conjugation endings, and especially the irregular paradigms that imply many sound alternations. In the latter case, GWDA indicates the respective infinitive followed by two further base forms in bold. The result is quasi-tripartite lemmas including three transcriptions (geben, gab, gegeben; sein, war, gewesen etc.). By contrast, the indicative forms of the second and third persons singular Präsens and Präteritum appear as independent, alphabetically ordered entries — but only for a number of irregular verbs, probably chosen according to the criterion of difficulty, or depending on whether or not they phonetically have a change in the stem in relation to the base form. Thus, one can find forms in the dictionary such as gib[s]t and gab[s]t (of geben), träg[s]t and trug[s]t (of tragen), nimm[s]t (but not nahm[s]t) of nehmen, läuf[s]t (but not lief[s]t) of laufen; occasionally relatively unproblematic cases have also been included (e.g. sprach[s]t of sprechen, traf[s]t of treffen), while some relevant forms, such as lässt (of lassen) are missing. In particular, the conjugation paradigms of modal verbs, rich in diverse variations, are incomplete. In consequence, users will fail to find not only the past participles (‘Partizip II’) of all modal verbs, but also a number of conjugated forms, such as kann (of können) or will (of wollen). Although darf[st] and mag[st] (of dürfen and mögen, respectively) are present, for alphabetic reasons they are not recorded in the same position as their infinitive forms.

A major deficiency of GWDA is that it does not list any conjunctive forms, not even the most popular ones, such as möchte (of mögen; being extremely frequent in modern German) or sei, wäre (both of sein), hätte (of haben), könnte, dürfte, müsste (of können, dürfen, müssen, respectively), käme (of kommen), gäbe (of geben), etc. Only occasionally imperatives with a change of the root vowel (e.g. gib of geben) are recorded, and in the case of regular verbs, as a general rule, only the infinitive is given.

In this respect, the other dictionaries certainly seem to be more advantageous. In both DUDEN editions and in DAWB, the three base forms of irregular verbs are always recorded as independent lemmas (thus, in the case of the verb sein with its past tense forms war and gewesen, they must be searched accordingly under S, W, and G).⁷ This also applies to the third person indicative Präsens, when a vowel change or umlaut in the stem occurs (gibt of geben, trägt of tragen), to the conjunctive forms with umlaut (äße of essen, böte of bieten, hätte of haben etc.), as well as often to the imperative with a vowel change in the stem (gib of geben, hilf of helfen). In the last case, DU2000 and (to a lesser extent) DU2015 have significantly more such forms than DAWB. Surprising is the fact that some important imperatives of this type, such as brich, lies, stich, stirb, vergiss, or wirf (of brechen, lesen, stechen, sterben, vergessen, and werfen, respectively), which are present in DU2000, are missing in its successor DU2015.

The two DUDEN editions are also more informative than DAWB in relation to the indicative personal forms of the past tense Präteritum of irregular verbs. For instance, they contain such lemmas as gab, gaben, gabst, and gabt (of geben); lag, lagen, and lagt (of liegen); trug, trugen, and trugt (of tragen), while DAWB confines the verb forms in question to the base form, such as gab, lag, trug. Occasionally, in DAWB, the pronunciation of a preterital personal form can be deduced from a homophonic entry or from an entry containing this form as a word-forming constituent, for example gaben ← Gaben, lagen ← lagenweise. This possibility, however, applies only to the conjugation forms ending in -en.

Moreover, only DU2000 consistently provides, always within the respective infinitive lemma, both imperatives and personal finite forms (third pers. sing. or second pers. pl. Präs. ind. act.) of regular and irregular verbs without vowel change, when adding the inflectional ending (or null ending at a bare stem) triggers final devoicing or alternation of the stem-final consonant, for example biegen: bieg, biegt; schreiben: schreib, schreibt; rauben: raub, raubt; sagen: sag, sagt; beruhigen: beruhig, beruhigt etc. A number of verbal forms which are of phonetical interest have been included in this way. A glance at the corresponding lemmas in DU2015 (biegen: bieg, --; schreiben: schreib, --; rauben: raub, --; sagen: sag, --; beruhigen: --) shows that it has transferred most of the imperatives, but no other forms, from its predecessor.

As in the case of the regular verb beruhigen mentioned above, which, incidentally, is very often used in German imperative sentences (especially in its reflexive form), in DU2015 the imperatives are also missing in other alphabetically adjacent verb entries, such as berücksichtigen, besänftigen, or bescheinigen. However, the respective imperative forms have been listed (or transferred from DU2000) with many other similarly built verbs, such as beschäftigen, beschleunigen, beschönigen, beseitigen, or bestätigen.

In DAWB, in contrast, such imperatives are generally not recorded, and personal forms of third pers. sing. or second pers. pl. are recorded rather sporadically for some irregular verbs. It refers to such forms as bringt (of bringen), liegt (of liegen), lügt (of lügen), trügt (of trügen), or (exceptionally) second pers. sing., bleibst (of bleiben). Incidentally, it should be said that the words lügt and trügt, in which the umlaut is not triggered by inflection, belong to very few verb forms without the root vowel change that have already been recorded in GWDA (beside the plural forms gebt of geben and grabt of graben). In both DAWB and GWDA, these forms are independent lemmas.

Finally, DU2000 as the only one of the four dictionaries, gives many, though by no means all, forms of the first pers. sing. Präs. ind. act. (identical with the singular imperative) of regular verbs ending in -eln and -ern, in the case of which the -e- belonging to the stem can often be omitted.⁸ This, usually optional, elimination manifests itself on the phonetic level in the first case by fading of the schwa [ə] (e.g. jubeln → juble, segeln → segle, tadeln → tadle) and in the second case by consonantalization of the vocalic r, when comparing to the infinitive, or also by schwa elision, when comparing to the respective unreduced form of the first person (ändern → ändere → ändre, schleudern → schleudere → schleudre, wandern → wandere → wandre etc.).

The sound-modifying influence of inflection is also observable in adjectives. In several cases, comparative and superlative forms get an umlaut in the stem (see Table 7) or are occasionally represented by suppletive forms (this applies to some adverbs as well, though to a lesser extent). Moreover, the declension of many adjectives results in the voicing or alternation of the stem final consonant after adding a suffix, and other declension-related changes may occasionally occur in the final position of base forms (see Table 8).

The irregular comparatives, including those with umlaut or suppletive forms, are always listed in GWDA and DAWB as independent lemmas arranged alphabetically. However, in DU2000 and DU2015, only the suppletive forms are treated in this way, while the forms with umlaut are recorded immediately next to the respective positive form. In DU2000 (though no longer in DU2015), there is usually an additional reference keyword in the appropriate alphabetical position. Within the forms with umlaut, the comparatives can generally be found in all four dictionaries (however, in both GWDA and DAWB such forms as kränker, kürzer, and schwächer are missing, in GWDA also forms like härter, kälter, schärfer, and stärker, and in DAWB — gesünder and klüger). The superlative forms are given for these adjectives only in individual cases: älteste — possibly because it is the first adjective of this type in alphabetical order (notably, in the two DUDEN editions and in DAWB the substantivized form Älteste, written with a capital letter, is given), größte — certainly because of the reduced superlative ending, höchst and nächst — because of the [x]-[ç] or zero-[ç] consonant change, as well as jüngst and längst, because they are also lexicalized in the language as semantically autonomous adverbs of time (meaning ‘recently’ and ‘long ago’ respectively; see Table 7). The most common and frequent suppletive forms, such as besser, best[e], mehr, meist[ens], minder, mindest[ens], have been recorded in all dictionaries analysed.

The change of the stem final consonant triggered by inflection (whether by mere voicing, or by [ç]-[g] alternation) is appropriately illustrated both in GWDA and in the two DUDEN editions, by using the respective declension form ending in -e (fies – fiese, trüb – trübe; ruhig – ruhige, etc.), which then appears next to the base word. This information is, however, missing in DAWB (see Table 8).

Another feature which is reflected neither in GWDA nor in DAWB, and is only partially reflected in the two DUDEN editions, is the elimination of -e triggered by declension in the case of the (relatively rare) adjectives ending in -el and -er.

In the first group (akzeptabel, dunkel, heikel, komfortabel etc.) it manifests on the articulatory level — similar to the verbs discussed above — by fading of the schwa-[ə]. In the second group (makaber, sauer, teuer etc.) it is reflected by consonantalization of the vocalic /r/ when comparing to the base form of the adjective, or also by schwa elision when comparing to the hypothetical unreduced inflected form, for example eine dunkle Nacht, ein makabres Verbrechen. Both DU2000 and DU2015 indicate only the schwa elimination in adjectives of foreign origin, for example in akzeptabel or komfortabel, but they fail to do so with the native German adjectives, such as dunkel or heikel. The indication is performed — as in the case of the consonant alternations mentioned above — by specifying the respective declension suffix -e (akzeptable, komfortable; see also Table 8).

Finally, with regard to other inflected word classes, it should be mentioned that declined pronoun and article forms are recorded in all four dictionaries as separate entries sorted alphabetically. In some cases, this results in individual terms of an inflectional paradigm (e.g. ich – mir; wir – uns; das – dem, etc.) being positioned in very different places of the dictionary.

To summarize the above, the principles and methods of selecting and recording inflectional forms in the pronunciation dictionaries in question give a highly complicated, sometimes even confusing picture. When it comes to positioning the forms, however, one of the works examined, namely DAWB, proves to be entirely consistent: all the inflected forms represent separate entries and are arranged accordingly in its word list.

This solution is without doubt clear and user-friendly. On the other hand, one may assume that such an approach requires significantly more space in a printed dictionary (which the above-average size of DAWB seems to confirm), and therefore can be problematic in some cases due to printing cost (cf. Atkins and Rundell 2008: 27, Lew 2011; see also Section 4). From the user’s perspective, there is also the fact that a physically much larger reference book is generally less handy to use.⁹

Thus, another conceivable method, which was not applied in any of the analysed dictionaries, might be to specify all phonetically relevant forms of a lexical word consistently under its base form, that is within one entry. A user who wants to find out more about the pronunciation of a word in an inflected form usually knows its lexical meaning and has a basic grammatical knowledge allowing him to look up words like äße, ging, größer, hätte, Häuser, klüger, Söhne, stirbt, or trägst under essen, gehen, groß, haben, Haus, klug, Sohn, sterben, or tragen, respectively. Moreover, additional reference keywords (such as those already used in DU2000) might be an acceptable auxiliary solution in the case of the altogether relatively rare suppletive forms which — due to morphemic differences — have to be recorded, considering the alphabetical order, in a completely different place than the base word.

Such a unified approach could help both the authors and the user to keep a general overview. For the former, it might facilitate optimizing the selection and coordination of the inflectional forms to be included in the process of compiling a dictionary, whereas the latter would be no longer forced to reach as often for two or more dictionaries because of the number of differences.

4. Sources of vocabulary used and current developments in this regard

Different sources are used in order to obtain the lemmas, that is various words and word forms, but in many cases also multiword expressions of which a dictionary’s list of entries is composed:

In the phase of the material collection the lexicographical work is based on the experience and knowledge of the staff members and on general search routines, which are used to collect data from the internet, from libraries or from existing dictionaries. These processes will help to provide the sources for building up the dictionary basis, which is distinguished into the primary sources (data which are used for the purpose of the dictionary making, e.g. citation collections, special text corpora), the secondary sources (other dictionaries), and the tertiary sources (all other linguistic material concerning the dictionary subject matter) […]. Schierholz 2015: 330-331

Among the pronunciation dictionaries under analysis, only GWDA, the oldest, fails to provide any information on the vocabulary inventories used. By contrast, the authors of the remaining three titles mention, above all, the most common universal dictionaries of German, as well as the DUDEN spelling dictionary and the DUDEN dictionary of foreign words in their most up-to-date editions.

Apart from these, other sources named in DU2000 are general reference works, dictionaries for special purposes, pronunciation dictionaries, and bilingual dictionaries, as well as radio and television broadcasts. As far as proper names are concerned, the additional sources include town indices, atlases, maps, timetables, biographical dictionaries, and such like (p. 18). All this data also applies to DU2015, since it is based on the word list of previous edition(s) (see DU2015: 14), which has been extended, updated, and modified accordingly in the process of creating the newest edition. However, as some of the examples presented in this paper — especially in Section Two — show, certain lemmas or lemma parts that were recorded in DU2000 have been removed in DU2015.

It cannot be determined without appropriate insider information on the book production process whether or not the saving of space in the print edition was the only motive for the deletions. However, the explanations in Agnes (1995: 47) suggest that technical and, in particular, financial factors often play a key role in the process of revision of large dictio-naries (see also Tarp 2008: 176). Given a certain requirement that has to be met, for example the pagination must not be changed (which, however, obviously does not apply to DU2015), a new lemma can be recorded on the same page only at the cost of other entries or their components (for a broader discussion on space restrictions in paper and electronic dictionaries see Atkins and Rundell 2008, Lew 2011, Lew and De Schryver 2014).

Compared with both DUDEN editions, the authors of DAWB list a number of specific works which served as additional vocabulary sources. In particular, the list includes various dictionaries: an etymological, a phraseological, and a neologism dictionary, but also atypi-cally structured reference works, such as an onomasiological and a retrograde dictionary. In addition, there are four different dictionaries of German as a foreign language and one universal dictionary (see DAWB: 17–18; for detailed bibliographic information, see 223–228).

Besides, two of the pronunciation dictionaries in question, namely DAWB and DU2015 as the most recent ones, also refer to digitized vocabulary sources, in addition to traditional paper-based ones. Thus, in DAWB, the online portal Wortschatz Universität Leipzig (http:// www.wortschatz.uni-leipzig.de), is listed even in the first place, that is before all titles already mentioned above. Furthermore, the LEO-Internetwörterbuch Deutsch – Englisch as well as different search engines and sources like Google and Wikipedia are also included (p. 18).

In DU2015, we learn that its total list of lemmas was frequency-prioritized and compared with DeReWo (2013), that is with a word list based on the German Reference Corpus (DeReKo) (cf. 14). The latter, which was created at the IDS, Institut für Deutsche Sprache,¹⁰ and is still being developed, is the largest linguistically motivated collection of German corpora, serving as an empirical basis for linguistic research. Under the project name DeReWo, frequency-based ranking lists of words are created on the basis of virtual corpora, in the main of the DeReKo, including lemma candidate lists for dictionaries (see http://www1.ids-mannheim.de/kl/projekte/methoden/derewo.html).

It is quite obvious that all these steps mark the beginning of a new development, during which the focus in the process of selecting vocabulary for pronunciation dictionaries is increasingly shifting to digital (online) sources. This transition provides a chance to make better use of the benefits offered by corpus linguistic methods for lemma selection. Čermák (2013: 1356) lists the advantages of using corpora:

Corpus-based dictionaries heavily draw on corpora, among other things, because of three primary advantages, whose character is either entirely or almost entirely new now, namely: (1) Evidence: sufficient quality and quantity of data, alowing us to cover their frequencies; (2) Context: sufficient size and sufficiently reflecting all relevant aspects of meaning; (3) Usage: sufficient coverage of relevant types of context as well as situations and functions underlying these contexts. (italics in original)

Whereas the last two aspects seem to concern primarily general dictionaries, the first one will undoubtedly apply to pronunciation dictionaries as well.

A number of linguists have gone one step further, suggesting to use the text content of the World Wide Web as a corpus (see, e.g., De Schryver 2002, Kilgarriff and Grefenstette 2003, Resnik and Smith 2003, Lew 2009). Kilgarriff and Grefenstette (2003: 333) explain this development as follows: ‘Language scientists and technologists are increasingly turning to [ the web ] as a source of language data, because it is so big, because it is the only available source for the type of language they are interested in, or simply because it is free and instantly available.’

On the other hand, using the web as a corpus for lexicographic purposes also involves certain problems, such as possibly dubious linguistic representativeness, poor balancing, and noisiness of corpora. When it comes to language errors, Kilgarriff and Grefenstette (2003: 342) quite rightly argue: ‘Web texts are produced by a wide variety of authors. Contrary to paper-based, copy-edited published texts, web-based texts may be produced cheaply and rapidly with little concern for correctness. […] The web is a dirty corpus, but expected usage is much more frequent than what might considered as noise.’ Lew (2009: 298), in turn, completes the discussion as follows: ‘[I]t can be concluded that the WWW, despite its noisiness and poor balancing, can be an attractive and useful tool for on-line language reference. Its main virtues lie in the impressive size of the resource, and the speed with which it can be trawled using a general-access search engine.’

It seems that a full transition to electronic (and increasingly also web-based) corpora is inevitable and might be much faster than one would have expected just a few years ago — if one only considers the dynamics with which the latest technologies are changing the world around us in all its aspects. However, of even greater importance is the fact that digitization also implies the transition from printed to electronic dictionaries (for some general issues surrounding the e-dictionary making process, the related user needs, user feedback, changes, and deletion of entries see, e.g., Klosa 2013 and Gouws 2016). As far as electronic pronunciation dictionaries are concerned, Sobkowiak (2005, 2007a) lists a number of their benefits for users, such as phonetic access, audio sound representation, possibility of customizing the display and phonetic content to the needs of the user, to name but a few. On the other hand, occasional mismatch between transcription and recorded audio, list-reading effects, recordings limited to the headwords only, deficient typographic design, and some others appear specifically as potential weaknesses of the respective dictionaries (see ibid.; for innovative ideas of designing phonetic interfaces in electronic dictionaries, see Sobkowiak 2007b).

The digital version of DU2015, being the first and only electronic pronunciation dictio-nary of standard German to date,¹¹ was the subject of a discussion in the first part of our analysis (Nycz and Tęcza 2019, in print).

5. Previous and forthcoming work

After having analysed the lexicographic construction and the multimedia components of four German pronunciation dictionaries in Nycz and Tęcza (2019, in print), and their principles of selection and lemmatization of lexical material in the current paper, we intend to take a closer look at their partially divergent understandings of phonetic standard in our next contribution (Nycz and Tęcza 2020, under review). The forthcoming article shall also examine the different ways of looking at some crucial phonetic phenomena in the dictio-naries concerned. Afterwards, the overall comparison will be summarized and final conclusions will be formulated.

6. Acknowledgements

We are indebted to Prof. Ursula Hirschfeld from The Martin Luther University of Halle-Wittenberg for kindly having made available to us the two following papers: Hirschfeld and Stock (2006a, b).

Footnotes

References