Parallel corpus in analysing Czech spoken expressions and their equivalents in English, French, and Polish

Parallel corpus in analysing Czech spoken expressions and their equivalents in English, French, and Polish Adrian Jan Zasina (Charles University, Prague) ABSTRACT This paper uses corpus data to analyse spoken expressions and discourse markers in Czech, applying these findings to corpus-based exercises for learners of Czech as aforeign language. The analytical section highlights the usefulness of parallel corpus in identifying suitable translation equivalents for prevalent Czech spoken vocabulary in English, French, and Polish as native languages from the learner’s perspective. The methodology outlines the process of finding appropriate translation equivalents in film subtitles, considering both meaning and spoken register. The pedagogical section introduces three corpus-based exercises designed to improve conversational skills, featuring authentic texts that familiarise learners with spoken vocabulary. This research builds on previous studies of the English language that did not use parallel corpora to identify translation equivalents in learners’ native languages— an essential factor for understanding aforeign language. In addition, tailor-made corpus-based exercises can be seamlessly integrated into everyday classroom activities to enhance language awareness among non-native speakers. KEYWORDS corpus, corpus-based exercises, Czech, data-driven learning, discourse markers, speaking skills, spoken expressions DOI https://doi.org/10.14712/18059635.2024.2.2 1 INTRODUCTION The process of acquiring aforeign language requires substantial exposure to natural language to yield discernible results. Therefore, employing authentic linguistic material is crucial for facilitating the development of alearner’s language skills, particularly for advanced learners aiming to attain fluency akin to that of native speakers, especially in conversational contexts. However, dialogues found in coursebooks often diverge significantly from real-life language use (Gilmore, 2004). Despite attempts to incorporate features typical of spoken language registers, these dialogues are artificial constructs created by coursebook authors (Holá, 2019). Consequently, they lack the authenticity inherent in everyday interactions, which include response tokens, natural repetitions, false starts, hesitations, etc. (Walsh, 2010). McCarthy and Carter (2001, p. 338) aptly emphasize the importance of appropriate input to achieve desirable language output: “Whatever else may be the result of imaginative methodologies for eliciting spoken language in the L2 classroom, there can be little hope for natural spoken output on the part of language learners if the OPEN ACCESS ADRIAN JAN ZASINA 101 input is stubbornly rooted in models that owe their origin and shape to the written language.” Fortunately, today we have access to language corpora that provide examples of authentic language usage. Corpora have been extensively employed in language teaching for over 30 years. The approach of using corpus methods in learning was popularized by Johns (Johns, 1991) in the early 1990s. He coined the term data-driven learning (DDL) in his seminal work, describing this innovative method. According to Johns, students can discover language rules based on corpus evidence, thereby becoming independent in their learning. In recent years, numerous studies have employed the DDL method; however, the majority of these do not focus on developing speaking skills (Boulton & Cobb, 2017, p.379). They primarily concentrate on writing, vocabulary, lexicogrammar, and grammar. Nevertheless, studies investigating typical spoken vocabulary, such as linking adverbials (Boulton, 2009), epistemic stance markers (Şahin Kızıl & Savran, 2018) or the discourse marker well (Huang, 2019), have emerged. Unfortunately, in the context of teaching Czech as aforeign language, such corpus studies are currently unavailable. The sociolinguistic situation of Czech is sometimes considered as near-diglossic (Bermel, 2014), characterized by two distinct varieties: Literary Czech (spisovná čeština) and Common Czech (obecná čeština). Literary Czech represents the prestigious variant used in the majority of situations, while Common Czech is employed in everyday informal spoken communication. There is ongoing debate regarding the incorporation of Common Czech into foreign language teaching, with scholars holding differing perspectives on this matter. Some advocate for the introduction of spoken language elements even at the basic level (Holá, 2019, pp.115–116), while others suggest its integration at an advanced level of communication competence (Hrdlička, 2019, p.101). Undoubtedly, advanced learners should recognize the marked differences between spoken and written styles, especially when striving for amore “native” sounding expression. Nonetheless, working with authentic language samples is crucial for advanced learners to attain appropriate fluency in spoken interactions with native speakers. Hence, this study aims to analyse spoken expressions that can later be used in teaching speaking skills in Czech as aforeign language. The primary goal is to conduct an analysis based on parallel corpus data to identify the most appropriate translation equivalents of frequently used Czech spoken vocabulary in English, French, and Polish, considering the learner’s perspective as native speakers of these languages. To effectively apply the features of spoken language, learners need to first understand their meaning in their mother tongue. Barlow (2000, p.110) notes that “learning asecond language involves some use of first language schemas as templates for creation of schemas for the second language, as well as the formation of completely new schemas for the second language.” Moreover, it is crucial to make learners aware of how to work with corpus data to obtain relevant translation equivalents. For this purpose, asubtitle component of aparallel corpus, emulating aspoken register, is used. The secondary goal is to apply the vocabulary under examination to corpus-based exercises. These exercises aim to practice and reinforce spoken vocabulary while producing dialogues that enhance conversational skills. They consist of authentic lanOPEN ACCESS 102 LINGUISTICA PRAGENSIA 2/2024 guage examples derived from parallel, written, and spoken corpora, intended for use in classroom activities as hands-on exercises. The paper is further divided into four sections. Section 2 presents data and methodology, describing the process of choosing the lexemes under examination. Section 3, devoted to corpus analysis, is the core part of the paper. Firstly, it comments on the semantics of spoken expressions and discourse markers, and the importance of using parallel corpus data in this study. Secondly, it provides an in-depth analysis of the subtitle corpus with regard to translation equivalents in English, French, and Polish, and underscores the importance of understanding the Czech spoken register through their equivalents derived from aparallel corpus. Thirdly, it summarizes the most common tendencies in translation strategies and offers general comments on working with quasi-spoken parallel data. Section 4 delivers ideas for corpus-based classroom activities that could improve the conversational skills of advanced Czech learners. It also highlights the usefulness of the DDL method in language teaching and learning. Section 5 discusses and concludes the main findings. 2 DATA AND METHODOLOGY First, it was necessary to find candidates for spoken words for further analysis of typical spoken vocabulary in the parallel corpus. To compile an appropriate list of spoken vocabulary, Iused frequency lists of the top 3,000 written and spoken Czech lemmas set by Škrabal, Laubeová and Štěpánková (2022, pp.32–92). The list is areadyto-use set based on written and spoken corpora. The lemmas in this list were marked with the letters W(for written) and S(for spoken), followed by anumber 1–3, indicating the first, second, and third thousand. Iexamined awordlist with the code S1— the most frequent lemmas marked as spoken within the first thousand— which consists of 180 lexemes. To reveal asuitable vocabulary for the learners, further categorisations were needed (see 2.2 Method). For the purpose of analysis, aparallel corpus was used; for preparing corpus-based exercises, corpora of contemporary spoken and written Czech were used (see 2.1 Data). 2.1 DATA This study makes use of data from the InterCorp release 12 parallel corpus (Čermák & Rosen, 2012) to establish alist of translation equivalents for typical spoken vocabulary. InterCorp consists of original texts and their translations in two main components, called the core and the collection. The core primarily contains fiction and is semi-manually aligned1, whereas the collection is automatically annotated and is further divided into six components: 1 First, the core undergoes automatic alignment using the hunalign software (Varga et al., 2007; http://mokk.bme.hu/en/resources/hunalign/). Second, the automatically aligned texts are uploaded to the parallel editor InterText (Vondřička, 2014; https://wanthalf.saga. cz/intertext) where they are checked semi-manually. OPEN ACCESS ADRIAN JAN ZASINA 103 — legal texts of the European Union from Acquis Communautaire corpus (https:// ec.europa.eu/jrc/en/language-technologies/jrc-acquis), — translation of the Bible, — political commentaries from Project Syndicate (https://www.project-syndicate.org/), — political commentaries from VoxEurop (https://voxeurop.eu/), — proceedings of the EU parliament from 2007–2011 (http://www.statmt.org/ europarl/), — film subtitles provided by the Open Subtitles database (https://www.opensubtitles.org/). For the purpose of my study, Iwork with the subtitle component and its Czech (Rosen, Vavřín, & Zasina, 2019a), English (Klégr et al., 2019), French (Nádvorníková & Vavřín, 2019), and Polish (Bańczyk, Dybalska, & Vavřín, 2019) parts. The film subtitles were chosen because of their similarities to the spoken register. It is not feasible to obtain parallel texts of spoken language; however, subtitles can effectively emulate natural and spontaneous spoken communication. Therefore, in terms of parallel data, they are the most suitable material for analysing the most frequent spoken vocabulary across languages. Moreover, this approach makes it possible to identify translation equivalents in the languages examined. Of course, Iam aware that this is not real spoken language (in this context, it is rather quasi-spoken language), so it is necessary to engage spoken corpora in the preparation of corpus exercises. In terms of corpus-based exercises, the ORTOFON corpus (Kopřivová, Komrsková, Lukeš, Poukarová, & Škarpová, 2017) and SYN2015 corpus (Křen et al., 2015) were used. ORTOFON is aCzech spoken corpus of spontaneous (informal) everyday communication that is lemmatized, morphologically tagged, and balanced with respect to several sociolinguistic categories (Komrsková, Kopřivová, Lukeš, Poukarová, & Goláňová, 2017). It contains recordings of spontaneous conversations between people who have familiar or friendly relations; none of the conversations is experimentally induced. On the other hand, SYN2015 is arepresentative corpus of contemporary written Czech, consisting of lemmatisation, morphological, and syntactic annotation (Křen et al., 2016). It is balanced in terms of various text genres, including fiction, non-fiction, newspapers and magazines. Predominantly, it consists of texts published in the last five-year period (2010–2014). Both corpora are appropriate materials for providing suitable examples of spoken and written language to highlight their differences. Thus, this material is used in the corpus-based exercises presented in Section 4. 2.2 METHOD The current study concentrates on spoken expressions intended for teaching Czech learners. Hence, it was imperative to select appropriate words reflecting the natural speech of native speakers. Iopted to pinpoint typical spoken expressions2 and 2 By spoken expressions, Imean lexemes that are frequent in the spoken register and have written counterparts with which they form pairs, such as: cop— policeman, bro— brother, granny— grandmother. OPEN ACCESS 104 LINGUISTICA PRAGENSIA 2/2024 discourse markers (DM)3 from awordlist comprising 180 lexemes obtained from Škrabal, Laubeová and Štěpánková’s frequency list (2022, pp.32–92). Several identified categories do not meet the established criteria: — exclamations (e.g., ježišmarjá ‘Jesus Christ’, ježíš ‘geez’, jé ‘oh’, ahoj ‘hello’, fuj ‘yuck’), — swear words (e.g., prdel ‘ass’, hovno ‘shit’, kráva ‘bitch’, píča ‘pussy’, debil ‘asshole’), — deixis (e.g., tenhleten ‘this’, tadyhle ‘here’, tamhle ‘there’, nikam ‘nowhere’), — hesitation (hmm, aha, emm), — dialectal words (e. g. aj ‘and’, tož ‘well’), — words with athematic focus (e.g., nouns: záchod ‘toilet’, angličtina ‘English language’, chata ‘cottage’, doktor ‘doctor’; verbs: sníst ‘to eat’, dojet ‘to reach’, vymyslet ‘to think up’, lézt ‘to climb’; adjectives: strašný ‘terrible’, zvědavý ‘curious’, šílený ‘crazy’, napsaný ‘written’), — other frequent parts of speech (e.g., numerals: šedesát ‘sixty’, osmdesát, ‘eighty’, devadesát ‘ninety’ adverbs: akorát ‘only, just’, trošku ‘alittle’, super ‘great’, hezky ‘nice’; conjunctions: čili ‘so’, poněvadž ‘because’). Two more categories were identified: spoken expressions (Table 1) and DM (Table 2), which are used in the analysis presented in the following paragraph. Lexemes in the tables are ranked based on their frequency in spoken corpora. The tables also include translations into English, ipm (instances per million) in spoken corpora, ipm in the written corpus SYN2015, and the spoken/written (S/W) ratio of ipm. When comparing both ipm values, it is noticeable that these words are predominantly used in spoken language. In terms of DM in Table 2, it must be highlighted that lexemes such as myslet, normálně, vyloženě, kdyžtak, schválně, víceméně have grammatical and semantic functions. However, they can also be considered DM in certain contexts, e.g., Vypadá úplně normálně (She looks totally normal) vs. Normálně žárlíš (You’re simply being jealous), as com mented on in the analytical part. It was not possible to differentiate between these two functions automatically; therefore, the numbers presented in the table are shown for the entire lemma, and the English translations correspond to the basic meaning. In the analytical part, the Treq tool (Škrabal & Vavřín, 2017; Vavřín & Rosen, 2015) was employed to generate alist of the most frequent translation equivalents for Czech spoken expressions (Tables 4, 6, 7) and DM (Table 8) under examination. Amaximum of three of the most frequent examples were chosen, except in one in3 DM are defined as “words or phrases that appear to have no grammatical or semantic function, such as you know, like, oh, well, Imean, actually, basically, OK as well as connectives like because, so, and, but and or” (Baker & Ellece, 2011, p.34). Schiffrin (1987, p.31) defines DM in her seminal work as “sequentially dependent elements that bracket units of talk”. In this study, DM are understood as words with no grammatical or semantic function but with apragmatic function; they are typical of spoken communication and are an important factor in maintaining aconversation. DM are involved in spontaneous interactions between interlocutors. OPEN ACCESS ADRIAN JAN ZASINA 105 No. Lemma Translation Frequency ipm SYN2015 ipm S/Wratio 1 furt still 6948 914.42 8.86 103.21 2 barák house 1998 262.96 18.76 14.02 3 mamka mum 1768 232.69 10.91 21.33 4 děcko kid 1214 159.77 14.15 11.29 5 taťka dad 1055 138.85 7.16 19.39 6 kafe coffee 1046 137.66 17.05 8.07 7 sranda fun 915 120.42 11.11 10.84 8 kilo kilo 809 106.47 20.79 5.12 9 brácha bro 764 100.55 14.46 6.95 10 ségra sis 681 89.63 3.06 29.29 11 babi granny 647 85.15 3.79 22.47 12 prachy money 629 82.78 12.20 6.79 13 ženská woman 563 74.10 28.73 2.58 14 chudák poor 559 73.57 19.01 3.87 15 polívka soup 417 54.88 6.37 8.62 16 policajt cop 415 54.62 15.60 3.50 17 krám shop 407 53.57 9.03 5.93 18 kámoš mate 396 52.12 11.16 4.67 19 flaška bottle 392 51.59 4.22 12.23 20 strejda uncle 386 50.80 7.91 6.42 21 borec guy 373 49.09 9.46 5.19 22 gympl high school 371 48.83 2.51 19.45 23 obývák living room 360 47.38 17.29 2.74 24 foťák camera 347 45.67 5.07 9.01 25 baterka torch 340 44.75 11.20 4.00 26 mail email 338 44.48 9.72 4.58 Table 1.Spoken expressions No. Lemma Translation Frequency ipm SYN2015 ipm S/Wratio 1 myslet to think 16030 2109.71 680.07 3.10 2 viď right 7353 967.73 14.47 66.88 3 hele hey, look 6694 881.00 21.91 40.21 4 normálně normally 5841 768.73 27.71 27.74 5 jakože like 3310 435.63 2.08 209.44 6 jó yeah 1453 191.23 2.72 70.30 7 cože what 977 128.58 20.46 1285.83 8 hej hey 956 125.82 8.49 14.82 9 vyloženě strictly, downright 710 93.44 10.43 8.96 10 kdyžtak just 679 89.36 0.31 288.27 11 holt just, well 463 60.94 6.18 9.86 12 schválně on purpose 349 45.94 10.64 4.32 13 víceméně more or less 344 45.27 19.73 2.29 Table 2.Discourse markers OPEN ACCESS 106 LINGUISTICA PRAGENSIA 2/2024 stance where four were selected. Translation equivalents were arranged based on absolute frequency for each language separately, and particular equivalents on the same line may not necessarily correspond to each other. Example sentences from the InterCorp corpus were provided for the chosen lexemes to illustrate the most suitable spoken equivalents in all four languages. These sentences can be utilized in aclassroom setting to demonstrate appropriate strategies used in spoken communication. In some cases, on-line dictionaries were consulted to verify whether the translation equivalents were identified as informal or spoken words. The English Cambridge Dictionary (https://dictionary.cambridge.org/), the French Le Dictionnaire (https://www. le-dictionnaire.com/), and the Polish Słownik języka polskiego PWN (https://sjp.pwn.pl/) were used for this purpose. 3 ANALYSIS This paragraph presents an analysis conducted on data from InterCorp. The objective of the examination is to identify the most suitable translation equivalents of spoken Czech expressions and DM (Tables 1 and 2) in English, French, and Polish, considering them from the perspective of the learner’s native languages. The aim is not only to find any equivalent but also to locate appropriate translation equivalents that suit the spoken register. The results should assist teachers in approaching corpus data and understanding the types of issues that must be considered when working with corpus evidence. Moreover, the obtained translation equivalents may be directly applied to classroom activities, and examples derived from the analysis could provide learners with authentic usage illustrations. This paragraph is divided into three subsections. Subsection 3.1 discusses the examined lexemes in terms of their meaning and function. Subsection 3.2 focuses on the subtitle material, explaining how to interpret corpus instances and analysing the obtained results. Subsection 3.3 recapitulates the main findings and formulates general tendencies in the investigated languages. 3.1 SPOKEN EXPRESSIONS AND DISCOURSE MARKERS Looking at the category of spoken expressions, it is possible to identify designations of family members (e.g., mamka, děcko, taťka), people (e.g., ženská, chudák, policajt), buildings (e.g., barák, krám, gympl), food and drinks (kafe, polívka), items (flaška, foťák, baterka), and other colloquial expressions (furt, sranda). This primarily highlights the thematic focus of typical conversation. Within the DM category, it is possible to identify verbs of thinking (myslet), confirmation checks (viď, cože), phatic expressions (hele, jó, hej), fillers (jakože, kdyžtak), boosters (vyloženě), downtoners (víceméně).4 They are important factors to maintain aconversation. 4 The division of DM was inspired by the lists of language futures (Cvrček, Laubeová, et al., 2020b) and the chapter on DM (Čermáková, Jílková, Komrsková, Kopřivová, & Poukarová, 2019). OPEN ACCESS ADRIAN JAN ZASINA 107 These types of expressions are not typically included in coursebooks; however, learners may frequently encounter them in interactions with native speakers.5 Foreigners might face challenges in comprehending and actively using such spoken expressions, as well as in finding suitable translations that correspond to spoken communication in their native languages (in this case, English, French, and Polish). Furthermore, printed dictionaries often fail to provide translation equivalents for all potential situations. They struggle to keep pace with evolving language use and cannot encompass equivalences that have emerged in recent years. Fortunately, modern learners now have the opportunity to acquire vocabulary through online access to corpus data. The collection of subtitles emulates spoken language that cannot typically be obtained in parallel texts. The analysis does not focus on the direction of translation but rather on the examples themselves. The direction of translations is not considered here because, in this study, capturing appropriate equivalents typical of the spoken register is more important than addressing interference from the source language. However, it is worth noting that most film subtitles are translated from English into other languages. 3.2 SUBTITLE SUBCORPUS It is important to carefully select examples from the subtitle subcorpus, as it contains alarge number of irrelevant instances that do not provide ameaningful situational context for non-native speakers. Sometimes, the concordance displays sentences that are too short or where the translation is misaligned (see Figure 1). In such cases, it is necessary to click on the sentence to expand the context. Figure 1.Parallel concordance of lexeme mamka ‘mum’ in the subtitle collection Teachers should also be aware that the translations from the Open Subtitles database are not authorized by any official film company. Therefore, it is crucial for teachers to choose examples wisely for their classroom activities. It is recommended to prepare material in advance to avoid any misunderstandings, especially when working with learners at alower proficiency level, which requires careful preparation. However, the subtitles represent avast collection of data that allows us to identify the most frequent occurrences. This frequency-based approach helps avoid unnecessary instances. Consequently, the following analysis deals with the most frequent trans5 As Bulejčíková (2015, p.76–86) concludes, Czechs tend to use quite often Common Czech in communication with foreigners (in more than aquarter of the records analysed). OPEN ACCESS 108 LINGUISTICA PRAGENSIA 2/2024 lation equivalents and outlines steps for approaching corpus data and interpreting results from the subtitle collection. To highlight the most appropriate strategies for learners engaged in spoken communication, it is essential to examine the chosen corpus instances, focusing on the most suitable spoken equivalents. 3.2.1 SPOKEN EXPRESSIONS The first group of spoken expressions under analysis comprises alist of family members: mamka, děcko, taťka, brácha, ségra, babi, strejda. It is important to be aware that each language pair (e.g., Czech-English or Czech-French) in InterCorp contains adifferent number of texts. Therefore, the number of instances varies for each lexeme depending on the language component. The table below shows the absolute frequency of these family members in the subtitle collection for the Czech-English, CzechFrench, and Czech-Polish components. Lexeme Component mamka děcko taťka brácha ségra babi strejda Czech-English 1,111 3,029 1,008 4,717 554 1,001 1,373 Czech-French 592 1,634 540 2,425 277 420 650 Czech-Polish 726 2,136 653 3,421 365 625 915 Table 3.Number of hits for family members in the subtitle collection of InterCorp No. Family member English Translation French Translation Polish Translation 1. mamka mom mère mama mother maman mamusia mum matka 2. děcko kid enfant dziecko baby bébé dzieciak child gosse 3. taťka daddy papa tata dad père tatuś papa tato 4. brácha brother frère brat bro frangin stary man pote brach 5. ségra sister sœur siostra sis sœurette siostrzyczka frangine 6. babi grandma mamie babcia granny mémé babunia 7. strejda uncle oncle wujek tonton wuj Table 4.Translation equivalents for the family members group, based on the Treq tool OPEN ACCESS ADRIAN JAN ZASINA 115 (13) cs Seš dost slavnej, viď? en Hey, you’re pretty famous, right? fr T’es célèbre, toi, hein ? pl Jesteś dość znany, co? The lexemes hele and hej are primarily used to draw attention, often at the beginning of an utterance. Hele is commonly translated as an imperative in the three examined languages: look, regarde, (po)słuchaj. In contrast, hej is more accurately translated as hey, hé, or hej, which align better with the spoken register. Example (14) demonstrates the most suitable translation equivalents for the lexeme hele. (14) cs Hele, něco ti řeknu. en Look, I’ve got some news for you. fr Regardez, j’ai des nouvelles pour vous. pl Słuchaj, mam dla ciebie wiadomość. No. Discourse markers English Translation French Translation Polish Translation 1. myslet to think penser myśleć to mean croire sądzić to guess réfléchir mieć na myśli 2. viď right hein prawda huh non co pas vrai tak 3. hele to look regarder hej hey écouter (po)słuchać to listen hé patrzeć 4. normálně normally normalement normalnie normal normal zwykle usually habitude zazwyczaj 5. jakože like comme jakby 6. jó yeah ouais tak taa(a) 7. cože what quoi co 8. hej hey hé hej yo hey hey yeah salut cześć 9. vyloženě strictly absolument szczególnie 10. kdyžtak if – wrazie czego just jakby co 11. holt just – – 12. schválně on purpose exprès celowo 13. víceméně more or less plus ou moins mniej więcej Table 8.Translation equivalents for DM, based on the Treq tool OPEN ACCESS 116 LINGUISTICA PRAGENSIA 2/2024 The lexeme normálně has three equivalents in each language, but the most common translations are normally, normalement, normalnie. These equivalents correspond closely with each other, as illustrated in Example (15). (15) cs On by pro mě normálně nehnul ani prstem. en You know, and normally he wouldn’t lift afinger to do anything for me. fr Normalement, il ne lèverait pas le petit doigt pour moi. pl Normalnie on nie kiwnąłby nawet palcem. The lexemes jakože, cože, vyloženě, schválně, and víceméně each have only one translation equivalent in English, French, and Polish. While these expressions often do not add significant content to aconversation, they are characteristic of spoken language. They provide speakers with additional time to organize their thoughts and structure their utterances. The lexeme jó is aparticularly interesting case. It translates into familiar variants such as yeah in English, and ouais in French. However, in Polish, the most frequent equivalent is the standard tak, with taa being amore familiar variant that aligns with the spoken register. The last two lexemes, kdyžtak and holt, are quite specific to Czech. While finding an equivalent for these lexemes in English (such as just) is relatively straightforward, it is more challenging to identify suitable translations in French and Polish. No direct translations were found for kdyžtak and holt in French. For Polish, however, atranslation equivalent for kdyžtak has been identified, as shown in Example (16). These observations underscore the difficulty in finding precise counterparts for certain Czech discourse markers in other languages, particularly in the case of less directly translatable terms. (16) cs Kdyžtak budu mluvit já. pl Wrazie czego, ja się tym zajmę. 3.3 SUMMARY The analysis of subtitle material has indicated afew relevant approaches to working with quasi-spoken parallel data. First, the corpus data have provided arelevant source of typical translation equivalents for spoken language. Thanks to the wide variety of subtitle material in Czech, English, French, and Polish, the most frequent equivalents reflect not only accurate translations but also asuitable register. This information is crucial for learners of Czech who want to improve their speaking skills and fluency. Second, the results of the analysis have revealed that Polish, in comparison with English and French, more often opts for familiar variants. This shows that Polish and Czech have more in common not only because of their genetic similarities but also because of their similar tendency to use familiar terms rather than official ones. Third, auser of parallel corpora should be aware that the highest frequency may not always indicate atranslation equivalent suitable for aparticular situation (in this OPEN ACCESS ADRIAN JAN ZASINA 117 case, the spoken register). Sometimes it is necessary to analyse the context to decide if agiven equivalent meets our requirements; for example, in the case of the lexeme prachy, more appropriate equivalents appeared in the second position. Fourth, in some cases, it is not possible to find any translation equivalent. There could be two reasons for this. Sometimes it is aquestion of data availability; too little data cannot provide enough translation examples to identify any equivalent. In other cases, it is amatter of language resources, which may be lacking in the target language, as in the example of the lexeme holt, for which no single reference in French or Polish was found. The analysis of subtitle collections is crucial in language learning and teaching. These results should be taken into account when considering the implementation of corpus methods in teaching conversational skills in the Czech language. It is feasible to build on the identified spoken expressions and DM to prepare suitable exercises that enrich learners’ knowledge of spoken Czech. Furthermore, the analysis of translation equivalents provides examples in the learners’ native language or awell-known second language, offering essential information for understanding the Czech spoken register. The issues covered in this research should be incorporated into activities that may help learners master typical spoken expressions. Thus, the following section focuses on the introduction of concrete corpus-based exercises. 4 CORPUS CLASSROOM EXERCISES6 This section presents ideas for corpus-based classroom activities designed to enhance conversational skills of advanced learners of Czech. The first two exercises centre around spoken expressions described in Table 1, while the third exercise focuses on discourse marker vocabulary presented in Table 2. All exercises are tailored for advanced learners at aproficiency level of C1 or higher. The first task involves material from the subtitle subcorpus of InterCorp release 12 (Rosen, Vavřín, & Zasina, 2019b). The objective is to introduce new spoken vocabulary to learners, requiring them to replace underlined expressions with formal or neutral equivalents familiar from their prior education at lower levels. All sentences are drawn from the subtitle subcorpus, which emulates spoken Czech. Subsequently, to practise these new expressions, learners collaborate in pairs to prepare adialogue. They then perform these roles in front of the class, with the teacher providing assistance and correcting any mispronunciations after the learners’ presentation. This exercise serves to acquaint learners with the topic of spoken expressions. 6 The corpus-based exercises presented here were published in aworkbook (Zasina, 2023, p.86–87). OPEN ACCESS 118 LINGUISTICA PRAGENSIA 2/2024 Exercise 1.Podívejte se na níže uvedené věty ze subkorpusu filmových titulků korpusu InterCorp v12. Nahraďte potržené mluvené varianty výrazů jejich formálními nebo neutrálními ekvivalenty nacházejícími se vrámečku. Následně spartnerem zkuste vytvořit dialog spoužitím těchto mluvených výrazů. pořád, dům, dítě, máma, kilogram, káva, legrace, táta, sestra, žena, bratr, polévka 1. Furt to ještě bolí. 2.Vtak velkém baráku jsem ještě nebyl. 3.Má tu děcko, máme být potichu. 4.Mamka nemá ráda růžovou. 5.Zhubla jsem o2 kila. 6.Možná se stavím večer na kafe. 7.Tohle už není žádná sranda! 8.Podívejte, co mi taťka koupil kpromoci. 9.Jseš ta nejlepší ségra, kterou mám. 10.Ta ženská mi bude tolik chybět. 11.Tvůj brácha na tebe moc tlačí. 12.Přinesl jsem vám trochu čínský polívky. The second task underscores the discernible distinctions between written and spoken language, with the objective of elucidating aclear contrast to foster learners’ awareness of employing appropriate vocabulary in specific contexts. This exercise utilizes sentence examples representing typical spoken utterances from the ORTOFON v1 spoken corpus (Kopřivová et al., 2017) and typical written utterances from the SYN2015 written corpus (Křen et al., 2015). Consequently, each sentence serves as an authentic illustration of real natural language usage. Learners engage with word pairs, requiring them to choose between spoken and neutral/formal expressions to fill gaps based on contextual cues and characteristic features of text style that denote aparticular register. Through this task, learners have the opportunity to enhance their ability to recognize the distinctive features of spoken and written styles. Exercise 2.Která varianta je vhodná pro psaný akterá pro mluvený styl? Doplňte správný výraz podle charakteristických vlastností stylu textu. legrace × sranda Zítra si možná užiji …………… No vidíš, to je přesně vono, že si zněj dělaj lidi …………… Občas, ale spíš si ze mě lidé dělají …………. nebo pokřikují Oto. peníze × prachy Tam jde vo …………. alidi… se teď nechtěj učit novejm věcem. Část …………. získaných prodejem složila před odjezdem ufaráře. Stát nemusel vyplácet …………, které tak jako tak neměl. kamarád × kámoš Slyšelas to vod jednoho …………, že to říkal, ateď jsem to četla ivknížce. Tři …………. ho stále tiše pozorovali. Tak ten jeho …………. to je nějakej několikanásobnej železnej muž vČechách. OPEN ACCESS ADRIAN JAN ZASINA 119 fotoaparát × foťák Budete-li si ………… vybírat vobchodě, nezapomeňte si vyzkoušet právě video. Mě to hrozně nebavilo fotit… navíc mi přišlo že ten ………… je nějakej špatnej. Ve výbavě nechybí 5Mpx …………, A-GPS nebo digitální kompas. policisté × policajti Když ………….. nehodu vyšetřovali, někdo je nasměroval právě na domnělého pachatele. Prostě zavolám …………. anechám si to prostě zaplatit od silnic, že? Atak se prej na sebe tak ti …………. podívali ateprv pak jakože rychle naskočili do auta. The third exercise is centred on an authentic dialogue extracted from the ORTOFON corpus (Kopřivová et al., 2017), featuring asubstantial use of DM. Initially, learners have the opportunity to immerse themselves in agenuine conversation and identify DM based on aprovided list in the included box. The dialogue can be read independently by learners or read aloud in the classroom by selected students. Following the reading, learners are allotted time for text analysis. Subsequently, the task involves contemplating the function of DM within the text. Learners collaborate in pairs to discuss their findings before engaging in an open class discussion to share individual insights. In the third phase, learners are tasked with applying the information acquired in practical terms. They work in pairs to prepare adialogue, specifically utilizing DM not present in the provided example dialogue. Learners may seek assistance from the teacher or consult corpus data to explore additional instances of DM usage. Finally, they present their dialogues to their classmates. Exercise 3.Přečtěte si pasáž neformální konverzace zkorpusu ORTOFON apokuste se na základě níže uvedené tabulky označit diskurzní markery vdialogu. Jakou plní roli vtextu? Společně spartnerem připravte dialog spoužitím dalších diskurzních markerů, jež se vdialogu neobjevily. myslet, viď, hele, normálně, jakože, jó/jo, cože, hej, vyloženě, kdyžtak, holt, schválně, víceméně Martina: (…) To se mi stalo oprázdninách, že jsem si otevřela, že jsem nějaké téma našla azajímalo mě… Tak jsem to otevřela, Mladou frontu iPrávo, aměli úplně kompletně stejný článek… Veronika: Jako úplně doslova? Martina: … jenom měli jinou fotku… hmm Veronika: Ty jo! Martina: Jakože no doslova jakože… některé ty pasáže byly fakt doslovné… některé byly jakože malinko jiné, ale jakože fakt se to lišilo víceméně vobrázku. Veronika: Ty jo! OPEN ACCESS 120 LINGUISTICA PRAGENSIA 2/2024 Martina: Takže to by bylo fakt jako docela drsné, protože já jsem četla nejprve Mladou frontu. Apak jsem si říkala, tak jako zkusím se ještě podívat třeba do tých jakože do Hospodářek ado Práva a… normálně to na mě vyjelo to samé. Veronika: Jo, jo, jo, jo. Martina: Ateďka člověk řekne, ty jo, však jsem to před chvílí četla toto… jako hmm. Veronika: Tak to je zvláštní jako. Martina: Ale je to zajímavé dělat to otom kraji. Veronika: Já mám strašnou žízeň normálně… po tom… výstupu tam. Martina: Já jsem myslela, že po víkendu. Veronika: To ne, to měla Renča už včera. Martina: To měla Renča. Veronika: Jak ona furt pila apila ajá říkám, ty jo, to máš ještě zté soboty určitě aona: myslíš? Martina: Však já jsem jí to taky říkala, já jsem si tam dělala šťávu vkuchyni… aona říkala já jsem měla strašnou žízeň, včera já jsem vypila během rána dva litry (…) The proposed exercises can be modified by each teacher to adapt them to the needs of learners. The exercises are also suitable as an introduction to the topic of spoken Czech or as revision material. The corpus evidence included in all three exercises provides clear information about the differences between spoken and written language. 5 CONCLUSION Bearing in mind the complex situation of the Czech language bordering on diglossia, this research attempts to analyse spoken expressions and DM that are crucial in communication with native speakers and might later be used to teach speaking skills in Czech as aforeign language. The analysis indicates that the subtitle component of the parallel corpus may be ahelpful source for examining translation equivalents of typical spoken vocabulary. The quasi-spoken parallel data enable the identification of relevant equivalents in English, French, and Polish that correspond to the spoken register as well. This is an especially valuable finding because it is not obvious to relate parallel data with spoken language, as they normally consist of fiction and non-fiction. The subtitle collection certainly brings benefits; however, it is necessary to take into account the potential bias that might influence the results. For instance, the most frequent translation equivalent may not always be appropriate for spoken language. Therefore, alinguist’s expertise and context analysis are also required. Furthermore, corpus-based exercises demonstrate how to approach the teaching features of the spoken register using corpus data. All presented exercises are comprised of authentic texts representing written, quasi-spoken, and spoken Czech. They effectively introduce learners to the topic of spoken Czech and provide clear information regarding differences and vocabulary usage. Moreover, they compel learners to practise oral production and utilize spoken expressions and discourse markers to sound more natural when speaking. Particularly valuable is the OPEN ACCESS ADRIAN JAN ZASINA 121 situational context delivered by corpus instances, allowing learners to familiarize themselves with appropriate usage. This research also illustrates that material such as subtitles is essential in aparallel corpus to capture spoken elements in quasi-spoken texts. Therefore, such sources should be expanded not only by increasing the collection of subtitles but also by including film or drama scripts to provide more relevant data for examining interlingual correspondence in spoken language. Despite their weaknesses, subtitles have certain advantages that are often overlooked by some researchers (Charciarek, 2018, 2019), who cite only afew benefits (such as being asource of idioms for translation dictionaries) and ignore the usefulness of subtitles in analysing the features of spoken language. Potential limitations of this study might arise from the analysis of data sourced from film subtitles. Although film subtitles strive to replicate spoken language, there is alack of empirical evidence to assess their proximity to real spoken language. Consequently, their authenticity could be questioned. Hence, further research is necessary to gauge the extent of similarity between film subtitles and naturally occurring spontaneous informal conversations. This could be achieved through acomparative analysis, utilising amulti-dimensional space where any dataset can be compared against amodel of the Czech language. Asimilar approach was proposed in previous studies that compared web-crawled and traditional corpora (Cvrček, Komrsková, et al., 2020) or elicited texts (Cvrček, Laubeová, et al., 2020a). By projecting film subtitles onto amulti-dimensional model of Czech, their prevalence within both spoken and non-spoken texts could be revealed. The study substantially adds to our understanding of spoken expressions and discourse markers in Czech and their English, French, and Polish equivalents relevant to spoken registers. The results offer valuable contributions to teaching and learning speaking skills in Czech through corpus-based exercises. It is afirst attempt to use corpus methods for familiarizing learners with spoken language using authentic Czech texts. This approach potentially shapes the further development of teaching methods, which have traditionally relied on prepared texts, and could considerably enhance the development of learners’ spoken language. Furthermore, parallel data provide awide array of translation equivalents; unlike printed dictionaries, they additionally offer contextual examples and rare, occasional translations that may be suitable only in specific situations. Traditional dictionaries typically provide the most frequent equivalents and often become outdated. The research provides aclear explanation of using corpus-based exercises in the classroom to foster learners’ autonomy in developing fluency and achieving amore native-like performance in oral communication. It also contributes to previous studies conducted on the English language (Boulton, 2009; Huang, 2019; Şahin Kızıl & Savran, 2018) that do not use parallel corpora to identify translation equivalents in the learners’ native languages, acrucial aspect for comprehending structures in aforeign language. OPEN ACCESS 122 LINGUISTICA PRAGENSIA 2/2024 Acknowledgements My thanks go to Petra Poukarová for fielding my questions about discourse markers and to Alexandr Rosen for his supportive comments. This work was supported by the Cooperatio Program, research area Linguistics, and the project Multilingual Lens: Investigating Large Text Corpora from Different Methodological Perspectives (UNCE/24/ SSH/009). REFERENCES Baker, P., & Ellece, S. (2011). Key terms in discourse analysis. London: Bloomsbury Publishing. Bańczyk, Ł., Dybalska, R., & Vavřín, M. (2019). InterCorp— Polish, Release 12 of 12December 2019 [Corpus]. Prague: Institute of the Czech National Corpus, Charles University. www. korpus.cz Barlow, M. (2000). Parallel texts in language teaching. In S.P.Botley, A.M.McEnery, & A.Wilson (Eds.), Multilingual Corpora in Teaching an Research (pp.106–115). Amsterdam-Atlanta: Rodopi. Bermel, N. (2014). Czech Diglossia: Dismantling or Dissolution? In J.Arokay, J.Gvozdanovic, & D.Miyajima (Eds.), Divided Languages? Diglossia, Translation and the Rise of Modernity in Japan, China, and the Slavic World (1st ed., pp.21–37). Dordrecht: Springer International Publishing. Boulton, A. (2009). Testing the limits of datadriven learning: Language proficiency and training. ReCALL, 21(1), 37–54. Boulton, A., & Cobb, T. (2017). Corpus Use in Language Learning: AMeta-Analysis. Language Learning, 67(2), 348–393. https://doi. org/10.1111/lang.12224 Bulejčíková, P. (2015). Problematika spisovnosti se zřetelem kvýuce češtiny jako cizího jazyka [Standard Language with Regard to Teaching Czech as aForeign Language] [Dissertation, Charles University]. Charles University, Prague. https://dspace.cuni.cz/ handle/20.500.11956/63386 Čermák, F., & Rosen, A. (2012). The case of InterCorp, amultilingual parallel corpus. International Journal of Corpus Linguistics, 17(3), 411–427. Čermáková, A., Jílková, L., Komrsková, Z., Kopřivová, M., & Poukarová, P. (2019). Diskurzní markery. In J.Hoffmannová, J.Homoláč, & K.Mrázková (Eds.), Syntax mluvené češtiny (pp.244–351). Praha: Academia. Charciarek, A. (2018). Možnosti využití korpusu InterCorp včesko-polské překladové lexikografii. Časopis pro moderní filologii, 100(2), 206–222. Charciarek, A. (2019). Využití paralelního korpusu vtranslatologii (na základě českopolského InterCorpu). Bohemistyka, XIX(2), 194–216. https://doi.org/10.14746/bo.2019.2.5 Cvrček, V., Komrsková, Z., Lukeš, D., Poukarová, P., Řehořková, A., Zasina, A.J., & Benko, V. (2020). Comparing web-crawled and traditional corpora. Language Resources and Evaluation, 54, 713–745. https://doi. org/10.1007/s10579-020-09487-4 Cvrček, V., Laubeová, Z., Lukeš, D., Poukarová, P., Řehořková, A., & Zasina, A.J. (2020a). Author and register as sources of variation: Acorpus-based study using elicited texts. International Journal of Corpus Linguistics, 25(4), 461–488. https://doi.org/10.1075/ijcl.19020. cvr Cvrček, V., Laubeová, Z., Lukeš, D., Poukarová, P., Řehořková, A., & Zasina, A.J. (2020b). Registry včeštině. Praha: Nakladatelství Lidové noviny. Gilmore, A. (2004). Acomparison of textbook and authentic interactions. ELT Journal, 58(4), 363–374. Holá, L. (2019). Běžně mluvená čeština ve výuce češtiny jako cizího jazyka. In M.Nekula & K.Šichová (Eds.), Variety češtiny ačeština jako cizí jazyk (pp.107–127). Praha: Akropolis. OPEN ACCESS ADRIAN JAN ZASINA 123 Hrdlička, M. (2019). Spisovná aobecná čeština ve výuce cizinců. In M.Nekula & K.Šichová (Eds.), Variety češtiny ačeština jako cizí jazyk (pp.85–106). Praha: Akropolis. Huang, L. (2019). Acorpus-based exploration of the discourse marker well in spoken interlanguage. Language and Speech, 62(3), 570–593. https://doi. org/10.1177/0023830918798863 Johns, T. (1991). Should you be persuaded: Two samples of data-driven learning materials. Classroom Concordancing: ELR Journal, (4), 1–16. Klégr, A., Kubánek, M., Malá, M., Rohrauer, L., Šaldová, P., & Vavřín, M. (2019). InterCorp— English, Release 12 of 12December 2019 [Corpus]. Prague: Institute of the Czech National Corpus, Charles University. www.korpus.cz Komrsková, Z., Kopřivová, M., Lukeš, D., Poukarová, P., & Goláňová, H. (2017). New Spoken Corpora of Czech: ORTOFON and DIALEKT. Journal of Linguistics/Jazykovedný Časopis, 68(2), 219–228. https://doi. org/10.1515/jazcas-2017-0031 Kopřivová, M., Komrsková, Z., Lukeš, D., Poukarová, P., & Škarpová, M. (2017). ORTOFON v1: Korpus neformální mluvené češtiny svíceúrovňovým přepisem. [Corpus] Praha: Ústav Českého národního korpusu FF UK. www.korpus.cz Křen, M., Cvrček, V., Čapka, T., Čermáková,A., Hnátková, M., Jelínek, T., Kováříková, D., Petkevič, V., Procházka, P., Škrabal, M., Truneček, P., Vondřička, P., Zasina, A.J. (2015). SYN2015: Reprezentativní korpus psané češtiny [Corpus]. Praha: Ústav Českého národního korpusu FF UK. www.korpus.cz Křen, M., Cvrček, V., Čapka, T., Čermáková, A., Hnátková, M., Jelínek, T., Kováříková, D., Petkevič, V., Procházka, P., Škrabal, M., Truneček, P., Vondřička, P., Zasina, A.J. (2016). SYN2015: Representative Corpus of Contemporary Written Czech. Proceedings of the Tenth International Conference on Language Resources and Evaluation, 2522–2528. Portorož: ELRA. McCarthy, M., & Carter, R. (2001). Size Isn’t Everything: Spoken English, Corpus, and the Classroom. TESOL Quarterly, 35(2), 337–340. https://doi.org/10.2307/3587654 Nádvorníková, O., & Vavřín, M. (2019). InterCorp— French, Release 12 of 12December 2019 [Corpus]. Prague: Institute of the Czech National Corpus, Charles University. www. korpus.cz Rosen, A., Vavřín, M., & Zasina, A.J. (2019a). InterCorp— Czech, Release 12 of 12December 2019 [Corpus]. Prague: Institute of the Czech National Corpus, Charles University. www. korpus.cz Rosen, A., Vavřín, M., & Zasina, A.J. (2019b). InterCorp, Release 12 of 12December 2019 [Corpus]. Prague: Institute of the Czech National Corpus, Charles University. www. korpus.cz Şahin Kızıl, A., & Savran, Z. (2018). The Integration of Corpus into EFL Speaking Instruction: AStudy of Learner Perceptions. International Online Journal of Education and Teaching, 5(2), 376–389. Schiffrin, D. (1987). Discourse markers. Cambridge: Cambridge University Press. https://doi.org/10.1017/CBO9780511611841 Škrabal, M., Laubeová, Z., & Štěpánková, B. (2022). Korpusové přístupy kčeské diglosii. Praha: Nakladatelství Lidové noviny. Škrabal, M., & Vavřín, M. (2017). Databáze překladových ekvivalentů Treq. Časopis pro moderní filologii, 99(2), 245–260. Varga, D., Halácsy, P., Kornai, A., Nagy, V., Németh, L., & Trón, V. (2007). Parallel corpora for medium density languages. In N.Nicoloy, R.Mitkov, G.Angelova, & K.Bontcheva (Eds.), Recent Advances in Natural Language Processing IV (pp.247–258). Amsterdam/Philadelphia: John Benjamins Publishing. Vavřín, M., & Rosen, A. (2015). Treq (2.0) [Computer software]. Praha: FF UK. https:// treq.korpus.cz/ Vondřička, P. (2014). Aligning parallel texts with InterText. In N.Calzolari, K.Choukri, T.Declerck, H.Loftsson, B.Maegaard, J.Mariani, A.Moreno, J.Odijk, & S.Piperidis (Eds.), Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC’14) (pp.1875–1879). OPEN ACCESS 124 LINGUISTICA PRAGENSIA 2/2024 Reykjavik: European Language Resources Association (ELRA). Walsh, S. (2010). What features of spoken and written corpora can be exploited in creating language teaching materials and syllabuses. In A.O’Keeffe & M.McCarthy (Eds.), The Routledge Handbook of Corpus Linguistics (pp.333–344). London— New York: Routledge. Zasina, A.J. (2023). Korpusová cvičebnice pro studenty češtiny jako cizího jazyka. Praha: Karolinum. Adrian Jan Zasina Institute of Czech and Deaf Studies Faculty of Arts, Charles University Na Příkopě 584/29 110 00 Praha 1 ORCID ID: 0000-0001-9348-5833 [email protected] OPEN ACCESS