Conference Presentations by Nina Szymor
Since the 1990s, various corpus-based studies have looked at the relations between translated and... more Since the 1990s, various corpus-based studies have looked at the relations between translated and non-translated texts, showing systematic differences at lexical, syntactic and discursive levels (e.g. Baker 2004; Mauranen 2000). This study looks at aspectual preferences of individual modal verbs and adverbial predicatives in Polish (musiec, nalezy, powinien, moc, mozna and wolno), comparing their use between a balanced corpus of Polish (NKJP National Corpus of Polish), legal subcorpus of NKJP, and a corpus of translated legal Polish (PELCRA English-Polish Parallel Corpora). Many studies attempted to describe the relationship between modality and aspect, both in general (e.g. Boogaert and Janssen 2007; Abraham and Leiss 2008) as well as Slavic linguistics (e.g. Divjak 2009; Paduceva 2008), but their focus is on the correlations between modality types (e.g. deontic, dynamic and epistemic) and aspect, rather than individual modals and aspect.
Analysis of the three corpora has shown that there are significant differences in the choice of aspectual forms following the modals expressing necessity and obligation. In the balanced representative corpus of Polish (NKJP), musieć, należy, and powinien each behave differently in terms of aspectual preferences. Musieć does not have preference for any aspect, należy shows preference for perfective infinitives, whereas powinien shows preference for imperfectives. In a random sample extracted from the legal subcorpus of NKJP, musieć and powinien show no aspectual preferences, and należy prefers perfective. Interestingly, a random sample taken from a corpus of translated legal texts shows a consistent preference for perfective infinitives following all three modals. Parallel analysis of the translations and their source texts has shown that translators choose to use perfective infinitives following musieć in sentences, where imperfective is also possible and correct, more often than authors of non-translated texts. Experimental validation of the corpus analysis takes the form of acceptability rating task and a translation task, completed by two groups of respondents: trainee translators and professional translators.
The observed differences in aspectual choice between translated and non-translated Polish texts are explained with reference to the characteristics of human cognition. The study contributes to uncovering ‘the conceptual operations which guide the use of translation strategies in the process of recreating meaning’ (Rojo and Ibarretxe-Antunano 2013: 10) – a Cognitive Linguistic trend in Corpus-Based Translation Studies.
Over the past two decades, several theoretical accounts have been proposed to capture modality (i... more Over the past two decades, several theoretical accounts have been proposed to capture modality (inter alia, Perkins 1983, Huddleston 1988, Sweetser 1990, Bybee et al. 1994, van der Auwera and Plungian 1998, Palmer 2001, Hengeveld 2004, Nuyts 2006). Given the abstract nature of this concept, it is unsurprising that linguists do not agree on the number of modality types to distinguish, with Palmer (2001) proposing two types (event modality and propositional modality), Perkins (1983) suggesting three (dynamic, deontic and epistemic modality), and van der Auwera and Plungian (1998) insisting on four (deontic, epistemic, participant‐internal, and participant‐external modality).
In this talk, we present the results of a study on Russian, Polish, Czech and Croatian planned that aims to shed light on the number and nature of modal categories that are supported by language use. We look at the role corpus and experimental data and quantitative methods can play as objective viewpoints on previous intuitive analyses of modal types. In a first step, we analyze random samples of 250 independent observations for the most frequent modal words, extracted from the national corpora. Observations are annotated for modal type as well as for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Multiple correspondence analysis explores the corpus data, and (polytomous) regression models which aspects of usage may be crucial in triggering a specific modal reading. The Polish data reveal low inter‐annotator agreement (60‐80%) for modal types and suggest that only deontic modality is supported by morphological, syntactic or semantic properties.
In a second step, we validate our findings with data from a forced choice design in which a number of naïve native speakers are trained to use different classifications of modality and are asked to label the modality expressed in a number of sentences taken from the random samples.
The analysis captures the way in which the different modality types correlate with usage data and the extent to which theoretical linguistic classifications correspond to semantic categories that naïve speakers employ.
Over the past 2 decades, several theoretical accounts have been proposed to capture modality (Per... more Over the past 2 decades, several theoretical accounts have been proposed to capture modality (Perkins 1983, Huddleston 1988, Sweetser 1990, Bybee et al. 1994, van der Auwera and Plungian 1998, Palmer 2001, Hengeveld 2004, Nuyts 2006). Given the abstract nature of this concept, it is unsurprising that linguists do not agree on the number of modality types to distinguish, with Palmer (2001) proposing two types (event modality and propositional modality), Perkins (1983) suggesting three (dynamic, deontic and epistemic modality), and van der Auwera and Plungian (1998) insisting on four (deontic, epistemic, participant-internal, and participant-external modality).
In this talk, we present the results of a pilot study on Polish data that aims to shed light on the number and nature of modal categories that are supported by language use. More in specific, we look at the role empirical data and quantitative methods can play as objective viewpoints on previous intuitive analyses of modal types. We base our investigation on corpus and experimental data. The corpus-based dataset was extracted from the NKJP [http://www.nkjp.pl/]; the entire dataset contains random samples of 250 independent observations for each of the 6 most frequent modal adverbial and adjectival predicatives. All observations are annotated for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Experimental data was obtained by means of a survey in which naive native speakers of Polish were given a definition and some examples of each modality type and were asked to label the modality expressed in a number of sentences taken from the random samples.
We use multiple correspondence analysis for a first exploration of the corpus data, and (polytomous) regression to model which aspects of usage may be crucial in triggering a specific modal reading. In a second step, we validate our model by contrasting our corpus findings with data from our experiments with native speakers. The analysis thus captures the way in which the different modality types correlate with usage data and the extent to which theoretical linguistic classifications correspond to semantic categories that naïve speakers employ.
Unpacking modality in Polish (Divjak, Szymor and Socha)
Over the past 2 decades, several theoretical accounts have been proposed to capture modality (Per... more Over the past 2 decades, several theoretical accounts have been proposed to capture modality (Perkins 1983, Huddleston 1988, Sweetser 1990, Bybee et al. 1994, van der Auwera and Plungian 1998, Palmer 2001, Hengeveld 2004, Nuyts 2006). Given the abstract nature of this concept, it is unsurprising that linguists do not agree on the number of modality types to distinguish, with Palmer (2001) proposing two types (event modality and propositional modality), Perkins (1983) suggesting three (dynamic, deontic and epistemic modality), and van der Auwera and Plungian (1998) insisting on four (deontic, epistemic, participant-internal, and participant-external modality). There is also lack of unanimity as to the definitions and labels given to the different types of modality. For example, there are differing opinions in respect of the category of dynamic modality - although some linguists agree that it is a separate category (Perkins 1983), others label (some aspects of) it together with deontic modality, under such names as event modality (Palmer 2001), root modality (Hofmann 1976, Coates 1983, Sweetser 1990), or agent-oriented modality (Bybee et al. 1994) (Nuyts 2006: 7).
In this talk as part of the theme session, we present the results of a pilot study on Polish data that aims to shed light on the number and nature of modal categories and looks at the role empirical data, quantitative methods, and visualization tools can play as objective viewpoints on previous intuitive analyses of modal types.
Practically, the analysis aims to capture the way in which the different modality types as defined by Nuyts (2006) correlate with usage data by tracking the behavior of words in modal contexts. The dataset consists of 250 sentences for each of the following 6 relatively frequent modal adverbial and adjectival predicatives as well as verbs: można (can, may), móc (can, may), musieć (need to, must, have to), mieć (have to, must), powinien (should), wolno (one is allowed, one may). The data are extracted from the NKJP and are annotated for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak 2004; Divjak and Gries 2006). A number of statistical techniques are then used to quantify the intuitive clarity of the proposed classifications, to examine which types of modality cluster together and could be grouped under one heading, and to determine how different types of modality correlate with aspects of usage.
Since the 1990s, various corpus-based studies have looked at the relations between translated and... more Since the 1990s, various corpus-based studies have looked at the relations between translated and non-translated texts, showing systematic differences at lexical, syntactic and discursive levels (e.g. Baker 2004; Mauranen 2000). However, differences at the semantic level (e.g. between the synonymous try and attempt, or the polysemous run, get, etc.) have not been given the same attention, perhaps due to lack of appropriate techniques for doing so. A relatively new approach - the Behavioral Profiling (Divjak 2004; Divjak and Gries 2006, and later work) - offers an objective and precise corpus-based methodology that can be used for various cognitively-inspired lexical semantic analyses and that can be applied within and across languages (Divjak & Gries 2009, among others) and to first and second language acquisition data (Deshors 2010).
The aim of this pilot study is to extend the application of Behavioral Profiling from lexical semantic studies within and across languages to the study of lexical semantics in translation. In this case study, I will focus on analyzing instances of modality in translated and non-translated Polish texts to see whether translation affects the ways in which the words are used and their semantic networks are constructed. Data extracted from corpora (PELCRA – Polish and English Language Corpora for Research and Applications) is annotated using Behavioral Profiling and analysed statistically to check whether the observed differences (if any) are significant.
Papers by Nina Szymor
Translation: universals or cognition?
Target. International Journal of Translation Studies
This paper contributes to the ongoing debate on the existence of translation universals by invest... more This paper contributes to the ongoing debate on the existence of translation universals by investigating the use of aspect in modal contexts in translated and non-translated legal Polish and by analysing the observed differences with reference to insights from cognitive linguistics. Corpus analysis highlights significant distributional differences in the use of the two aspectual forms of Polish verbs (imperfective and perfective) in modal contexts. I argue that cognitive mechanisms called ‘chunking’ (Langacker 1988; Bybee 2006) and ‘entrenchment’ (Bybee 2010) underlie these differences. I show that what may at first glance seem as behaviour unique to the translation process, is in fact caused by general cognitive processes. The study has implications for both translation studies and cognitive linguistics: it offers support for the basic assumptions about the usage-based nature of linguistic knowledge and highlights the importance of taking these assumptions into consideration when i...
In this paper we present the results of an empirical study into the cognitive reality of existing... more In this paper we present the results of an empirical study into the cognitive reality of existing classifications of modality using Polish data.
We analyzed random samples of 250 independent observations for the 7 most frequent modal words (móc, można, musieć, należy, powinien, trzeba, wolno), extracted from the Polish national corpus. Observations were annotated for modal type according to a number of classifications, including van der Auwera and Plungian (1998), as well as for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Multiple correspondence analysis and (polytomous) regression models were used to determine how well modal type and usage align. These corpus-based findings were validated experimentally. In a forced choice task, naive native speakers were exposed to definitions and prototypical examples of modal types or functions, then labeled a number of authentic corpus sentences accordingly. In the sorting task, naive native speakers sorted authentic corpus sentences into semantically coherent groups.
We discuss the results of our empirical study as well as the issues involved in building usage-based accounts on traditional linguistic classifications.
Uploads
Conference Presentations by Nina Szymor
Analysis of the three corpora has shown that there are significant differences in the choice of aspectual forms following the modals expressing necessity and obligation. In the balanced representative corpus of Polish (NKJP), musieć, należy, and powinien each behave differently in terms of aspectual preferences. Musieć does not have preference for any aspect, należy shows preference for perfective infinitives, whereas powinien shows preference for imperfectives. In a random sample extracted from the legal subcorpus of NKJP, musieć and powinien show no aspectual preferences, and należy prefers perfective. Interestingly, a random sample taken from a corpus of translated legal texts shows a consistent preference for perfective infinitives following all three modals. Parallel analysis of the translations and their source texts has shown that translators choose to use perfective infinitives following musieć in sentences, where imperfective is also possible and correct, more often than authors of non-translated texts. Experimental validation of the corpus analysis takes the form of acceptability rating task and a translation task, completed by two groups of respondents: trainee translators and professional translators.
The observed differences in aspectual choice between translated and non-translated Polish texts are explained with reference to the characteristics of human cognition. The study contributes to uncovering ‘the conceptual operations which guide the use of translation strategies in the process of recreating meaning’ (Rojo and Ibarretxe-Antunano 2013: 10) – a Cognitive Linguistic trend in Corpus-Based Translation Studies.
In this talk, we present the results of a study on Russian, Polish, Czech and Croatian planned that aims to shed light on the number and nature of modal categories that are supported by language use. We look at the role corpus and experimental data and quantitative methods can play as objective viewpoints on previous intuitive analyses of modal types. In a first step, we analyze random samples of 250 independent observations for the most frequent modal words, extracted from the national corpora. Observations are annotated for modal type as well as for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Multiple correspondence analysis explores the corpus data, and (polytomous) regression models which aspects of usage may be crucial in triggering a specific modal reading. The Polish data reveal low inter‐annotator agreement (60‐80%) for modal types and suggest that only deontic modality is supported by morphological, syntactic or semantic properties.
In a second step, we validate our findings with data from a forced choice design in which a number of naïve native speakers are trained to use different classifications of modality and are asked to label the modality expressed in a number of sentences taken from the random samples.
The analysis captures the way in which the different modality types correlate with usage data and the extent to which theoretical linguistic classifications correspond to semantic categories that naïve speakers employ.
In this talk, we present the results of a pilot study on Polish data that aims to shed light on the number and nature of modal categories that are supported by language use. More in specific, we look at the role empirical data and quantitative methods can play as objective viewpoints on previous intuitive analyses of modal types. We base our investigation on corpus and experimental data. The corpus-based dataset was extracted from the NKJP [http://www.nkjp.pl/]; the entire dataset contains random samples of 250 independent observations for each of the 6 most frequent modal adverbial and adjectival predicatives. All observations are annotated for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Experimental data was obtained by means of a survey in which naive native speakers of Polish were given a definition and some examples of each modality type and were asked to label the modality expressed in a number of sentences taken from the random samples.
We use multiple correspondence analysis for a first exploration of the corpus data, and (polytomous) regression to model which aspects of usage may be crucial in triggering a specific modal reading. In a second step, we validate our model by contrasting our corpus findings with data from our experiments with native speakers. The analysis thus captures the way in which the different modality types correlate with usage data and the extent to which theoretical linguistic classifications correspond to semantic categories that naïve speakers employ.
In this talk as part of the theme session, we present the results of a pilot study on Polish data that aims to shed light on the number and nature of modal categories and looks at the role empirical data, quantitative methods, and visualization tools can play as objective viewpoints on previous intuitive analyses of modal types.
Practically, the analysis aims to capture the way in which the different modality types as defined by Nuyts (2006) correlate with usage data by tracking the behavior of words in modal contexts. The dataset consists of 250 sentences for each of the following 6 relatively frequent modal adverbial and adjectival predicatives as well as verbs: można (can, may), móc (can, may), musieć (need to, must, have to), mieć (have to, must), powinien (should), wolno (one is allowed, one may). The data are extracted from the NKJP and are annotated for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak 2004; Divjak and Gries 2006). A number of statistical techniques are then used to quantify the intuitive clarity of the proposed classifications, to examine which types of modality cluster together and could be grouped under one heading, and to determine how different types of modality correlate with aspects of usage.
The aim of this pilot study is to extend the application of Behavioral Profiling from lexical semantic studies within and across languages to the study of lexical semantics in translation. In this case study, I will focus on analyzing instances of modality in translated and non-translated Polish texts to see whether translation affects the ways in which the words are used and their semantic networks are constructed. Data extracted from corpora (PELCRA – Polish and English Language Corpora for Research and Applications) is annotated using Behavioral Profiling and analysed statistically to check whether the observed differences (if any) are significant.
Papers by Nina Szymor
We analyzed random samples of 250 independent observations for the 7 most frequent modal words (móc, można, musieć, należy, powinien, trzeba, wolno), extracted from the Polish national corpus. Observations were annotated for modal type according to a number of classifications, including van der Auwera and Plungian (1998), as well as for morphological, syntactic and semantic properties using the Behavioral Profiling approach (Divjak and Gries 2006). Multiple correspondence analysis and (polytomous) regression models were used to determine how well modal type and usage align. These corpus-based findings were validated experimentally. In a forced choice task, naive native speakers were exposed to definitions and prototypical examples of modal types or functions, then labeled a number of authentic corpus sentences accordingly. In the sorting task, naive native speakers sorted authentic corpus sentences into semantically coherent groups.
We discuss the results of our empirical study as well as the issues involved in building usage-based accounts on traditional linguistic classifications.