Radoslava Trnavac
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Radoslava Trnavac has a total of 8 co-authors. Here are the top co-authors by number of shared publications:
- Maite Taboada
Simon Fraser University
3 shared publications - Nele Põldvere
University of Oslo
2 shared publications - Cliff Goddard
Griffith University
1 shared publication Darya Kharlamova
1 shared publication- Debopam Das
Humboldt-Universität zu Berlin
1 shared publication Encarnacion Hidalgo Tenorio
1 shared publication
Latest Publications
- 12 October 2026
- Published by John Benjamins Publishing Company in Pragmatics and Society
From persuasion techniques to pragmatic mechanisms
In this study, we describe the language of persuasion in a specific type of fake news used to mislead people by often implicit means, namely, propaganda. In this way, propaganda lends itself well to a pragmatic analysis of the underlying mechanisms that are exploited to convince people to adopt a specific ideology or worldview. Our aim is to identify these pragmatic mechanisms, and the linguistic forms through which they are realized, in a systematic way using the Fakespeak-ENG corpus. We both draw on and add to an existing framework of persuasion by taking it as our starting point for identifying persuasion in the corpus texts and by adding to it another analytical layer that takes language more seriously. The results show that the most common persuasion techniques in our propaganda texts are Loaded Language, Appeal to Authority and Appeal to Time, which are characterized by negative sentiment words, reporting verbs and temporal deictic markers, respectively. In addition, we emphasize the manipulative use of figurative language, false implicatures and presuppositions, and typographic emphasis to convey persuasion in more or less subtle ways. Unpacking these linguistic strategies will add to existing work on the language of fake news in general and propaganda in particular and have real-world implications for automatic propaganda detection and media literacy education.
- 12 October 2026
- Published by John Benjamins Publishing Company in Pragmatics and Society
From persuasion techniques to pragmatic mechanisms
In this study, we describe the language of persuasion in a specific type of fake news used to mislead people by often implicit means, namely, propaganda. In this way, propaganda lends itself well to a pragmatic analysis of the underlying mechanisms that are exploited to convince people to adopt a specific ideology or worldview. Our aim is to identify these pragmatic mechanisms, and the linguistic forms through which they are realized, in a systematic way using the Fakespeak-ENG corpus. We both draw on and add to an existing framework of persuasion by taking it as our starting point for identifying persuasion in the corpus texts and by adding to it another analytical layer that takes language more seriously. The results show that the most common persuasion techniques in our propaganda texts are Loaded Language, Appeal to Authority and Appeal to Time, which are characterized by negative sentiment words, reporting verbs and temporal deictic markers, respectively. In addition, we emphasize the manipulative use of figurative language, false implicatures and presuppositions, and typographic emphasis to convey persuasion in more or less subtle ways. Unpacking these linguistic strategies will add to existing work on the language of fake news in general and propaganda in particular and have real-world implications for automatic propaganda detection and media literacy education.
- Open Access
- 4 March 2026
- Published by Elsevier in SSRN Electronic Journal
Simple Solutions for Complex Tasks: Pre-LLM Approaches to Automated Coherence Assessment in Non-Native English Essays
Discourse coherence is crucial in evaluating student writing, especially in second language contexts, where the ability to produce logically connected, readable text indicates academic proficiency. However, currently, no universally accepted linguistically backed automated coherence evaluation system for student writing exists. This study aims to achieve three goals: first, it presents a novel linguistically supported annotation system for automated coherence assessment in student writing that takes into consideration both local and global coherence, with the best-performing models made publically available. Secondly, it investigates the possibility of developing a linguistically driven coherence evaluation model utilizing LLM-based annotation. Third, it proposes a methodology that uses computationally inexpensive pre-LLM techniques to approximate LLM performance, hence addressing the issues over excessive LLM dependency.Several pre-LLM techniques, including traditional machine learning algorithms (logistic regression, decision tree classifier, gradient boosting) and encoder-based language model fine-tuning (BERT, RoBERTa, ModernBERT) in the learner corpus of English (REALEC), were trained using the LLM-annotated dataset across four coherence dimensions (Thesis Relatedness, Grammatical Correctness, Syntactic Coherence, and Semantic Coherence). In all four dimensions, the top-performing models obtained F1-scores of 0.88–0.91, which nearly resembled the LLM baseline performance while using substantially less computational power and, for the most part, producing deterministic, comprehensible results.
- 4 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSimple Solutions for Complex Tasks: Pre-LLM Approaches to Automated Coherence Assessment in Non-Native English Essays
Discourse coherence is crucial in evaluating student writing, especially in second language contexts, where the ability to produce logically connected, readable text indicates academic proficiency. However, currently, no universally accepted linguistically backed automated coherence evaluation system for student writing exists. This study aims to achieve three goals: first, it presents a novel linguistically supported annotation system for automated coherence assessment in student writing that takes into consideration both local and global coherence, with the best-performing models made publically available. Secondly, it investigates the possibility of developing a linguistically driven coherence evaluation model utilizing LLM-based annotation. Third, it proposes a methodology that uses computationally inexpensive pre-LLM techniques to approximate LLM performance, hence addressing the issues over excessive LLM dependency.Several pre-LLM techniques, including traditional machine learning algorithms (logistic regression, decision tree classifier, gradient boosting) and encoder-based language model fine-tuning (BERT, RoBERTa, ModernBERT) in the learner corpus of English (REALEC), were trained using the LLM-annotated dataset across four coherence dimensions (Thesis Relatedness, Grammatical Correctness, Syntactic Coherence, and Semantic Coherence). In all four dimensions, the top-performing models obtained F1-scores of 0.88–0.91, which nearly resembled the LLM baseline performance while using substantially less computational power and, for the most part, producing deterministic, comprehensible results.
- Open Access
- 2 October 2025
- Published by Peoples' Friendship University of Russia in Russian Journal of Linguistics
- Vol. 29 (3) , 560-585
The interplay of conceptual metaphors and evaluation in press reports on the AUKUS agreement
The linguistic literature has shown that metaphor provokes evaluative meanings, but it is unclear how different types of metaphor, as well as different genres of discourse influence the realization of such meanings. To partially answer this question, our study aims to investigate the relationship between conceptual metaphors and evaluation in Australian broadsheet and tabloid articles on the AUKUS alliance, the trilateral security partnership between Australia, the U.K., and the U.S. The data comprise all Australian news and opinion articles gathered from Nexis Uni through the University of Granada’s e-library between September 15, 2021, and October 31, 2021. The study employs a parameter-based approach to evaluation and its methodology involves the annotation of metaphors according to the parameters of creativity/conventionality, as well as the examination of their source and target domains. The quantitative analysis indicates that there is a connection between the types of metaphors (conventional and creative) and the characteristics of their evaluation (polarity and explicitness), which are affected by the genre of the corpus and the topic of the content where the metaphors appear. Additionally, the corpus shows that negative emotivity is a prevalent evaluative component of metaphors in both newspapers. The qualitative analysis demonstrates that conceptualizations of metaphors alternate between biological and mechanistic images of the world, with the dominant metaphor domains being almost unilaterally used in both broadsheets and tabloids. The implications of the study suggest that the complex relationship between metaphor and evaluation should be interpreted in light of the text’s genre and topic rather than separately.
- 2 October 2025
- Published by Peoples' Friendship University of Russia in Russian Journal of Linguistics
- Vol. 29 (3) , 560-585
Open AccessThe interplay of conceptual metaphors and evaluation in press reports on the AUKUS agreement
The linguistic literature has shown that metaphor provokes evaluative meanings, but it is unclear how different types of metaphor, as well as different genres of discourse influence the realization of such meanings. To partially answer this question, our study aims to investigate the relationship between conceptual metaphors and evaluation in Australian broadsheet and tabloid articles on the AUKUS alliance, the trilateral security partnership between Australia, the U.K., and the U.S. The data comprise all Australian news and opinion articles gathered from Nexis Uni through the University of Granada’s e-library between September 15, 2021, and October 31, 2021. The study employs a parameter-based approach to evaluation and its methodology involves the annotation of metaphors according to the parameters of creativity/conventionality, as well as the examination of their source and target domains. The quantitative analysis indicates that there is a connection between the types of metaphors (conventional and creative) and the characteristics of their evaluation (polarity and explicitness), which are affected by the genre of the corpus and the topic of the content where the metaphors appear. Additionally, the corpus shows that negative emotivity is a prevalent evaluative component of metaphors in both newspapers. The qualitative analysis demonstrates that conceptualizations of metaphors alternate between biological and mechanistic images of the world, with the dominant metaphor domains being almost unilaterally used in both broadsheets and tabloids. The implications of the study suggest that the complex relationship between metaphor and evaluation should be interpreted in light of the text’s genre and topic rather than separately.
- 5 October 2024
- Published by Elsevier in Applied Corpus Linguistics
- Vol. 4 (3)
Breach of pacta sunt servanda: A corpus-assisted analysis of newspaper discourse on the AUKUS agreement
- 5 October 2024
- Published by Elsevier in Applied Corpus Linguistics
- Vol. 4 (3)
Breach of pacta sunt servanda: A corpus-assisted analysis of newspaper discourse on the AUKUS agreement
- Open Access
- 27 September 2024
- Published by Cambridge University Press (CUP) in English Language and Linguistics
- p. 1-30
The ‘adverb-ly adjective’ construction in English: meanings, distribution and discourse functions
We investigate a class of adjective phrases composed of a deadjectival adverb ending in -ly and an adjective head (e.g. staggeringly incompetent, absolutely terrific, fiscally responsible), a compact construction whereby two adjectives may jointly contribute to evaluative meaning. Using corpus methodologies on more than 1 million examples and relying on semantic analyses of about 1,000 instances, we propose that the construction can be divided into different semantic subtypes, including Degree (deeply disturbing), Focus (utterly ridiculous), Manner (delightfully performed), Reaction (strangely compelling), Topical (historically inaccurate) and Epistemic (intuitively obvious), among others. Using this typology, we investigate the relative distribution of each subtype across several registers of written English. We found a high frequency of the Reaction subtype in book, film and art reviews, and we suggest a discourse-functional explanation for this, linked to the perceived value of originality in expressive writing. This investigation reveals the power of semantically informed, corpus methodologies to shed light on the distribution of specific constructions.
- 27 September 2024
- Published by Cambridge University Press (CUP) in English Language and Linguistics
- p. 1-30
Open AccessThe ‘adverb-ly adjective’ construction in English: meanings, distribution and discourse functions
We investigate a class of adjective phrases composed of a deadjectival adverb ending in -ly and an adjective head (e.g. staggeringly incompetent, absolutely terrific, fiscally responsible), a compact construction whereby two adjectives may jointly contribute to evaluative meaning. Using corpus methodologies on more than 1 million examples and relying on semantic analyses of about 1,000 instances, we propose that the construction can be divided into different semantic subtypes, including Degree (deeply disturbing), Focus (utterly ridiculous), Manner (delightfully performed), Reaction (strangely compelling), Topical (historically inaccurate) and Epistemic (intuitively obvious), among others. Using this typology, we investigate the relative distribution of each subtype across several registers of written English. We found a high frequency of the Reaction subtype in book, film and art reviews, and we suggest a discourse-functional explanation for this, linked to the perceived value of originality in expressive writing. This investigation reveals the power of semantically informed, corpus methodologies to shed light on the distribution of specific constructions.
- Open Access
- 28 February 2024
- Published by Springer Nature in Corpus Pragmatics
- Vol. 8 (2) , 107-130
Investigating Appraisal and the Language of Evaluation in Fake News Corpora
The present corpus study, which is grounded in Appraisal Theory, investigates evaluative language use in fake news in English. The primary aim is to find out how and why, if at all, evaluative meanings are construed differently in fake news compared to genuine news. The secondary aim is to explore potential differences between types of fake news based on contextual factors. The data are from two carefully-designed corpora containing both fake and genuine news: a single-authored corpus and a multi-authored corpus. Both corpora contain false information that is meant to deceive, but they also differ from each other in terms of register, genre and the motivational goals of the authors. Through qualitative and quantitative analyses, we show that there are systematic differences in the occurrence of Appraisal expressions across fake and genuine news, with Appraisal being more common in the former. However, the exact nature of the affective, dialogic and modal expression of fake news is influenced by contextual factors that, so far, have largely been ignored in fake news research. Therefore, the study has important implications for the development of fake news detection systems based on data sources of different kinds, a task which is in grave need of the input of corpus linguists.
- 28 February 2024
- Published by Springer Nature in Corpus Pragmatics
- Vol. 8 (2) , 107-130
Open AccessInvestigating Appraisal and the Language of Evaluation in Fake News Corpora
The present corpus study, which is grounded in Appraisal Theory, investigates evaluative language use in fake news in English. The primary aim is to find out how and why, if at all, evaluative meanings are construed differently in fake news compared to genuine news. The secondary aim is to explore potential differences between types of fake news based on contextual factors. The data are from two carefully-designed corpora containing both fake and genuine news: a single-authored corpus and a multi-authored corpus. Both corpora contain false information that is meant to deceive, but they also differ from each other in terms of register, genre and the motivational goals of the authors. Through qualitative and quantitative analyses, we show that there are systematic differences in the occurrence of Appraisal expressions across fake and genuine news, with Appraisal being more common in the former. However, the exact nature of the affective, dialogic and modal expression of fake news is influenced by contextual factors that, so far, have largely been ignored in fake news research. Therefore, the study has important implications for the development of fake news detection systems based on data sources of different kinds, a task which is in grave need of the input of corpus linguists.
- 11 October 2021
- Published by Walter de Gruyter GmbH in Text & Talk
- Vol. 43 (2) , 235-262
Engagement and constructiveness in online news comments in English and Russian
We investigate the relationship between Engagement and constructiveness in online news comments by analyzing the frequency and type of Engagement expressions in a corpus of English and Russian comments, following the Appraisal framework. The comments in question, 10,000 words in each language, were posted in response to opinion articles in the Canadian newspaper The Globe and Mail and the Russian online news channel RT . In the context of online news comments, users generally characterize constructive comments as posts that tend to create a civil dialogue through remarks that are relevant to the article and do not provoke an emotional response. Through quantitative and qualitative analyses, we conclude that the language of constructive comments is more explicitly subjective in both languages. The main difference in the use of Engagement expressions in constructive and non-constructive comments lies along the lines of certainty/uncertainty and reliability/unreliability. As for cross-linguistic differences, it seems that English constructive comments place emphasis on the reliability of a commenter’s knowledge, while Russian constructive comments employ more modals of necessity, which have a prescriptive function.
- 11 October 2021
- Published by Walter de Gruyter GmbH in Text & Talk
- Vol. 43 (2) , 235-262
Engagement and constructiveness in online news comments in English and Russian
We investigate the relationship between Engagement and constructiveness in online news comments by analyzing the frequency and type of Engagement expressions in a corpus of English and Russian comments, following the Appraisal framework. The comments in question, 10,000 words in each language, were posted in response to opinion articles in the Canadian newspaper The Globe and Mail and the Russian online news channel RT . In the context of online news comments, users generally characterize constructive comments as posts that tend to create a civil dialogue through remarks that are relevant to the article and do not provoke an emotional response. Through quantitative and qualitative analyses, we conclude that the language of constructive comments is more explicitly subjective in both languages. The main difference in the use of Engagement expressions in constructive and non-constructive comments lies along the lines of certainty/uncertainty and reliability/unreliability. As for cross-linguistic differences, it seems that English constructive comments place emphasis on the reliability of a commenter’s knowledge, while Russian constructive comments employ more modals of necessity, which have a prescriptive function.
- Open Access
- 1 August 2016
- Published by Edinburgh University Press in Corpora
- Vol. 11 (2) , 169-190
Discourse relations and evaluation
In this paper, we examine the role of discourse relations (relations between propositions) in the interpretation of evaluative or opinion words. Through a combination of Rhetorical Structure Theory (or RST; Mann and Thompson, 1988 ) and Appraisal Theory ( Martin and White, 2005 ), we analyse how different discourse relations modify the evaluative content of opinion words, and what impact the nucleus–satellite structure in RST has on the evaluation. We conduct a corpus study, examining and annotating over 3,000 evaluative words in fifty movie reviews in the SFU Review Corpus ( Taboada, 2008 ) with respect to five parameters: word category (noun, verb, adjective or adverb), prior polarity (positive, negative or neutral), RST structure (both nucleus–satellite status and relation type) and change of polarity as a result of being part of a discourse relation (Intensify, Downtone, Reversal or No Change). Results show that relations such as Concession, Elaboration, Evaluation, Evidence and Restatement most frequently intensify the polarity of opinion words, although the majority of evaluative words do not undergo changes in their polarity related to the type of relation that they are a part of. We also find that most opinion words (about 70 percent) are positioned in the nucleus, confirming a hypothesis based on the literature that nuclei are the most important units when extracting opinion automatically.
- 1 August 2016
- Published by Edinburgh University Press in Corpora
- Vol. 11 (2) , 169-190
Open AccessDiscourse relations and evaluation
In this paper, we examine the role of discourse relations (relations between propositions) in the interpretation of evaluative or opinion words. Through a combination of Rhetorical Structure Theory (or RST; Mann and Thompson, 1988 ) and Appraisal Theory ( Martin and White, 2005 ), we analyse how different discourse relations modify the evaluative content of opinion words, and what impact the nucleus–satellite structure in RST has on the evaluation. We conduct a corpus study, examining and annotating over 3,000 evaluative words in fifty movie reviews in the SFU Review Corpus ( Taboada, 2008 ) with respect to five parameters: word category (noun, verb, adjective or adverb), prior polarity (positive, negative or neutral), RST structure (both nucleus–satellite status and relation type) and change of polarity as a result of being part of a discourse relation (Intensify, Downtone, Reversal or No Change). Results show that relations such as Concession, Elaboration, Evaluation, Evidence and Restatement most frequently intensify the polarity of opinion words, although the majority of evaluative words do not undergo changes in their polarity related to the type of relation that they are a part of. We also find that most opinion words (about 70 percent) are positioned in the nucleus, confirming a hypothesis based on the literature that nuclei are the most important units when extracting opinion automatically.