AI-Based Syntactic Complexity Metrics and Sight Interpreting Performance

Zou, Longhui; Carl, Michaël; Mirzapour, Mehdi; Hélène, Jacquenet,; Vieira, Lucas Nunes

doi:10.1007/978-3-030-98404-5_49

Cited by 5 publications

(5 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Additionally, syntactic features derived from parse tree information have also proven important indicators of a text's complexity as first revealed in the work by Schwarm and Ostendorf (2005) relying on a constituency parser and later also corroborated using dependency parsing information (De Clercq and Hoste, 2016). Considering the latter, many efforts have been devoted towards universal dependencies (De Marneffe et al, 2021), which was employed in recent research to operationalize how cognitive effort can be increased by complex syntax during translation through a variety of syntactic complexity metrics (Zou et al, 2022).…”

Section: Text Complexitymentioning

confidence: 93%

“…Second, the following syntactic metrics were calculated based on the work of Zou et al (2022), these metrics can be viewed as proxies to measure the impact of syntactic complexity on cognitive effort:…”

Section: Complexity Metricsmentioning

confidence: 99%

“…NND is the aggregation (i.e., sum, max or average) of nested noun distances. Following the research of Zou et al (2022) we used sum as the aggregating metric.…”

Section: Complexity Metricsmentioning

confidence: 99%

See 2 more Smart Citations

Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection

Singh,

De Bruyne,

De Clercq

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

Given the omnipresence of social media in our society, thoughts and opinions are being shared online in an unprecedented manner. This means that both positive and negative emotions can be equally and freely expressed. However, the negativity bias posits that human beings are inherently drawn to and more moved by negativity and, as a consequence, negative emotions get more traffic. Correspondingly, when writing about emotions this negativity bias could lead to expressions of negative emotions that are linguistically more complex. In this paper, we attempt to use readability and linguistic complexity metrics to better understand the manifestation of emotions on social media platforms like Reddit based on the widely-used GoEmotions dataset. We demonstrate that according to most metrics, negative emotions indeed tend to generate more complex text than positive emotions. In addition, we examine whether a higher complexity hampers the automatic identification of emotions. To answer this question, we finetuned three state-of-the-art transformers (BERT, RoBERTa, and SpanBERT) on the same emotion detection dataset. We demonstrate that these models often fail to predict emotions for the more complex texts. More advanced LLMs like RoBERTa and SpanBERT also fail to improve by significant margins on complex samples. This calls for a more nuanced interpretation of the emotion detection performance of transformer models. We make the automatically annotated data available for further research at hf.co/datasets/pranaydeeps/CAMEO.

show abstract

Section: Text Complexitymentioning

confidence: 93%

Section: Complexity Metricsmentioning

confidence: 99%

See 1 more Smart Citation

Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection

Singh,

De Bruyne,

De Clercq

et al. 2023

Findings of the Association for Computational Linguistics: EMNLP 2023

View full text Add to dashboard Cite

show abstract

“…For instance, the use of metrics of syntactic equivalence in the assessment of translation difficulty of sentences translated from English into Dutch was explored [21]. In a similar vein, the use of six AI-based syntactic complexity metrics in the performance of some sight interpreters in English-to-Chinese sight interpreting was examined [22]. The six metrics used in the study are incomplete dependency theory (IDT), dependency locality theory (DLT), combined IDT and DLT (IDT + DLT), left embeddedness (LE), nested nouns distance (NND), and bilingual complexity ratio (BRC).…”

Section: Literature Reviewmentioning

confidence: 99%

Translating Islamic Media Discourse from Arabic into English: An Analysis of Translation Process

A.S. Mohammed

2023

Translatology, Translation and Interpretation - Toward a New Scientific Endeavor

View full text Add to dashboard Cite

This study deals with the translation of Islamic media discourse from Arabic to English. It investigates both the process and product of translating Islamic media texts to determine the problems that translators encounter and the strategies that they employ to provide a communicative, target-reader friendly translation. This study uses an analytical and conceptual framework that stems from various taxonomies of translation strategies and cognitive translation studies. The translation process is investigated through the use of eye-tracking technology, keystroking, and user activity software. A parallel corpus of Islamic media texts is also analyzed to determine the most common strategies that are employed by translators of Islamic media. Analysis of the parallel corpus indicates that the translators have adopted various strategies to render Islamic media texts into English, including transference, functional equivalence, transposition, componential analysis, and foreignization, among others. The behavioral data generated by eye tracking, keystroking, Translog protocols, and user activity software show that the translation process involves a considerable number of fixations, pauses, insertions, deletions, and negotiations that may justify the decisions of the translators of a text.

show abstract

“…The annotators further classified each error as "Major" or "Minor", depending on its seriousness. The error annotation interface for the annotators in this experiment is described in Zou et al (2021). Once an error was discovered, annotators were requested to perform word-level alignment between the erroneous TT and their associated ST before assigning the error to the alignment group (AG).…”

Section: Manual Assessment Of Simtxt Outputmentioning

confidence: 99%

Patterns of Attention and Quality in English-Chinese Simultaneous Interpreting with Text

Zou¹,

Carl²,

Feng³

2022

IJCETI

View full text Add to dashboard Cite

Simultaneous interpreting with text (SIMTXT) is generally seen as being cognitively more demanding than either simultaneous interpreting (SI) or sight translation (STT). However, little research has been done on how interpreters allocate their attention proportionally during SIMTXT between their auditory and visual attention. This study examines attention patterns during SIMTXT and investigates the relationship between the interpreters’ attention patterns and the quality of the interpreting output. Nine professional interpreters were recruited for SIMTXT, interpreting six English STs into Chinese. The interpreters listened to the audio input while the transcribed text of the audio input was displayed on the screen by Translog-II. The eye movements were recorded by an eye-tracker (Tobii TX300) while their interpretations were recorded by Audacity. The gaze and the spoken data (source and target) were synchronized on a word-level and aligned with the final STs and TTs. We explored and categorized the visual and auditory attention patterns based on their ear-voice span (EVS), eye-voice span (IVS), and ear-eye span (EIS) and found three types of attention patterns in our data, namely, ear-dominant (ED), eye-dominant (ID), and ear-eye-balanced (EIB). In addition, the interpreting output was annotated by three professional interpreters based on an error taxonomy adapted from Multidimensional Quality Metrics (MQM). We found that the EIB interpreters produced the lowest translation quality in terms of total error numbers and accuracy, followed by ED, and then ID interpreters. ED interpreters produced the highest translation quality in terms of fluency, followed by EIB, and then ID interpreters.

show abstract

AI-Based Syntactic Complexity Metrics and Sight Interpreting Performance

Cited by 5 publications

References 15 publications

Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection

Misery Loves Complexity: Exploring Linguistic Complexity in the Context of Emotion Detection

Translating Islamic Media Discourse from Arabic into English: An Analysis of Translation Process

Patterns of Attention and Quality in English-Chinese Simultaneous Interpreting with Text

Contact Info

Product

Resources

About