Related Words Research Articles

A syntactic evidence network model for fact verification

In natural language processing, fact verification is a very challenging task, which requires retrieving multiple evidence sentences from a reliable corpus to verify the authenticity of a claim. Although most of the current deep learning methods use the attention mechanism for fact verification, they have not considered imposing attentional constraints on important related words in the claim and evidence sentences, resulting in inaccurate attention for some irrelevant words. In this paper, we propose a syntactic evidence network (SENet) model which incorporates entity keywords, syntactic information and sentence attention for fact verification. The SENet model extracts entity keywords from claim and evidence sentences, and uses a pre-trained syntactic dependency parser to extract the corresponding syntactic sentence structures and incorporates the extracted syntactic information into the attention mechanism for language-driven word representation. In addition, the sentence attention mechanism is applied to obtain a richer semantic representation. We have conducted experiments on the FEVER and UKP Snopes datasets for performance evaluation. Our SENet model has achieved 78.69% in Label Accuracy and 75.63% in FEVER Score on the FEVER dataset. In addition, our SENet model also has achieved 65.0% in precision and 61.2% in macro F1 on the UKP Snopes dataset. The experimental results have shown that our proposed SENet model has outperformed the baseline models and achieved the state-of-the-art performance for fact verification.

Lexical-semantic activation in French-Spanish or French-English bilingual toddlers: An event-related potential (ERP) investigation

Both behavioral and neurophysiological evidence shows that lexical-semantic organization emerges by two years in monolingual children. Research in bilingual children is more scarce, and there is only a limited amount of neurophysiological evidence of the effect of language dominance on lexical-semantic activation. In the present event-related potential (ERP) study, we investigated whether bilingual French-Spanish and French-English learning 24- to 30-month-olds activate semantic relations between words similarly in their both languages, and whether the priming effects are similar in children learning two different language pairs. Participants were presented with related and unrelated dominant and non-dominant language word pairs in a within-language lexical-semantic priming paradigm. The amplitudes of N400 were modulated by trial type, language dominance and language group. A language-independent priming effect - more pronounced N400 amplitudes for unrelated than for related target words - was found in the group of toddlers learning French and Spanish. In the group of toddlers learning French and English, a priming effect was observed only in their non-dominant language. Our results propose that the language pair may contribute to lexical-semantic facilitation in priming tasks during early childhood.

Middle-schoolers' reading and lexical-semantic processing depth in response to digital and print media: An N400 study.

We report the first use of ERP measures to identify text engagement differences when reading digitally or in print. Depth of semantic encoding is key for reading comprehension, and we predicted that deeper reading of expository texts would facilitate stronger associations with subsequently-presented related words, resulting in enhanced N400 responses to unrelated probe words and a graded attenuation of the N400 to related and moderately related words. In contrast, shallow reading would produce weaker associations between probe words and text passages, resulting in enhanced N400 responses to both moderately related and unrelated words, and an attenuated response to related words. Behavioral research has shown deeper semantic encoding of text from paper than from a screen. Hence, we predicted that the N400 would index deeper reading of text passages that were presented in print, and shallower reading of texts presented digitally. Middle-school students (n = 59) read passages in digital and print formats and high-density EEG was recorded while participants completed single-word semantic judgment tasks after each passage. Following digital text presentation, the N400 response pattern to moderately-related words indicated shallow reading, tracking with responses to words that were unrelated to the text. Following print reading, the N400 responses to moderately-related words patterned instead with responses to related words, interpreted as an index of deeper reading. These findings provide evidence of differences in brain responses to texts presented in print and digital media, including deeper semantic encoding for print than digital texts.

Making Judgments of Learning Either Enhances or Impairs Memory: Evidence From 17 Experiments With Related and Unrelated Word Pairs

Published studies found that predicting one’s future memory during learning (judgments of learning, JOLs) consistently improved cued-recall performance for related word pairs. In contrast, making JOLs had inconsistent effects on memory for unrelated pairs, with most studies finding null effects and some finding detrimental effects. This study reports data from 17 experiments in which participants either made or did not make JOLs for related and unrelated word pairs in their everyday language. Making JOLs increased the difference in memory performance between related and unrelated pairs in every experiment. Although almost all experiments showed numerically positive JOL reactivity for related pairs and numerically negative JOL reactivity for unrelated pairs, either effect was reliable in just half of the experiments. Small-scale meta-analyses revealed small-to-moderate positive reactivity for related pairs and small-to-moderate negative reactivity for unrelated pairs. Language of word pairs (German, Hebrew, or English) moderated positive reactivity and experimental setting (controlled or unsupervised online) moderated negative reactivity, but none of the other moderators we examined—presence of an additional pair type, study time, total number of pairs—impacted reactivity. Experiments that showed positive reactivity for related pairs tended not to show negative reactivity for unrelated pairs, and vice versa. Overall, these findings indicate that negative JOL reactivity for unrelated pairs is similarly large and robust as positive reactivity for related pairs. They favor the cue-strengthening hypothesis with dual-task costs over other accounts and raise the practically relevant possibility that monitoring could have detrimental effects on learning in educational settings. All data are freely available online.

Bilingual Translation in Spiderman the Movie Storybook

The language transfer can be seen in the lexical choices in a translation. A translator may have different considerations when applying its use in a text since adaptation of context is an essential part to be employed through the translation. Problems with transferring the source language into the target language certainly have an effect on delivering messages within a text. This study aims to analyze the lexical usage in the bilingual translation of the movie storybook Spiderman and the strategies in the translation, including their meaning. The qualitative method used DeepL to assist in the analysis of differences that might come into the translation. In some cases, words translated from English into Indonesian were found non-equivalent, however, the rest of them were understandable. Translation by a more general word strategies were also used to be placed in the story rather than using the specific words and vice versa, paraphrasing using a related word, omission, and neutral words or less expressive words found to correspond the meaning. The equivalence of transferring the source language (SL) into the target language (TL) can be achieved when optional words are found in the translation in which meaning is also taken for consideration within the human translation. The product of translation was presented bilingually to analyze the differences in meaning between the two languages’ context of use. These showed DeepL as one of the machine translations with translation memories using Artificial Intelligence technology which compares the differences and similarities of sense in translation study. They can work together for the analysis to understand the use and results. Keywords: equivalence, non-equivalence, lexical, strategy

A Meta-Analysis of Anomia Treatment in Bilingual Aphasia: Within- and Cross-Language Generalization and Predictors of the Treatment Outcomes.

The present meta-analysis investigated the efficacy of anomia treatment in bilingual and multilingual persons with aphasia (BPWAs) by assessing the magnitudes of six anomia treatment outcomes. Three of the treatment outcomes pertained to the "trained language": improvement of trained words (treatment effect [TE]), within-language generalization of semantically related untrained words (WLG-Related), and within-language generalization of unrelated words (WLG-Unrelated). Three treatment outcomes were for the "untrained language": improvement of translations of the trained words (cross-language generalization of trained words [CLG-Tx]), cross-language generalization of semantically related untrained words (CLG-Related), and cross-language generalization of unrelated untrained words (CLG-Unrelated). This study also examined participant- and treatment-related predictors of these treatment outcomes. This study is registered in the International Prospective Register of Systematic Reviews (PROSPERO) under the number CRD42023418147. Nine electronic databases were searched to identify word retrieval treatment studies of poststroke BPWAs of at least 6 months postonset. Pre- and posttreatment single-word naming scores were extracted for each eligible participant and used to calculate effect sizes (within-case Cohen's d) of the six treatment outcomes. Random-effects meta-analyses were conducted to assess weighted mean effect sizes of the treatment outcomes across studies. Multiple linear regression analyses were used to examine the effects of participant-related variables (pretreatment single-word naming and comprehension representing poststroke lexical processing abilities) and treatment-related variables (type, language, and duration). The methodological quality of eligible studies and the risk of bias in this meta-analysis were assessed. A total of 17 published studies with 39 BPWAs were included in the meta-analysis. The methodological quality of the included studies ranged from fair (n = 4) to good (n = 13). Anomia treatment produced a medium effect size for TE (M = 8.36) and marginally small effect sizes for WLG-Related (M = 1.63), WLG-Unrelated (M = 0.68), and CLG-Tx (M = 1.56). Effect sizes were nonsignificant for CLG-Related and CLG-Unrelated. TE was significantly larger than the other five types of treatment outcomes. TE and WLG-Related effect sizes were larger for BPWAs with milder comprehension or naming impairments and for treatments of longer duration. WLG-Unrelated was larger when BPWAs received phonological treatment than semantic and mixed treatments. The overall risk of bias in the meta-analysis was low with a potential risk of bias present in the study identification process. Current anomia treatment practices for bilingual speakers are efficacious in improving trained items but produce marginally small within-language generalization and cross-language generalization to translations of the trained items. These results highlight the need to provide treatment in each language of BPWAs and/or investigate other approaches to promote cross-language generalization. Furthermore, anomia treatment outcomes are influenced by BPWAs' poststroke single-word naming and comprehension abilities as well as treatment duration and the provision of phonological treatment. https://doi.org/10.23641/asha.25595712.

PRINCIPLES OF COMPILING A GERMAN-UKRAINIAN DICTIONARY OF PARONYMS

This article addresses the principles of selecting paronyms and the methodology for compiling an electronic German-Ukrainian dictionary of paronyms using a corpus-based approach and systematic analysis to identify similarities and differences in the use of paronyms in speech. The author presents a concept of paronymy as a systemic phenomenon involving etymologically related words with similar morphemic structures but different meanings. By conducting a comparative analysis of collocations, the study identified 1,836 paronyms organized into paronymic series, each comprising two to nine components. The article examines existing dictionaries of paronyms and highlights their lexicographic shortcomings, such as excessive normative/prescriptive orientation, inadequate development of dictionary entries from both linguistic and didactic perspectives, and a lack of rigorous evaluations of modern data to inform dictionary compilation. The ideal paronym dictionary should rely on corpus data, providing reliable information with generalizations about individual word uses within conventionalized language usage. Corpus-based tools, such as those used for contrasting near-synonyms, systematically identify similarities and differences between paronyms in a contrastive manner. This process determines the degree of semantic closeness among expressions with similar contextual usage by comparing their immediate collocational patterns. The recorded semantic closeness or distance between paronyms is based on their contextual overlap in usage. The lexicographic processing of the selected paronyms was carried out with TshwaneLex, a computer-based dictionary creation software that allows lexicographers to develop dictionaries for any language without requiring advanced IT skills. TshwaneLex offers various features such as automatic cross-reference tracking, advanced dictionary comparison/merge functionality, and support for all world languages through full Unicode compatibility. The lemma editing process uses a tree-based interface to represent the hierarchical structure of the lemma, including different meanings, submeanings, word formations, multi-word units, examples of usage, cross-references, and more. Furthermore, the input/output architecture is designed to support the development of additional interfaces for various data sources as add-ons or plug-ins, facilitating other output formats. The article notes that a PHP-based software module is currently available for hosting the TshwaneLex dictionary on the internet, providing an accessible platform for electronic dictionary publication.

Words and Roots – Polysemy and Allosemy – Communication and Language

AbstractMost substantive (content-bearing) words are polysemous, but polysemy is cross-categorial; for instance, the lexical forms ‘stone’ and ‘front’ are associated with families of interrelated senses and these senses are spread across their manifestations as three words, noun, verb and adjective. So, the ultimate unit underpinning polysemy is not a word but the categoryless root of the related words, which must, in some sense, track the interrelated families of senses. The main topic of this paper is the vexed question of the meaning of roots and the backdrop is a view of words as delineated syntactic domains which allow assignment of atomic content (non-compositional meaning), and whose actual meanings are, in the first instance, pragmatically inferred in the throes of communication, some of them subsequently becoming established, so stored in a lexicon and directly retrieved in comprehension. Three different positions on the meanings of roots are outlined, and their merits and shortcomings are discussed: (a) inherent underspecified meanings; (b) meanings conditioned by grammatical context (allosemy); (c) meaninglessness. I argue that, overall, the current state of the evidence favours the third position: roots are categoryless, meaningless (perhaps phonological) indices.

The Effects of Problematic Internet Use and Emotional Connotation on Internet Slang Processing: Evidence from a Lexical Decision Task.

Young people use slang for identifying themselves with a particular social group, gaining social recognition and respect from that group, and expressing their emotional state. One feature of Internet slang is its active use by youth in online communication, which, under certain conditions, may cause problematic Internet use (PIU). We conducted two studies in young Russian speakers (n1 = 115, n2 = 106). In study 1, participants were asked to rate a set of slang and common words using Self-Assessment Manikin. The study revealed that the most reliable predictor of higher emotional ratings was word familiarity. There were no significant effects of slang vs. common words or word frequency. In study 2, we used a dual lexical decision task to reveal the effects of word characteristics and propensity for PIU on reaction time (RT) for Internet slang words in pairs with semantically related vs. unrelated common words. Study 2 did not reveal any significant semantic priming effect. Word frequency was a significant predictor of lexical decision facilitation. Common, but not slang, word valence and dominance significantly affected RT in the opposite direction. Individuals with higher cognitive preoccupation with the Internet responded significantly faster, while those more likely to use online communication for mood regulation responded significantly slower to the stimuli. Apparently, on explicit and implicit levels, in-depth knowledge of Internet slang can be one the PIU markers. The results are discussed in line with Davis' approach to determining the general pathological Internet use.

Unveiling Gendered 3MT Presentations: A Corpus-based Analysis of Evaluative Language Important

Originating at the University of Queensland (UQ), the 3 Minute Thesis (3MT) has evolved into a pivotal form of academic communication. The linguistic resource important is often employed by presenters to underscore the significance of specific concepts, discoveries, or viewpoints, revealing their epistemic and evaluative stance. By constructing a corpus encompassing all award-winning 3MT presentations from UQ, with a particular focus on the usage of important and its related words, we found that (1) there was no statistically significant gender-based discrepancy in the frequency of employing important; (2) male presenters employed important in isolation, while their female counterparts opted for more expansive terms such as significant or crucial, encompassing a broader range of interpretations; (3) in terms of contexts, female speakers tended to extend their use of these terms beyond academic contexts, linking them with lived experiences more often than male presenters. The observed patterns related to how male and female presenters convey the importance of their research can enhance our comprehension of the subtle ways in which language mirrors and perpetuates gendered norms and expectations. More importantly, they may facilitate progress towards fostering more inclusive and equitable academic environments.

Empowering educational policy management: understanding student emotions and sentiments

PurposeUsing sentiment analysis (SA), this study aims to examine the impact of COVID-19 on mental health and virtual learning experiences among 1,125 students at a public Argentinean faculty.Design/methodology/approachA study was conducted during the COVID-19 pandemic, surveying 1,125 students to gather their opinions. The survey data was analysed using text mining tools and SA. SA was used to extract the students’ emotions, views and feelings computationally and identify co-occurrences and patterns in related words. The study also examines educational policies implemented after the pandemic.FindingsThe prevalent emotions expressed in the comments were trust, sadness, anticipation and fear. A combination of trust and fear resulted in submission. Negative comments often included the words “virtual”, “virtual classroom”, “virtual classes” and “professor”. Two significant issues were identified: teachers’ inexperience with virtual classes and inadequate server infrastructure, leading to frequent crashes. The most effective educational policies addressed vital issues related to the “virtual classroom”.Practical implicationsText mining and SA are valuable tools for decision-making during uncertain times, such as the COVID-19 pandemic. They can also provide insights to recover quality assurance processes at universities impacted by health concerns or external shocks.Originality/valueThe paper makes two main contributions: it conducts a SA to gain insights from comments and analyses the relationship between emotions and sentiments to identify optimal educational policies. The study pioneers exploring the link between emotions, policies and the pandemic at a public university in Argentina. This area of research still needs to be explored.

Adaptive lexical processing of semantic competitors extends to alternative names: Evidence from blocked-cyclic picture naming.

Naming a picture (e.g., "duck") in the context of semantically related pictures (e.g., "eagle," "stork," "parrot") takes longer than naming it in the context of unrelated pictures (e.g., "knave," "toast," "atlas"). Adaptive models of word production attribute this semantic interference effect in blocked-cyclic naming (BCN) to an adaptive mechanism that makes competitor words, (e.g., the semantically related word "eagle" for the target word "duck") which are activated but not selected for production, less accessible for future retrieval. Results from a recent picture-word-interference study, however, suggested that alternative names (e.g., "bird" for "duck") might be exempt from this mechanism, challenging adaptive lexical processing as a general mechanism. We tested whether converging evidence is obtained in BCN. In Experiment 1, we embedded pictures responded to with alternative (category) names (e.g., "bird") into contexts composed of pictures responded to with specific (exemplar) names (e.g., "duck," "eagle," "stork," and "parrot"). If alternative names are exempt from adaptive lexical processing, interference in the homogeneous context should be found for specific name items but not for alternative name items. In contrast to this prediction, there was similar-sized interference for both types of items. In Experiment 2, we replaced the alternative name items with unrelated items. For these items, interference was largely diminished, ruling out that the effect found in Experiment 1 is a general set effect. Overall, our data suggest that alternative names are not special with respect to adaptive lexical processing.

The Emotion Analysis of Indian Political Tweets using Machine Learning

In this day and age web-based entertainment is a major region for information examination and exploration work. For Feeling Examination, I select Tweeter handle. I use Tweepy for getting to tweeter information. I perform opinion examination on Indian Political information. I got 117545 tweets of 2019 Indian Political race. I use SVM (Backing Vector Machine) Classifier for feeling Examination. Feeling assessment oversees recognizing and portraying evaluations or sentiments conveyed in source message. Electronic diversion is creating an enormous proportion of feeling rich data as tweets, sees, blog sections, etc. Feeling examination of this client made data is especially useful in knowing the appraisal of the gathering. Twitter feeling assessment is problematic stood out from general assessment examination on account of the presence of work related conversation words and erroneous spellings. The most outrageous limitation of characters that are allowed in Twitter is 140. Data base philosophy and AI approach are the two frameworks used for separating suppositions from the text. In this paper, we endeavor to analyze the twitter posts about electronic things like mobiles, workstations, etc using AI approach.

Automated Marking System for Essay Questions

The stress of marking assessment scripts of many candidates often results in fatigue that could lead to low productivity and reduced consistency. In most cases, candidates use words, phrases and sentences that are synonyms or related in meaning to those stated in the marking scheme, however, examiners rely solely on the exact words specified in the marking scheme. This often leads to inconsistent grading and in most cases, candidates are disadvantaged. This study seeks to address these inconsistencies during assessment by evaluating the marked answer scripts and the marking scheme of Introduction to File Processing (CSC 221) from the Department of Computer Science, University of Uyo, Nigeria. These were collected and used with the Microsoft Research Paraphrase (MSRP) corpus. After preprocessing the datasets, they were subjected to Logistic Regression (LR), a machine learning technique where the semantic similarity of the answers of the candidates was measured in relation to the marking scheme of the examiner using the MSRP corpus model earlier trained on the Term Frequency-Inverse Document Frequency (TF-IDF) vectorization. Results of the experiment show a strong correlation coefficient of 0.89 and a Mean Relative Error (MRE) of 0.59 compared with the scores awarded by the human marker (examiner). Analysis of the error indicates that block marks were assigned to answers in the marking scheme while the automated marking system breaks the block marks into chunks based on phrases both in the marking scheme and the candidates’ answers. It also shows that some semantically related words were ignored by the examiner.

A study of the semantic preference and semantic prosody associated with the denominations of aging people

This paper studies the representation and conceptualization of aging people depending on the way they are referred to through a series of near synonyms: old / older / elderly / middle-aged / mature + N. A specific focus will be laid on the notions of "semantic preference," i.e., "the tendency for an item to co-occur with a set of semantically related words," and "semantic prosody," a.k.a. "evaluative, pragmatic, emotive or attitudinal prosody," to account for the evaluative aspect linked to certain denominations. This paper compares semantic preference and semantic prosody with five near-synonymous adjectives. It shows that denominations of aging people acquire a specific evaluation – a negative or positive semantic prosody – through their semantic preference, in that specific case the different attributive adjectives used in prenominal position of the sequences under scrutiny. This study addresses the question of the frequent negative axiology linked to the discourses on old age and aging through the notion of semantic prosody and will mostly examine the potential gender differences in the denominations of aging people.

Reduced false memory in the second language of Turkish-English bilinguals

In the Deese-Roediger-McDermott (DRM) paradigm, studying a list of semantically related words leads to false memory for the critical non-studied word that is related to all the words in the study list. Previous studies questioning whether bilinguals are more prone to false memory in their first language (L1) or second language (L2) in the DRM paradigm revealed mixed results. The present study investigated the same question with Turkish-English bilinguals. The revised hierarchical model proposes that the link between the lexicon and the semantic system is weaker in L2 than in L1, suggesting that false memory in the DRM paradigm that relies on semantic relatedness would be higher in L1 than in L2. Furthermore, previous studies showed that L2 is more resistant to errors in decision-making when the two languages are dissimilar, but not when they are similar, and Turkish and English are historically distant and typologically dissimilar languages. We tested Turkish-English bilingual participants whose L1 is Turkish with Turkish and English DRM word lists that had similar prior norms for generating false recognition. In the recognition test, some of the studied items and the critical non-studied items were presented and participants identified the studied items. False recognition for the critical non-studied items was lower and correct recognition for studied items was higher in L2 than in L1. The results suggest that L2 is more resistant to false memory due to its weaker lexicon and semantic system associations, at least when the two languages spoken by the bilingual are dissimilar.

The impact of spatial and verbal working memory load on semantic relatedness judgements.

Studies using a relatedness judgement task have found differences between prime-target word pairs that vary in the degree of semantic relatedness. However, the influence of working memory load on semantic processing in this task and the role of the type of working memory task have not yet been investigated. The present study therefore investigated for the first time the effect of working memory load (low vs. high) and working memory type (verbal vs. spatial) on semantic relatedness judgements. Semantically strongly related (e.g., hip - KNEE), weakly related (e.g., muscle - KNEE) and unrelated (e.g., office - KNEE) Polish word pairs were presented in an experiment involving a dual working memory and semantic relatedness task. The data revealed that, relative to semantically unrelated word pairs, responses were faster for strongly related pairs but slower for weakly related pairs. Importantly, the verbal working memory task decreased facilitation for strongly related pairs and increased inhibition for weakly related pairs relative to the spatial working memory task. Furthermore, working memory load impacted only weakly related pairs in the verbal but not in the spatial working memory task. These results show that working memory type and load influence semantic relatedness judgements, but the direction and size of the impact depend on the strength of semantic relations.

Derivation of Key Attributes and Clusters of Korean Taekwondo Policies and Systems on Social Media: Comparative Big Data Analysis Based on Regime Change

This study analyzed the key attributes and clusters of policies and systems of Korean Taekwondo based on Park Geun-Hye and Moon Jae-in’s regime changes using big data, including 8,030 data points on social media over a period of 8 years. Frequency analysis, term frequency inverse document frequency analysis, and degree centrality analysis were performed using TEXTOM 4.5. In addition, a convergent and correlation analysis was performed using UCINET 6 to visualize related words and analyze clusters. Frequency, term frequency inverse document frequency, and degree centrality analyses revealed 30 high-order terms. A convergence of iterated correlations analysis of the derived words identified three common clusters: improvement, problem, and support. In addition, there were two distinct clusters in the Park Geun-hye government (research and sports policy), and three in the Moon Jae-in government (interview, thoughts, and athlete). The Korean government attempted to improve and address problems related to policies and systems for the development of Taekwondo. However, as Taekwondo policies and systems were implemented from different perspectives by each regime, systematic policies and systems could not be implemented.

A Comparison of Expressive Vocabulary in Spontaneous Language Sample and Parent Checklist of Young Children with Autism Spectrum Disorder

Objectives: This study aims to examine the differences in vocabulary abilities in children with autism spectrum disorder (ASD) using a parent checklist and spontaneous language analysis.Methods: Participants included 10 children with ASD, aged 18-30 months, and 10 typically developing children matched for language age. Data were collected using the Korean version of the MacArthur-Bates Communicative Development Inventory (K M-B CDI) vocabulary checklist and spontaneous language analysis. The total number of vocabulary and the number of 5 word classes of vocabulary were measured and a repeated measures analysis of variance (ANOVA) and Pearson correlation analysis was performed to compare the two groups.Results: Children with ASD demonstrated a higher total vocabulary according to the parent checklist compared to typically developing children. However, there was no significant difference in the results of spontaneous language analysis. While the parent checklist indicated higher vocabulary across all five word classes for ASD children, spontaneous language analysis revealed that they predominantly used nouns and verbs, with fewer occurrences of determiners, independent words, and relational words. Moreover, the vocabulary overlap between the two methods was lower in the ASD group compared to the typically developing group. Conclusion: This study highlights the challenges faced by children with ASD in appropriately using their known vocabulary. It emphasizes the importance of considering not only the parent checklist but also analyzing spontaneously used vocabulary in real communication situations. This integrated approach may provide valuable insights for developing personalized interventions and therapeutic strategies for language development in children with ASD.

Measuring Alliance and Symptom Severity in Psychotherapy Transcripts Using Bert Topic Modeling

We aim to use topic modeling, an approach for discovering clusters of related words (“topics”), to predict symptom severity and therapeutic alliance in psychotherapy transcripts, while also identifying the most important topics and overarching themes for prediction. We analyzed 552 psychotherapy transcripts from 124 patients. Using BERTopic (Grootendorst, 2022), we extracted 250 topics each for patient and therapist speech. These topics were used to predict symptom severity and alliance with various competing machine-learning methods. Sensitivity analyses were calculated for a model based on 50 topics, LDA-based topic modeling, and a bigram model. Additionally, we grouped topics into themes using qualitative analysis and identified key topics and themes with eXplainable Artificial Intelligence (XAI). Symptom severity could be predicted with highest accuracy by patient topics (r\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$r$$\\end{document}=0.45, 95%-CI 0.40, 0.51), whereas alliance was better predicted by therapist topics (r\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$r$$\\end{document}=0.20, 95%-CI 0.16, 0.24). Drivers for symptom severity were themes related to health and negative experiences. Lower alliance was correlated with various themes, especially psychotherapy framework, income, and everyday life. This analysis shows the potential of using topic modeling in psychotherapy research allowing to predict several treatment-relevant metrics with reasonable accuracy. Further, the use of XAI allows for an analysis of the individual predictive value of topics and themes. Limitations entail heterogeneity across different topic modeling hyperparameters and a relatively small sample size.