Word Pairs Research Articles

Improved Hypernymy Detection Algorithm Based on Heterogeneous Graph Neural Network

Concept mapping is a knowledge representation method used to represent and understand concepts, entities, and the relationships between them, which are referred to as hyponymy–hypernymy semantic relations. These relations are primarily used to describe the hierarchical and categorical relationships between different concepts or entities. The detection of hyponymy–hypernymy semantic relations is an important task in the field of natural language processing, crucial to many downstream tasks such as information extraction, automatic reasoning, and personalized recommendations. These tasks often require understanding the semantic relations between concepts or entities in text for more accurate analysis and reasoning. Currently, algorithms for identifying and detecting hyponymy–hypernymy semantic relations face two main challenges: first, candidate hyponymy–hypernymy relation tuples do not exist in the same contextual sentence, failing to meet the co-occurrence requirement; second, distributed algorithms have issues with lexical memory. To address these issues, this paper proposes an improved algorithm for detecting hyponymy–hypernymy relations based on heterogeneous graph neural networks, aiming to detect hyponymy–hypernymy relations in various candidate word sets and then construct a hierarchical system. To meet the co-occurrence requirement of candidate word pairs, an open-source large model is utilized to generate contextual sentences for the candidate pairs. Sub-word features are adopted to capture the intrinsic semantic connections between nested phrases, thus alleviating the issue of reversed predictions for nested phrases. The representation of relation nodes is modified by encoding relation definitions through a pre-trained model, enabling the model to understand the semantic relations between concept nodes. To address the problem of overfitting in traditional graph attention networks, the calculation order of adjacency node aggregation is changed in the heterogeneous graph to capture dynamic attention features. In addition, a pipeline for hierarchical system construction is designed and implemented, combining the divide-and-conquer approach with loop detection algorithms. Compared to baseline metrics, the proposed method achieves a 4.14% improvement in accuracy and a 0.62% increase in F1 score on the EVALution dataset; a 4.89% increase in accuracy and a 0.71% improvement in F1 score on the Bansal dataset; and a 1.05% increase in accuracy, a 2.21% increase in recall, and a 1.79% improvement in F1 score on a self-annotated Chinese dictionary dataset.

Korean Hangul is more robust to a serial bottleneck: Co-occurring and semantically related Korean words can be processed in parallel.

Can readers process multiple words simultaneously, and are there cultural differences in attentional bottleneck in lexical processing? To answer these questions, we asked participants to view two words and categorize only one (single-task) or both words (dual-task), using Korean word pairs that frequently co-occur and are semantically related. We hypothesized that the coactivation of related words could facilitate lexical processing, and that the unique characteristics of Korean Hangul, such as its shallow orthographic depth and clear-cut syllabic boundaries, would enhance this effect. The results suggest that Korean Hangul is more robust to a serial processing bottleneck. Unlike the previous findings in English, independent or unrelated Korean word pairs were not processed in a strictly serial manner. Furthermore, the results for co-occurring and semantically related words supported the parallel processing model. Interestingly, for regularly co-occurring words, accuracy for one word increased if the participant was also correct about the other. These findings suggest that the relation between the words and the properties of writing systems should be considered in the long-standing serial versus parallel debate in reading research. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Familiar Music Reduces Mind Wandering and Boosts Behavioral Performance During Lexical Semantic Processing

Music has been shown to increase arousal and attention and even facilitate processing during non-musical tasks, including those related to speech and language functions. Mind wandering has been studied in many sustained attention tasks. Here, we investigated the intersection of these two phenomena: the role of mind wandering while listening to familiar/unfamiliar musical excerpts, and its effects on concurrent linguistic processing. We hypothesized that familiar music would be less distracting than unfamiliar music, causing less mind wandering, and consequently benefit concurrent speech perception. Participants (N = 96 young adults) performed a lexical-semantic congruity task where they judged the relatedness of visually presented word pairs while listening to non-vocal classical music (familiar or unfamiliar orchestral pieces), or a non-music environmental sound clip (control) played in the background. Mind wandering episodes were probed intermittently during the task by explicitly asking listeners if their mind was wandering in that moment. The primary outcome was accuracy and reactions times measured during the lexical-semantic judgment task across the three background music conditions (familiar, unfamiliar, and control). We found that listening to familiar music, relative to unfamiliar music or environmental noise, was associated with faster lexical-semantic decisions and a lower incidence of mind wandering. Mind wandering frequency was similar when performing the task when listening to familiar music and control environmental sounds. We infer that familiar music increases task enjoyment, reduces mind wandering, and promotes more rapid lexical access during concurrent lexical processing, by modulating task-related attentional resources. The implications of using music as an aid during academic study and cognitive tasks are discussed.

The role of phonetic overlap for speaker discrimination.

Linguistic information influences processing of speaker information in a multitude of ways, whether this arises from the listener's familiarity with the language or dialect spoken, or existing linguistic relationships between spoken words. Specifically, phonological and semantic relationships between spoken words have been observed to influence a listeners' ability to discriminate voices. This study aims to develop our understanding of how different kinds of linguistic relationships, namely, phonetic relationships, influence the processing of speaker information. We conducted two experiments, a voice discrimination task and a voice similarity rating task, in which listeners were presented with pairs of speakers producing two words with various degrees of phonetic overlap. On the whole, higher quantities of phonetic overlap corresponded to higher speaker discrimination performance and higher similarity scores; however, the type of the phonetic overlap also mattered. Overlapping vowel segments showed substantial utility, while overlap of the phonological rhyme alone substantially lower performance. Results from this condition suggest that a phonological relationship within the word pair can interfere with otherwise increased quantities of phonetic overlap. These findings highlight the salience of the phonological rhyme in voice processing, as well as the overall impact of phonetic overlap.

Integrating design and system approaches for analyzing road traffic collisions in low-income settings.

Integrating design and system approaches for analyzing road traffic collisions in low-income settings.

Involvement of episodic memory in language comprehension: Naturalistic comprehension pushes unrelated words closer in semantic space for at least 12h.

Recent experience with a word significantly influences its subsequent interpretation. For instance, encountering bank in a river-related context biases future interpretations toward 'side of a river' (vs. 'financial bank'). To explain this effect, the episodic context account posits that episodic memory helps bind word meanings in the language input, creating a temporary, context-specific representation that can bias subsequent lexical interpretation. This account predicts that even unrelated words would be linked together in episodic memory, potentially altering their interpretation. In Experiments 1-3, participants read unrelated word pairs (e.g., sword-microwave, privacy-export) embedded in meaningful sentences, then completed a speeded relatedness judgement task after delays of 5min, 20min, or 12h (including sleep). Results showed that sentence exposure increased the likelihood of the unrelated pairs being judged as related-a robust effect observed across all delay intervals. Experiment 4 showed that this exposure effect was abolished when words in a target pair were read in separate sentences, suggesting that the exposure effect may be dependent on lexical co-occurrence. Experiment 5, also with a 12-h delay (including sleep), additionally used an innovative word arrangement task to assess word relatedness without presenting the target pairs simultaneously or successively. In line with relatedness judgement, sentence exposure pushed the unrelated words closer in semantic space. Overall, our findings suggest that a context-specific representation, supported by episodic memory, is generated during language comprehension, and in turn, these representations can influence lexical interpretation for at least 12h and across different linguistic circumstances. We argue that these representations endow the mental lexicon with the efficiency to deal with word burstiness and the dynamic nature of language.

병렬 말뭉치 기반 L1(한국어) 화자를 위한 L2(한국수어) 교육 연어 탐색 방법 연구

Objectives This study explores a method for identifying collocations in Korean Sign Language (KSL) using a Korean-KSL parallel corpus and examines its applicability to L1 (Korean) speakers learning KSL as an L2. Methods The Korean-KSL parallel corpus was preprocessed using Python, and high-frequency word pairs were extracted based on N-grams using #LancsBox. To systematically outline the collocation exploration process, the study focused on the most frequent preceding word, {돕다} (to help), and applied Log Dice and Delta P metrics to assess collocational strength, leading to the selection of collocation candidates. Results The application of the proposed collocation exploration method based on a parallel corpus enabled the identification of collocation candidates at the two-word level. Additionally, certain expressions that exhibited weak collocation potential at the two-word level were confirmed as collocations when expanded to three-word sequences. Moreover, the use of Log Dice and Delta P in the word-unit analysis quantitatively measured differences in word order between Korean and KSL, highlighting the significance of bidirectional collocational strength analysis. Conclusions The study confirmed the validity of the proposed collocation exploration method and demonstrated that word-unit analysis incorporating collocational strength is essential for identifying collocation candidates, beyond simple frequency-based approaches. Furthermore, the findings empirically show that collocation learning contributes not only to vocabulary acquisition but also to reducing word order errors in KSL education. Future research will expand the range of central words and genres to enhance the collocation list and establish a foundation for developing more refined educational materials for KSL learners.

Lexical search and social reasoning jointly explain communication in associative reference games.

Effective linguistic communication depends upon many different cognitive processes working together in concert. Yet, our computational models of these processes are often developed in isolation, without considering how these processes fit together. In this work, we study a simplified variant of the popular board game Codenames, which highlights the integration of two important processes: (1) lexical retrieval and (2) pragmatic reasoning. In this task, speakers must generate a "clue" word from their full lexicon that allows their partner to select a pair of target words from a context of distractors. In Experiment 1, we evaluate a suite of different models on an existing corpus of production data, finding that models with both semantic search and pragmatic reasoning components significantly outperform ablated variants. Experiment 2 elicits targeted endorsements for a particularly diagnostic set of clues, providing further evidence for the pragmatic selection component in isolation. Finally, Experiments 3 and 4 elicit lists of clues that speakers are considering for a target pair in the absence and presence of the board context, respectively, providing further evidence for the way context is infused into semantic search. Taken together, our findings shed new light on the way that retrieval processes (generating promising candidates) and selection processes (evaluating the likely communicative success of these candidates) must work in tandem to support effective communication. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Evaluation of color impression using abstract and color-related semantic words and selection of suitable colors that express the word impression of these words

In this study, we investigate two hypotheses for determining color impression. Color impression can be directly determined by color appearance, especially in the case of the paired comparison, or by the combination of semantic impressions defined by words [especially in the semantic differential (SD) method]. We measured the color impression of 12 chromatic colors, black, gray, and white, using seven paired semantic words related to color properties and impressions and eight paired abstract words in the SD method. We also asked 10 observers to select the appropriate colors from the 15 colors in the paired comparison method for each word. Principal component analysis was applied to the data. In the results of the SD method, the coordinates of the colors on the space of the first and second principal component axes are basically spread as the hue circle but are more clustered than in the results of the paired comparison because the semantic words at higher grading points of the SD method are similar between neighboring colors. In the results of the paired comparison, the coordinates of the colors as loading values maintained the hue circle, and the color selection rates for the words in the paired comparison showed strong correlations with one of the color properties (L∗,a∗,b∗, and C∗), except for Murky, as do the grading points for the colors in the SD method, showing that color appearance can mostly determine color impression using words directly and relatively related to colors (except Vivid and Murky), but not using some of the abstract words (such as Clean, Magnificent, and Deserted). The direct comparison between the two methods suggests that the evaluation of color-related words is not directly determined by the color appearance and color category of the stimulus colors. Instead, the color image and color impression that have already been obtained through daily life and experiences influence color impressions more in color-related words. Color evaluations using words directly and relatively related to colors in the SD method is a more complex process and can be influenced more by the interaction between the semantic meanings of the words and the color appearance. In contrast, the evaluation of the paired comparison suggests that color appearance simply and directly determines word evaluation.

Zwischenstand und Nachträge zur Erforschung der stabreimenden Wortpaare im Frühdeutschen

Abstract Using findings in a magisterial catalogue of early German works and a new edition of prose texts through ca. 1200, the growing catalogue of alliterating word pairs in the early stages of German (Old High and Middle High) can be enlarged and improved with the most recent data. Interlinear versions, especially of the Psalms, have now been included.

Jón Ólafsson from Grunnavík and the I celandic language purism in the first half of the 18th century

This article examines aspects of Icelandic linguistic purism in the early 18th century, as revealed in a wordlist compiled by Jón Ólafsson from Grunnavík (1705–1779) and preserved in ms. AM 1013 4to (fol. 37v). After a brief introduction (§ 1), there follows an overview of the development of Icelandic language purism from its late-sixteenth-century origins to the first half of the 18th century (§ 2). The description of Jón Ólafsson’s life which follows (§ 3) highlights his genealogical links with Arngrímur Jónsson the Learned (1568–1648), a pioneering figure in the linguistic purism movement. The next section (§ 4) explores the wordlist itself and is divided into three subsections. The first (§ 4.1) summarises Jón’s linguistic scholarship as a whole. The second (§ 4.2) presents a diplomatic edition of the list and an analysis of eight of its word pairs, which have been chosen for their relevance with respect to their formation and ideological background. The analytical approach adopted describes the words as single lexical units in accordance with the main tenets of loanword studies, and examines the list in the light of language policy and planning studies. The final subsection (§ 4.3) highlights the relationship between the wordlist and Jón Ólafsson’s most important work, i.e. the Icelandic dictionary (ms. AM 433 fol.). The last section (§ 5) summarises the main points of the article and sets them within the context of pre-1750 Icelandic language purism. It will be argued that Jón Ólafsson’s work is of paramount importance in the development of Icelandic purist attitudes, both lexically, because of its close relationship with the principles set out by Arngrímur Jónsson in his Crymogæa, and in terms of the later history of the movement.

Phase-Amplitude Coupling of NREM Sleep Oscillations Shows Between-Night Stability and is Related to Overnight Memory Gains.

There is growing evidence in humans linking the temporal coupling between spindles and slow oscillations during NREM sleep with the overnight stabilization of memories encoded from daytime experiences in humans. However, whether the type and strength of learning influence that relationship is still unknown. Here we tested whether the amount or type of verbal word-pair learning prior to sleep affects subsequent phase-amplitude coupling (PAC) between spindles and slow oscillations (SO). We measured the strength and preferred timing of such coupling in the EEG of 41 healthy human participants over a post-learning and control night to compare intra-individual changes with inter-individual differences. We leveraged learning paradigms of varying word-pair (WP) load: 40 WP learned to a minimum criterion of 60% correct (n = 11); 40 WP presented twice (n = 15); 120 WP presented twice (n = 15). There were no significant differences in the preferred phase or strength between the control and post-learning nights, in all learning conditions. We observed an overnight consolidation effect (improved performance at delayed recall) for the criterion learning condition only, and only in this condition was the overnight change in memory performance significantly positively correlated with the phase of SO-spindle coupling. These results suggest that the coupling of brain oscillations during human NREM sleep is stable traits that are not modulated by the amount of pre-sleep learning, yet are implicated in the sleep-dependent consolidation of memory-especially when overnight gains in memory are observed.

Dataset of 37-subject EEG recordings using a low-cost mobile EEG headset during a semantic relatedness judgment task.

Dataset of 37-subject EEG recordings using a low-cost mobile EEG headset during a semantic relatedness judgment task.

Larger lexicons enable representation of fine-grained phonological similarity structure: Evidence from English L2 speakers’ sound similarity judgments of word pairs

Larger lexicons enable representation of fine-grained phonological similarity structure: Evidence from English L2 speakers’ sound similarity judgments of word pairs

빅데이터를 활용한 아동권리 연구 의미연결망 및 토픽 모델링 분석: 2000년 ~ 2024년까지

Objectives By examining research trends on children's rights using big data, we sought to understand social awareness of children's rights and suggest future research directions and implications related to children's rights. Methods We collected 25 years of data from January 2000 to December 2024 using children's rights as a keyword. To collect data, a data collection and purification process was performed using the Textome program, and text mining analysis, network analysis, and topic modeling (LDA) analysis were performed. Results First, looking at the text mining analysis results using big data, the word pairs that commonly exist in the word co-occurrence frequency analysis results are ‘Children's rights-respect’, ‘Children's rights-awareness’, ‘Children's rights-UN’, ‘It appeared as ‘Children’s Rights-Convention’. In addition, the word frequency (TF) and inverse document frequency (IDF) analysis results showed the following order: ‘teacher’, ‘childcare’, ‘awareness’, ‘rights’, ‘education’, and ‘respect’.Lastly, in the time series analysis, the word that appeared most highly by year was ‘children’s rights,’ and the word that appeared the least was ‘welfare.’ Second, looking at the results of network analysis using big data, 50 key words were extracted from the ego network results. As a result of CONCOR and clustering analysis, four clusters were formed. Third, looking at the topic modeling analysis results, a total of 10 topics were named for each topic. Conclusions Most of the research from the beginning of children's rights research in the 1990s to the present, which is not included in the data of this study, has emphasized only education and the role of teachers. The world of a child is made up of many different ecological systems, including primary caregivers, relatives, teachers, society, law, and the state. All of these members must cooperate to promote children's rights, but current research trends are limited to specific topics and limited research is being conducted.Therefore, it suggests that in order to truly promote children's rights, all systems should work together in a complex manner to promote children's rights, without being limited to specific topics. Through this study, it is hoped that research topics that have received relatively little attention so far in the field of children's rights research will be made visible, and that it will be used as basic data to expand the research scope of subsequent researchers and promote convergence research.

Negative emotional events retroactively disrupt semantic scaffolding of temporal-order memory.

Emotional responses pervade everyday life and exert temporally extended effects on cognition. Prior work shows that these modulatory effects of emotion on memory are highly selective, with semantic overlap helping to determine which nearby neutral details are prioritized in long-term memory. Although this has been demonstrated in item recognition, less is known about how emotion interacts with semantic information to influence temporal-order memory. Here, we developed an emotional oddball task in which participants encoded lists of neutral words that were either semantically related to or unrelated to a perceptually deviant emotional or neutral oddball word. We hypothesized that an adaptive memory system should selectively enhance temporal-order and recall memory for information that precedes or follows a conceptually related emotional stimulus. We found that order memory was enhanced for word pairs that preceded a semantically related neutral oddball, suggesting that semantics helps to scaffold temporal encoding processes. By contrast, emotional oddballs retroactively disrupted this mnemonic benefit of semantic overlap on temporal-order memory. Emotional oddballs also led to proactive impairments in order memory irrespective of semantic relatedness. After a 24-hr delay, emotion enhanced recall of preceding, semantically unrelated words. Encountering an emotional oddball also enhanced recall for subsequent words irrespective of semantic relatedness. Our findings suggest that emotion bidirectionally and selectively disrupts the temporal organization of memory, while also enhancing memory for individualized, unrelated elements of an emotional episode. (PsycInfo Database Record (c) 2025 APA, all rights reserved).

Functional magnetic resonance imaging of taxonomic and thematic processing of abstract and concrete word pairs.

This study used fMRI to examine the effect that the abstract-concrete dimension may have on the thematic-taxonomic distinction. The dual hub theory (DHT) posits that left angular gyrus supports thematic relationships and left temporal pole supports taxonomic relationships; however, the DHT is largely based on evidence from concrete words. It is important that theories of semantic organization include abstract words as they are ubiquitous in everyday discourse (Lupyan & Winter, Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 2018, 373, 20170137). Additionally, there is reason to believe that there may bean interaction between the abstract-concrete dimension and the thematic-taxonomic distinction, based on the different representational frameworks (DRF) hypothesis, which posits that abstract concepts are primarily organized by association/theme, and that concrete concepts are primarily organized by similarity/taxonomy. However, there appears to be a mismatch among the DHT, predictions of the DRF hypothesis for brain activation, and existing neuroimaging data for the concreteness effect. Thus, we sought to include abstract words in a test of the DHT and determine whether any interactions exist between the abstract-concrete dimension and the thematic-taxonomic distinction. While we replicated the localization of abstract and concrete word processing, we did not extend the DHT to include abstract words, nor did we find an interaction that would support the DRF hypothesis. Rather, our results align more closely with the hub-and-spoke model (Patterson etal., Nature Reviews Neuroscience, 2007, 8, 976).

The pretesting effect under divided attention

Completing a pretest (e.g., star—?) before receiving correct-answer feedback (e.g., star—night) can improve long-term retention of the material compared to material that was initially only studied. The present study examined whether this pretesting effect requires attentional resources during the initial pretest stage and the subsequent feedback stage. Two experiments were conducted in which participants studied word pairs which were either presented in full for 12 s and thus could be studied immediately (study-only trials) or were first only presented with the cue word of a pair and asked to guess the target word for 6 s before the complete pair was shown for another 6 s (pretest trials). Critically, learning occurred either under full attention or under distraction by a secondary task, with the distraction occurring either during the first 6 s or the last 6 s of a trial. While results showed the typical pretesting effect in the absence of any distraction, the effect remained intact when distraction occurred during the first 6 s of a pretest trial, but was eliminated when distraction occurred during the last 6 s. This pattern of results arose when distraction induced material-general (Experiment 1) and when it induced material-specific (Experiment 2) interference. Consistently, additional analyses showed greater recall impairments for pretested pairs when distraction occurred during Stage 2 than during Stage 1, although such impairment was present in both situations. The findings align with theoretical accounts suggesting critical roles of attentional processes for the pretesting effect.

Effects of Selective Retrieval Practice on Older Adults: Lesser Benefits, Greater Losses.

Retrieval practice enhances memory for practiced information, but at the price of impairing memory for unpracticed information, a phenomenon known as retrieval-induced forgetting (RIF). Evidence has shown that, for young adults, RIF can be eliminated after a long interval and when textual information is used as a memorandum. The current study aims to determine whether RIF is more durable and difficult to overcome for older adults due to their cognitive deficits. Both young and older participants completed a learning session on Day 1, during which they studied word pairs (Experiment 1) or scientific prose (Experiment 2). Then, they engaged in selective retrieval practice on Days 3, 5, or 7. Finally, they undertook a final test on Day 8. Experiment 1 showed no RIF for young but a robust RIF for older participants. Experiment 2 observed retrieval-induced facilitation for young but RIF for older participants. Although both young and older participants were encouraged to use an integration technique to facilitate learning during Experiment 2, the levels of integration only predicted the magnitudes of retrieval-induced facilitation for young but not for older participants. This study shows that older adults should be careful of carrying out selective retrieval because this may produce a more durable impairment in their memory for unpracticed information.

Optimizing forensic file classification: enhancing SFCS with βk hyperparameter tuning.

In forensic topical modelling, the α parameter controls the distribution of topics in documents. However, low, high, or incorrect values of α lead to topic sparsity, model overfitting, and suboptimal topic distribution. To control the word distribution across topics, the β parameter is introduced. However, low, high, or inappropriate β values lead to sparse distribution, disjointed topics, and abundant highly probable words. The βj parameter, in conjunction with seed-guided words based on Term Frequency and Inverse Document Frequency, is introduced to address the issues. Nevertheless, the data often suffers from skewness or noise due to frequent co-occurrences of unrelated polysemic word pairs generated using Pointwise Mutual Information. By integrating α, β, and βj into file classification systems, classification models converge to local optima with O(n log n* |V|) time complexity. To combat these challenges, this research proposes the SDOT Forensic Classification System (SFCS) with a functional parameter βk that identifies seed words by evaluating semantic and contextual similarity of word vectors. As a result, the topic distribution (Θd) is compelled to model the curated seed words within the distribution, generating pertinent topics. Incorporating βk into SFCS allowed the proposed model to remove 278 k irrelevant files from the corpus and identify 5.6 k suspicious files by extracting 700 blacklisted keywords. Furthermore, this research implemented hyperparameter optimization and hyperplane maximization, resulting in a file classification accuracy of 94.6%, 94.4% precision and 96.8% recall within O(n log n) complexity.