Lexical Model Research Articles

Complex Words as Shortest Paths in the Network of Lexical Knowledge.

Lexical models diverge on the question of how to represent complex words. Under the morpheme-based approach, each morpheme is treated as a separate unit, while under the word-based approach, morphological structure is derived from complex words. In this paper, we propose a new computational model of morphology that is based on graph theory and is intended to elaborate the word-based network approach. Specifically, we use a key concept of network science, the notion of shortest path, to investigate how complex words are learned, stored, and processed. The notion of shortest path refers to the task of finding the shortest or most optimal path connecting two non-adjacent nodes in a network. Building on this notion, the current study shows (i) that new complex words can be segmented into morphemesthrough the shortest path analysis; (ii) that attested English words tend to represent the shortest paths in the morphological network; and (iii) that novel (unattested) words receive higher acceptability ratings in experiments when they are formed along established optimal paths. The model's performance is tested in two experiments with human participants as well as against the behavioral data from the English Lexicon Project. We interpret our empirical results from the perspective of a usage-based model of grammar and argue that network science provides a powerful tool for analyzing language structure.

Sequential lexicon enhanced bidirectional encoder representations from transformers: Chinese named entity recognition using sequential lexicon enhanced BERT

Lexicon Enhanced Bidirectional Encoder Representations from Transformers (LEBERT) has achieved great success in Chinese Named Entity Recognition (NER). LEBERT performs lexical enhancement with a Lexicon Adapter layer, which facilitates deep lexicon knowledge fusion at the lower layers of BERT. However, this method is likely to introduce noise words and does not consider the possible conflicts between words when fusing lexicon information. To address this issue, we advocate for a novel lexical enhancement method, Sequential Lexicon Enhanced BERT (SLEBERT) for the Chinese NER, which builds sequential lexicon to reduce noise words and resolve the problem of lexical conflict. Compared with LEBERT, it leverages the position encoding of sequential lexicon and adaptive attention mechanism of sequential lexicon to enhance the lexicon feature. Experiments on the four available datasets identified that SLEBERT outperforms other lexical enhancement models in performance and efficiency.

ONTODIC

Abstract After analyzing other linguistic ontologies and lexical models with a variety of perspectives, objectives and results, it was found that they were unsuitable for building dictionaries. To help overcome these issues, a model of linguistic knowledge representation was designed. The ONTODIC model presents two major differences with respect to other ontologies. On the one hand, the aim of ONTODIC has been to model language from the premises of linguistics. This means studying the linguistic elements in their natural contexts, that is, the texts. On the other hand, the design of ONTODIC is based on the principles of description logic.

Determining the optimal environmental information for training computational models of lexical semantics and lexical organization.

Experiential theories of cognition propose that the external environment shapes cognitive processing, shifting emphasis from internal mechanisms to the learning of environmental structure. Computational modelling, particularly distributional models of lexical semantics (e.g., Landauer & Dumais, 1997) and models of lexical organization (e.g., Johns, 2021a), exemplifies this, highlights the influence of language experience on cognitive representations. While these models have been successful, comparatively less attention has been paid to the training materials used to train these models. Recent research has explored the role of social/communicatively oriented training materials on models of lexical semantics and organization (Johns, 2021a, 2021b, 2023, 2024), introducing discourse- and user-centred text training materials. However, determining the optimal training materials for these two model types remains an open question. This article addresses this problem by using experiential optimization (Johns, Jones, & Mewhort, 2019), which selects the materials that maximize model performance. This study will use experiential optimization to compare user-based and discourse-based corpora in optimizing models of lexical organization and semantics, offering insight into pathways towards integrating cognitive models in these areas. (PsycInfo Database Record (c) 2024 APA, all rights reserved).

Semantic and Pragmatic Analysis of a Turkish Sadness Verb Hüzünlen-

This study provides a semantic and pragmatic analysis of the Turkish verb "hüzünlen-" which conveys the meaning of sadness. Following a corpus-driven approach, query sequences from the Turkish National Corpus (TNC) were analysed according to the Extended Lexical Units model proposed by Stubbs (2005). In this context, 284 query sequences found in real spoken and written language samples were examined to determine the lexical profile of the verb "hüzünlen-." The aim of the study is to reveal the semantic and pragmatic usage differences of the verb "hüzünlen-" compared to other verbs expressing sadness. In this regard, the syntactic features, collocational tendencies, and context-dependent semantic and pragmatic characteristics of the verb "hüzünlen-" were thoroughly investigated. As a corpus-driven study, this research is the result of an inductive examination. The study demonstrated that the lexical framework of the verb "hüzünlen-" implies a sudden, intense sadness often related to past events and circumstances. The findings contribute to a better understanding of the cultural and linguistic features of expressing sadness in Turkish, as well as to comprehending the cognitive aspects of the emotion of sadness. Furthermore, the results of the study offer significant insights into how expressions of sadness are shaped in linguistic and cultural contexts in Turkish, allowing for a broader perspective on the linguistic representation of sadness. This research suggests that the verb "hüzünlen-" should be examined not only in Turkish but also in comparison with similar emotional expressions in other languages. Thus, it can contribute to understanding the universal and culture-specific features of language and emotion.

Neural tracking of speech acoustics in noise is coupled with lexical predictability as estimated by large language models.

Adults heard recordings of two spatially separated speakers reading newspaper and magazine articles. They were asked to listen to one of them and ignore the other, and EEG was recorded to assess their neural processing. Machine learning extracted neural sources that tracked the target and distractor speakers at three levels: the acoustic envelope of speech (delta- and theta-band modulations), lexical frequency for individual words, and the contextual predictability of individual words estimated by GPT-4 and earlier lexical models. To provide a broader view of speech perception, half of the subjects completed a simultaneous visual task, and the listeners included both native and non-native English speakers. Distinct neural components were extracted for these levels of auditory and lexical processing, demonstrating that native English speakers had greater target-distractor separation compared to non-native English speakers on most measures, and that lexical processing was reduced by the visual task. Moreover, there was a novel interaction of lexical predictability and frequency with auditory processing; acoustic tracking was stronger for lexically harder words, suggesting that people listened harder to the acoustics when needed for lexical selection. This demonstrates that speech perception is not simply a feedforward process from acoustic processing to the lexicon. Rather, the adaptable context-sensitive processing long known to occur at a lexical level has broader consequences for perception, coupling with the acoustic tracking of individual speakers in noise.Significance Statement In challenging listening conditions, people use focused attention to help understand individual talkers and ignore others, which changes their neural processing for speech at auditory through lexical levels. However, lexical processing for natural materials (e.g., conversations, audiobooks) has been difficult to measure, because of limitations of tools to estimate the predictability of individual words in longer discourses. The present investigation uses a contemporary large language model, GPT-4, to estimate word predictability, and demonstrates that listeners make online adaptations to their auditory neural processing in accord with these predictions; neural activity more closely tracks the acoustics of the target talker when words are less predictable from the context.

Modeling Lexical Tones for Speaker Discrimination.

Fundamental frequency (F0) has been widely studied and used in the context of speaker discrimination and forensic voice comparison casework, but most previous studies focused on long-term F0 statistics. Lexical tone, the linguistically structured and dynamic aspects of F0, has received much less research attention. A main methodological issue lies on how tonal F0 should be parameterized for the best speaker discrimination performance. This paper compares the speaker discriminatory performance of three approaches with lexical tone modeling: discrete cosine transform (DCT), polynomial curve fitting, and quantitative target approximation (qTA). Results show that using parameters based on DCT and polynomials led to similarly promising performance, whereas those based on qTA generally yielded relatively poor performance. Implications modeling surface tonal F0 and the underlying articulatory processes for speaker discrimination are discussed.

Injecting the score of the first-stage retriever as text improves BERT-based re-rankers

In this paper we propose a novel approach for combining first-stage lexical retrieval models and Transformer-based re-rankers: we inject the relevance score of the lexical model as a token into the input of the cross-encoder re-ranker. It was shown in prior work that interpolation between the relevance score of lexical and Bidirectional Encoder Representations from Transformers (BERT) based re-rankers may not consistently result in higher effectiveness. Our idea is motivated by the finding that BERT models can capture numeric information. We compare several representations of the Best Match 25 (BM25) and Dense Passage Retrieval (DPR) scores and inject them as text in the input of four different cross-encoders. Since knowledge distillation, i.e., teacher-student training, proved to be highly effective for cross-encoder re-rankers, we additionally analyze the effect of injecting the relevance score into the student model while training the model by three larger teacher models. Evaluation on the MSMARCO Passage collection and the TREC DL collections shows that the proposed method significantly improves over all cross-encoder re-rankers as well as the common interpolation methods. We show that the improvement is consistent for all query types. We also find an improvement in exact matching capabilities over both the first-stage rankers and the cross-encoders. Our findings indicate that cross-encoder re-rankers can efficiently be improved without additional computational burden or extra steps in the pipeline by adding the output of the first-stage ranker to the model input. This effect is robust for different models and query types.

Dialect Recognition via Lexical Processing: Is It a Viable Litmus Test?

For decades, linguists have been working to formulate an objective means of distinguishing dialects from languages, but dialect recognition has largely remained a subjective enterprise. Only recently have some studies proposed a processing-based psycholinguistic approach toward dialect recognition. These studies argued that dialect words are stored as a co-dependent representation, not as an independent representation of the words of bilingual speakers. Based on these studies, we investigated the lexical selection and processing mechanisms of bilingual and bidialectal speakers of two understudied languages, Oromo and Amharic, using the picture–word interference paradigm. We found independent lexical representations for both the bilingual and the bidialectal groups, which implies the involvement of the same cognitive mechanisms in both language and dialect processing. Thus, we argue that bidialectal speakers have flexible lexical representation and selection mechanisms that are dependent on the speakers’ previous language experience. Here, we propose a dynamic lexical selection model that accommodates diverse dialect ecologies.

Investigating lexical categorization in reading based on joint diagnostic and training approaches for language learners

Efficient reading is essential for societal participation, so reading proficiency is a central educational goal. Here, we use an individualized diagnostics and training framework to investigate processes in visual word recognition and evaluate its usefulness for detecting training responders. We (i) motivated a training procedure based on the Lexical Categorization Model (LCM) to introduce the framework. The LCM describes pre-lexical orthographic processing implemented in the left-ventral occipital cortex and is vital to reading. German language learners trained their lexical categorization abilities while we monitored reading speed change. In three studies, most language learners increased their reading skills. Next, we (ii) estimated, for each word, the LCM-based features and assessed each reader’s lexical categorization capabilities. Finally, we (iii) explored machine learning procedures to find the optimal feature selection and regression model to predict the benefit of the lexical categorization training for each individual. The best-performing pipeline increased reading speed from 23% in the unselected group to 43% in the machine-selected group. This selection process strongly depended on parameters associated with the LCM. Thus, training in lexical categorization can increase reading skills, and accurate computational descriptions of brain functions that allow the motivation of a training procedure combined with machine learning can be powerful for individualized reading training procedures.

DNN-based Multilingual Acoustic Modeling for Four Ethiopian Languages

In this paper, we present the results of experiments conducted on multilingual acoustic modeling in the development of an Automatic Speech Recognition (ASR) system using speech data of phonetically much related Ethiopian languages (Amharic, Tigrigna, Oromo and Wolaytta) with multilingual (ML) mix and multitask approaches. The use of speech data from only phonetically much related languages brought improvement over results reported in a previous work that used 26 languages (including the four languages). A maximum Word Error Rate (WER) reduction from 25.03% (in the previous work) to 21.52% has been achieved for Wolaytta, which is a relative WER reduction of 14.02%. As a result of using multilingual acoustic modeling for the development of an automatic speech recognition (ASR) system, a relative WER reduction of up to 7.36% (a WER reduction from 23.23% to 21.52%) has been achieved over a monolingual ASR. Compared to the ML mix, the multitask approach brought a better performance improvement (a relative WERs reduction of up to 5.9%). Experiments have also been conducted using Amharic and Tigrigna in a pair and Oromo and Wolaytta in another pair. The results of the experiments showed that languages with a relatively better language resources for lexical and language modeling (Amharic and Tigrigna) benefited from the use of speech data from only two languages. Generally, the findings show that the use of speech corpora of phonetically related languages with the multitask multilingual modeling approach for the development of ASR systems for less-resourced languages is a promising solution.

Examining the relations between early language skills and environmental variables and literacy skills: A longitudinal study from 2 to 9 years

A 7-year longitudinal study was conducted as part of the French national cohort ELFE ( N = 1095). The aim was to identify how and why early language skills at 2 years might predict later literacy skills assessed successively at 5, 7, and 9 years (LitSk5y; 7y; 9y). Using one and the same model, we also examined the relations between literacy skills in the three sessions and the potential impact of three environmental variables: the home learning environment (HLE) and two variables related to socioeconomic status (SES), namely, parental education level (PEL) and household income (HInc). The main findings reveal that expressive vocabulary at 2 years is a significant predictor of LitSk5y. No significant link was found for HLE. The other significant links reveal that PEL is associated with LitSk5y and LitSk7y, and HInc with LitSk5y and LitSk9y. The differential impact of these two SES variables is discussed. Moreover, the restructuring lexical model is proposed to explain how the development of early expressive vocabulary is linked to the development of literacy skills via an implicit stimulation of phonemic awareness during the acquisition of new words.

Manifestations of cancel culture in relation to the Roman Catholic Church in Poland

The aim of the article is to examine the media strategies of contesting the Roman Catholic Church in Poland on the example of three Polish opinion weeklies: "Polityka", "Newsweek" and "Tygodnik Powszechny". The period of our search was set at twelve months, starting from the outbreak of protests on October 22, 2020. On the basis of the media material analyzed above, three main strategies of contesting the Roman Catholic Church in the Polish opinion-forming press were identified: 1) exposing the negative aspects of the analyzed issues; 2) profiling of interlocutors and interviews; and 3) lexical and visual content modeling techniques.

A trilemma for the lexical utility model of the precautionary principle

AbstractBartha and DesRoches (Synthese 199(3–4):8701–8740, 2021) and Steel and Bartha (Risk Analysis 43(2):260–268, 2023) argue that we should understand the precautionary principle as the injunction to maximise lexical utilities. They show that the lexical utility model has important pragmatic advantages. Moreover, the model has the theoretical advantage of satisfying all axioms of expected utility theory except continuity. In this paper I raise a trilemma for any attempt at modelling the precautionary principle with lexical utilities: it permits choice cycles or leads to paralysis or implies that the smallest value difference that is possible in a context has extreme axiological implications.

Lexical semantic modeling as a means of effective vocabulary acquisition and expansion

English vocabulary is a challenging task for school students, so EFL teachers’ main aims is to find effective ways for faster and better vocabulary acquisition. This study implements using lexical semantic models based on semantic fields as an alternative strategy of teaching foreign vocabulary for vocabulary mastery and expansion in a structured way. The present study aims to investigate the effect of lexical semantic modeling on students’ foreign vocabulary acquisition and vocabulary expansion. The research design of this study is based on experimental teaching, modeling, observation, and post experimental testing. Thirty-three university students participated in this study. During the treatment students in experimental group learnt new words in paradigmatic and syntagmatic relationship by means semantic modeling, while students in the control group used the wordlists strategy. The research results has shown that learning words through lexical semantic modeling is more beneficial for EFL students comparing to wordlist strategy, as it facilitates better memorization and vocabulary retention, significantly enriches learners’ vocabulary through synonyms and antonyms, improves understanding of word relationship and correct word usage and collocations in different contexts and enlarges learner’s active vocabulary. The study results have shown that lexical semantic models has a positive effect on vocabulary acquisition and expansion when it is implemented together with lexical semantic exercises, and students practice new vocabulary using various types of lexical semantic exercises.

Information-Restricted Neural Language Models Reveal Different Brain Regions' Sensitivity to Semantics, Syntax, and Context.

A fundamental question in neurolinguistics concerns the brain regions involved in syntactic and semantic processing during speech comprehension, both at the lexical (word processing) and supra-lexical levels (sentence and discourse processing). To what extent are these regions separated or intertwined? To address this question, we introduce a novel approach exploiting neural language models to generate high-dimensional feature sets that separately encode semantic and syntactic information. More precisely, we train a lexical language model, GloVe, and a supra-lexical language model, GPT-2, on a text corpus from which we selectively removed either syntactic or semantic information. We then assess to what extent the features derived from these information-restricted models are still able to predict the fMRI time courses of humans listening to naturalistic text. Furthermore, to determine the windows of integration of brain regions involved in supra-lexical processing, we manipulate the size of contextual information provided to GPT-2. The analyses show that, while most brain regions involved in language comprehension are sensitive to both syntactic and semantic features, the relative magnitudes of these effects vary across these regions. Moreover, regions that are best fitted by semantic or syntactic features are more spatially dissociated in the left hemisphere than in the right one, and the right hemisphere shows sensitivity to longer contexts than the left. The novelty of our approach lies in the ability to control for the information encoded in the models' embeddings by manipulating the training set. These "information-restricted" models complement previous studies that used language models to probe the neural bases of language, and shed new light on its spatial organization.

Authorship identification methods in student plagiarism detection

In the modern educational context the problem of plagiarism is urgent and requires the development of effective methods of detection and prevention of this phenomenon. The application of authorship identification methods in the field of student plagiarism detection is considered. Different check, detect and analyze plagiarism approaches in various works are investigated. Both classical methods, which include text comparison and similarity search, and modern methods based on machine learning algorithms, as well as their combination and potential modifications, are considered. The advantages and limitations of each method are also discussed, and recommendations are given for choosing one or another approach according to the specific requirements of the research.Special attention is paid to such modern methods as metadata analysis and the application of neural networks. Stylistic analysis reveals authorial peculiarities such as word choice, preferred wording, and even punctuation. Lexical and syntactic models are used to identify repetitive phrases and structures that may indicate plagiarism. Statistical methods can identify anomalies in the use of words and phrases, and machine learning can create models to calculate the probability of plagiarism based on large amounts of data.Ultimately, an comparison of authorship identification techniques in the field of student plagiarism detection is provided, which aims to provide valuable information about different approaches and their applicability, and to help researchers and educators develop effective strategies for detecting and preventing plagiarism in educational environments.

A lexical network approach to second language development

This study investigates lexical development in second language (L2) learning from the perspective of complex dynamic system theory (CDST) using a complex network method. Based on authentic written output texts from L2 Chinese learners of different proficiency levels and language backgrounds, we successfully differentiate between different proficiency levels using a bi-gram lexical network model at a corpus level. A more in-depth investigation reveals that when compared to traditional lexical complexity indices, such as average word length and hapax legomena percentage (though Guiraud proves to be a robust predictor), the lexical network indices, such as network density and network clusters, provide a more profound understanding of L2 proficiency distinctions and a more precise approximation of the target language. Moreover, our findings illuminate the consistent manifestation of complex network characteristics within L2 Chinese lexical networks across all proficiency levels. Additionally, word association features, encompassing more than just word frequency information, provide comprehensive properties of the interlanguage system, as supported by their information gain values. We argue that studies within the CDST framework should integrate both lexical frequency and lexical network features to gain a comprehensive understanding of L2 lexical development.

A Combined Semantic Dependency and Lexical Embedding RoBERTa Model for Grid Field Relational Extraction

Relationship extraction is a crucial step in the construction of a knowledge graph. In this research, the grid field entity relationship extraction was performed via a labeling approach that used span representation. The subject entity and object entity were used as training instances to bolster the linkage between them. The embedding layer of the RoBERTa pre-training model included word embedding, position embedding, and paragraph embedding information. In addition, semantic dependency was introduced to establish an effective linkage between different entities. To facilitate the effective linkage, an additional lexically labeled embedment was introduced to empower the model to acquire more profound semantic insights. After obtaining the embedding layer, the RoBERTa model was used for multi-task learning of entities and relations. The multi-task information was then fused using the parameter hard sharing mechanism. Finally, after the layer was fully connected, the predicted entity relations were obtained. The approach was tested on a grid field dataset created for this study. The obtained results demonstrated that the proposed model has high performance.

Типы пространственной семантики в модели фрейма «Киберпространство» (на материале медиатекстов на русском языке)

The aim of the study is to identify the cognitive models and language means used to express locative semantics in the ‘Cyberspace’ frame involving the texts of Russian-language media. The study is based on the hypothesis that global news stories concerning events in cyberspace often receive a specific framing in the Russian media discourse due to both socio-cultural singularity and the peculiarities of spatial semantics expression in the Russian language. The scientific novelty of the study is determined by the fact that cyberspace representation in the mass media in Russian has not previously been the object of linguistic research, although new lexical units describing phenomena from the cybertechnology industry are being actively included both in the media discourse and in the common Russian language and require a systematic description. As a result of the study, the process of adapting the global news frame in the Russian-language media has been described for the first time taking into account lexical and word-formation models and linguocultural aspects. It has been found that the semantics of cyberspace is actualised within the framework of three thematic clusters: real, associative-background and conceptual locations described using borrowed vocabulary, authorial neologisms, metaphorisation and homonymy.