Word Sequences Research Articles

Semantic coherence markers: The contribution of perplexity metrics

Devising automatic tools to assist specialists in the early detection of mental disturbances and psychotic disorders is to date a challenging scientific problem and a practically relevant activity. In this work we explore how language models (that are probability distributions over text sequences) can be employed to analyze language and discriminate between mentally impaired and healthy subjects. We have preliminarily explored whether perplexity can be considered a reliable metrics to characterize an individual’s language. Perplexity was originally conceived as an information-theoretic measure to assess how much a given language model is suited to predict a text sequence or, equivalently, how much a word sequence fits into a specific language model. We carried out an extensive experimentation with healthy subjects, and employed language models as diverse as N-grams – from 2-grams to 5-grams – and GPT-2, a transformer-based language model. Our experiments show that irrespective of the complexity of the employed language model, perplexity scores are stable and sufficiently consistent for analyzing the language of individual subjects, and at the same time sensitive enough to capture differences due to linguistic registers adopted by the same speaker, e.g., in interviews and political rallies. A second array of experiments was designed to investigate whether perplexity scores may be used to discriminate between the transcripts of healthy subjects and subjects suffering from Alzheimer Disease (AD). Our best performing models achieved full accuracy and F-score (1.00 in both precision/specificity and recall/sensitivity) in categorizing subjects from both the AD class, and control subjects. These results suggest that perplexity can be a valuable analytical metrics with potential application to supporting early diagnosis of symptoms of mental disorders.

The The Connotative Meaning of Emotional Words in Kate Chopin’s The Awakening

As connotative meaning is the associative meaning of words and it depends on personal usage and perception, it requires analysis and research in different areas such as literary, journal, cultural contexts. In this paper, the researcher chooses a context from a literary work that is a novel written by Cate Chopin which is entitled “The Awakening”. She selects the emotional words used in the text. Then, the connotative meaning of the selected emotional words are detected and classified according to their types in an organized table. Stating the positive, negative, and the neutral type. Later, the findings will announce the usage sequence (the most usage to the fewest usage) of the emotional words’ connotative meaning types. First, the positive connotative meanings are mostly used. Secondly, the negative connotative meanings and lastly the neutral connotative meanings are used.

Inside the Mind of an AI: Materiality and the Crisis of Representation

The advent of Transformer architectures for neural nets and the high language proficiency of programs like GPT-3 confront us with a fundamental question. How can, or should, literary criticism proceed when the text's creator is not a human but a machine? The query shakes to its core literary criticism, and indeed the entire enterprise of critical inquiry. This essay confronts the issue head-on, arguing that it does matter whether language is produced by humans or AIs. It provides context for the development of neural nets that process natural language, looking at competitors to GPT-3 and similar Transformer novels. It then goes into depth on GPT-3's Transformer architecture and how it processes word sequences. It compares how the AI learns language with how human children learn it, arguing that the differences result in a systemic fragility of reference for AI's language understanding. It interrogates the null strategy of assuming there is no difference between human- and machine-generated text and explores its implications, arguing for a middle position between the program understanding nothing about meaning or everything. Finally, it offers four strategies for how literary criticism can engage with machine-generated language, arguing that machine narratives can be of significant interest in themselves. The point is not to ignore the powerful capabilities of GPT-3 to generate compelling narratives and texts, but rather to devise and deploy a new kind of literary criticism that can adequately interpret its complexities.

Traitement de l’ambiguïté syntaxique et sémantique en TA neuronale : analyse de la traduction de l’anglais vers le français, l’espagnol et l’italien

Syntactic and Semantic Ambiguity Processing in Neural MT from English to French, Spanish and Italian
 Despite recent advances in artificial intelligence, human translators outperform MT for at least three types of tasks: identifying referents in anaphora (especially of the interphrastic kind), resolving semantic ambiguity (which is mainly due to polysemy or homonymy), and resolving syntactic ambiguity (especially with poorly inflected source languages such as English). Using the results obtained by two freely available online machine translation programs, Google Translate and DeepL, we examine how these two types of ambiguity are processed in translation from English into French, Spanish and Italian. Our results show that the two programs perform well overall in resolving the simplest cases of syntactic ambiguity, with difficulties arising more frequently for noun phrases featuring atypical syntactic divisions and rarely used collocations. MT output for ambiguous structures involving verb roots followed by the –ING morpheme (flying planes, growing pains) is studied, as well as syntactic structures in which two or more nouns are preceded by one or more adjectives. MT handles relatively well the longest of those structures (ADJ ADJ N N N N), probably because their subsets are part of the bilingual or target language monolingual corpora that underlie MT systems. Structures involving head modification and coordination (ADJ N AND N) are also known to pose problems for MT and human translators alike. But since many of the most frequent N AND N structures involve cohyponyms (men and women, brothers and sisters), antonyms (rights and duties, costs and benefits) or near- synonyms (aid and advice), their translation as a whole unit generally triggers the choice of correct syntact dependencies in translation. Structures in which the adjective only modifies the first noun (fresh air and exercise, social sciences and humanities) are much less frequent and are also probably translated as a whole unit. Structures involving premodification, coordination and post-modification may give rise to four distinct types of structures depending on whether long-range dependencies apply (detailed [knowledge and understanding] of the IT industry, [ethnic group] and [place of birth], invaluable [context and [source of information], [close friend] and confidant] of Mr Jones. Structures in which both long-range dependencies apply (integrated prevention and control of pollution) are the ones which most frequently cause errors for MT. Semantic ambiguity cases have been processed with increasing success by MT, especially when collocates vary widely for the main two meanings of homonyms (a well-known example is the word pen). Processing polysemy (for instance the medical use of conditions in pre-existing conditions) is a bit more of a challenge for MT. Other cases involving concentration of several polysemic terms in the same sentence (Changing the placement of beams relative to the staff involves changing the direction of the stems in the beam) also create difficulties for MT when the polysemic terms are used without any of their usual collocates (here in the specialised field of musical edition). Homonymy cases involving grammatical category changes (N-to-V or V-to-N conversion) seem to continue to pose the most difficulties to neural MT, despite increasing consideration of intra- and extraphrastic context. Potentially ambiguous word sequences (treatment increase in as the daily dose and duration of treatment increase), which were processed incorrectly before neuronal MT, are now correctly translated. But word sequences in which one word belongs to a part of speech which is not the most commonly used one (e.g. the noun remains in what remains can be considered) may cause occasional errors. Several examples that involve the verb founder are studied, and they frequently trigger translation of the noun in all three Romance languages (or translations of the verbs find or found due to incorrect segmentation).

Complexity resolved: profile of a Slovenian child with protracted phonological development over an intervention period

ABSTRACT This paper describes the complex phonological patterns of one Slovenian-speaking girl with protracted phonological development who received phonological intervention. At 4;6, the child’s word length, stress and vowels were age-level, but she showed reduced match levels (accuracy) for syllable structure and consonants. Unusual constraints on word position and sequences, particularly concerning fricatives and stops, resulted in many assimilations, reduplications and metatheses. Addressing these constraints in the first intervention period (to 4;9) led to higher overall accuracy and a transition from less typical to more typical mismatches (“errors”), e.g., sibilant grooving (stridency) and rhotic production as singletons and in clusters. The profile focuses on that initial pre- and post-treatment period, giving a brief overview of the follow-up intervention and outcomes and reflecting on implications for future research and clinical practice. 1 1 The abstract is available in Slovenian language (see Supplementary material).

Las relaciones entre prosodia y comprensión lectora en alumnos hispanohablantes: revisión sistemática y metaanálisis

Fluency has been considered one of the relevant factors involved in reading comprehension, by providing a bridge between decoding and comprehension. At present, fluency includes, alongside rate and accuracy, prosody as one the parts which plays a preponderant role in the construction of meaning during the reading process. Prosodic reading is the use of oral language features when reading a text, such as pausing, stress and intonation, in order that the text will be read aloud with the tonal and rhythmic characteristics of speech. Furthermore, prosodic reading makes it possible to organize word sequences into syntactically cohesive units of meaning. The aim of this study was to conduct a systematic review and meta-analysis, firstly, in order to determine the situation of research articles which enquire into the relation between prosody and reading comprehension, and secondly, to measure the significance of that relation in students who are native speakers of Spanish. To that end, a bibliographic search was performed without time restrictions in the following databases: Dialnet, Redalyc, SciELO, ERIC, Scopus and PubMed. The systematic review included 18 studies. The qualitative synthesis showed four categories of articles: (1) validation studies of instruments to assess reading fluency; (2) intervention studies on different prosodic aspects and enquiries into their impact on reading comprehension; (3) comparisons of good and poor comprehenders; and (4) correlational studies. A multilevel meta-analysis was performed in the correlational studies (n = 12), examining 59 correlation coefficients. The analysis showed a significant association between reading comprehension and prosody (t(58) = 9.77, p < .001) with a moderate effect of .46 (z = .49 [95 % CI = .39, .59]). Once outlier values had been removed, no variation was found in the size of the estimated effect (r = .46). Two additional models were adjusted, including as moderating variables, respectively, the type of assessment performed on comprehension (sentence or text level) and the educational level (primary or secondary). The results obtained from these models suggest the association between comprehension and prosody is neither moderated by the type of assessment carried out (F(1, 57) = 1.51, ρ = .22), nor by the educational level (F(1, 57) = 0.10, ρ = .75). In summary, the meta-analysis showed a moderate relation between prosody and reading comprehension. These relations continued beyond the level at which the latter variable (sentence or text) was examined, and throughout the school trajectory. In other words, and in contrast to other aspects of fluency, such as accuracy and rate, prosody seems to maintain its contribution to comprehension. In sum, the findings support the claim that suprasegmental skills, both at lexical level (i. e. the skills related to the identification, stress and manipulation of stressed syllables in words) and at metrical level (the capabilities responsible for processing intonation, rhythm and pausing when reading), contribute to the semantic processing of sentences and texts in Spanish, regardless of the educational level under analysis. These results should be moderated given the limited number of studies found and the smaller amount of research carried out at secondary school level. In spite of these limitations, the relationships observed among the variables studied support the importance of prosodic reading to psychoeducational diagnosis and intervention in reading comprehension, and they provide significant background for future research. https://doi.org/10.16888/interd.2022.39.3.7

Development online models for automatic speech recognition systems with a low data level

Speech recognition is a rapidly growing field in machine learning. Conventional automatic speech recognition systems were built based on independent components, that is an acoustic model, a language model and a vocabulary, which were tuned and trained separately. The acoustic model is used to predict the context-dependent states of phonemes, and the language model and lexicon determine the most possible sequences of spoken phrases. The development of deep learning technologies has contributed to the improvement of other scientific areas, which includes speech recognition. Today, the most popular speech recognition systems are systems based on an end-to-end (E2E) structure, which trains the components of a traditional model simultaneously without isolating individual elements, representing the system as a single neural network. The E2E structure represents the system as one whole element, in contrast to the traditional one, which has several independent elements. The E2E system provides a direct mapping of acoustic signals in a sequence of labels without intermediate states, without the need for post-processing at the output, which makes it easy to implement. Today, the popular models are those that directly output the sequence of words based on the input sound in real-time, which are online end-to-end models. This article provides a detailed overview of popular online-based models for E2E systems such as RNN-T, Neural Transducer (NT) and Monotonic Chunkwise Attention (MoChA). It should be emphasized that online models for Kazakh speech recognition have not been developed at the moment. For low-resource languages, like the Kazakh language, the above models have not been studied. Thus, systems based on these models have been trained to recognize Kazakh speech. The results obtained showed that all three models work well for recognizing Kazakh speech without the use of external additions.

Only Words Count; the Rest Is Mere Chattering: A Cross-Disciplinary Approach to the Verbal Expression of Emotional Experience.

The analysis of sequences of words and prosody, meter, and rhythm provided in an interview addressing the capacity to identify and describe emotions represents a powerful tool to reveal emotional processing. The ability to express and identify emotions was analyzed by means of the Toronto Structured Interview for Alexithymia (TSIA), and TSIA transcripts were analyzed by Natural Language Processing to shed light on verbal features. The brain correlates of the capacity to translate emotional experience into words were determined through cortical thickness measures. A machine learning methodology proved that individuals with deficits in identifying and describing emotions (n = 7) produced language distortions, frequently used the present tense of auxiliary verbs, and few possessive determiners, as well as scarcely connected the speech, in comparison to individuals without deficits (n = 7). Interestingly, they showed high cortical thickness at left temporal pole and low at isthmus of the right cingulate cortex. Overall, we identified the neuro-linguistic pattern of the expression of emotional experience.

Neurophysiological evidence for goal-oriented modulation of speech perception.

Speech perception depends on the dynamic interplay of bottom-up and top-down information along a hierarchically organized cortical network. Here, we test, for the first time in the human brain, whether neural processing of attended speech is dynamically modulated by task demand using a context-free discrimination paradigm. Electroencephalographic signals were recorded during 3 parallel experiments that differed only in the phonological feature of discrimination (word, vowel, and lexical tone, respectively). The event-related potentials (ERPs) revealed the task modulation of speech processing at approximately 200ms (P2) after stimulus onset, probably influencing what phonological information to retain in memory. For the phonological comparison of sequential words, task modulation occurred later at approximately 300ms (N3 and P3), reflecting the engagement of task-specific cognitive processes. The ERP results were consistent with the changes in delta-theta neural oscillations, suggesting the involvement of cortical tracking of speech envelopes. The study thus provides neurophysiological evidence for goal-oriented modulation of attended speech and calls for speech perception models incorporating limited memory capacity and goal-oriented optimization mechanisms.

Automatic Description Method for Sports Videos Based on Economic Management Effectiveness Visualization Techniques

The automatic description (AD) of sports videos is a fundamental task for archiving the content of broadcasters, as well as understanding video scenes, and economic management effectiveness visualization techniques are key to the classification of sports videos. In this paper, a freestyle gymnastics video is used as an example to study the automatic video description by observing the set of movements of an athlete in a freestyle gymnastics video to generate the terminology of the movements performed by that athlete. The technique used in this paper to visualize the effectiveness of economic management is the long and short-term memory (LSTM) network model, which is used to learn the mapping relationship between word sequences and video frame sequences. Attention mechanisms (AM) are also introduced to highlight the importance of keyframes that determine freestyle gymnastics movements. The study is carried out by building a dataset of free gymnastics (FG) breakdown movements from professional events and applying a planned sampling method. Experimental results show that the method can improve the accuracy of an automatic free gymnastics video (FGV) description. The proposed method has a wide range of applications in sports analysis and instruction.

Narrative thinking lingers in spontaneous thought

Some experiences linger in mind, spontaneously returning to our thoughts for minutes after their conclusion. Other experiences fall out of mind immediately. It remains unclear why. We hypothesize that an input is more likely to persist in our thoughts when it has been deeply processed: when we have extracted its situational meaning rather than its physical properties or low-level semantics. Here, participants read sequences of words with different levels of coherence (word-, sentence-, or narrative-level). We probe participants’ spontaneous thoughts via free word association, before and after reading. By measuring lingering subjectively (via self-report) and objectively (via changes in free association content), we find that information lingers when it is coherent at the narrative level. Furthermore, and an individual’s feeling of transportation into reading material predicts lingering better than the material’s objective coherence. Thus, our thoughts in the present moment echo prior experiences that have been incorporated into deeper, narrative forms of thinking.

Protracted phonological development of a Portuguese six-year-old from the perspective of nonlinear phonology

ABSTRACT 1 1 is also available in European Portuguese. As a contribution to a special issue with individual profiles in protracted phonological development (PPD), we present a European Portuguese-speaking six-year-old, “Vicente”. By age six years, Portuguese-learning children have generally mastered most of the phonology. However, Vicente showed severe persisting PPD, which was negatively impacting his general socialisation. Although word length and stress often matched the adult targets as did many consonants and vowels, consonants were restricted in distribution and sequences. Consonant clusters showed a particularly high proportion of reduction. A constraints-based nonlinear phonological analysis led to a proposed intervention plan to address needs across the phonological hierarchy in the context of his greater need for enhanced socialisation: starting with the more attainable new word positions and sequences for consonants in his inventory, before addressing a major need for new word structure (clusters) and minor needs for segmental development (additional coronals).

Time-series classification with SAFE: Simple and fast segmented word embedding-based neural time series classifier

Dictionary-based classifiers are an essential group of approaches in the field of time series classification. Their distinctive characteristic is that they transform time series into segments made of symbols (words) and then classify time series using these words. Dictionary-based approaches are suitable for datasets containing time series of unequal length. The prevalence of dictionary-based methods inspired the research in this paper. We propose a new dictionary-based classifier called SAFE. The new approach transforms the raw numeric data into a symbolic representation using the Simple Symbolic Aggregate approXimation (SAX) method. We then partition the symbolic time series into a sequence of words. Then we employ the word embedding neural model known in Natural Language Processing to train the classifying mechanism. The proposed scheme was applied to classify 30 benchmark datasets and compared with a range of state-of-the-art time series classifiers. The name SAFE comes from our observation that this method is safe to use. Empirical experiments have shown that SAFE gives excellent results: it is always in the top 5%–10% when we rank the classification accuracy of state-of-the-art algorithms for various datasets. Our method ranks third in the list of state-of-the-art dictionary-based approaches (after the WEASEL and BOSS methods).

Ontology based approach in solving collision situations at sea

The article presents an innovative approach to the implementation of semi-automatic communication in solving collision situations at sea based on navigational and communication ontology. Data used in this process are obtained from navigational systems and equipment, including AIS, ARPA and ECDIS. The problem of communication processes management has been analysed. Based on this, navigational and communication ontology has been developed and presented. The construction of ontology takes into account, inter alia, IMO's Standard Marine Communication Phrases. The developed ontology allows the implementation of communication processes as the concatenation of sequences of words corresponding to utterances used by humans. The use of the said ontology in the process of semi-automatic communication may contribute to the reduction of errors in communication and facilitate communication to those with poor English skills. The developed ontology was used to support solving collision situations at sea. Examples of its use in real and simulated cases are presented. The authors made use of official reports of marine accident investigation boards and own scenarios simulating typical situations of ship encounters. The results of research will be used to create an automatic communication system for manned and autonomous ships.

Post-Authorship Attribution Using Regularized Deep Neural Network

Post-authorship attribution is a scientific process of using stylometric features to identify the genuine writer of an online text snippet such as an email, blog, forum post, or chat log. It has useful applications in manifold domains, for instance, in a verification process to proactively detect misogynistic, misandrist, xenophobic, and abusive posts on the internet or social networks. The process assumes that texts can be characterized by sequences of words that agglutinate the functional and content lyrics of a writer. However, defining an appropriate characterization of text to capture the unique writing style of an author is a complex endeavor in the discipline of computational linguistics. Moreover, posts are typically short texts with obfuscating vocabularies that might impact the accuracy of authorship attribution. The vocabularies include idioms, onomatopoeias, homophones, phonemes, synonyms, acronyms, anaphora, and polysemy. The method of the regularized deep neural network (RDNN) is introduced in this paper to circumvent the intrinsic challenges of post-authorship attribution. It is based on a convolutional neural network, bidirectional long short-term memory encoder, and distributed highway network. The neural network was used to extract lexical stylometric features that are fed into the bidirectional encoder to extract a syntactic feature-vector representation. The feature vector was then supplied as input to the distributed high networks for regularization to minimize the network-generalization error. The regularized feature vector was ultimately passed to the bidirectional decoder to learn the writing style of an author. The feature-classification layer consists of a fully connected network and a SoftMax function to make the prediction. The RDNN method was tested against thirteen state-of-the-art methods using four benchmark experimental datasets to validate its performance. Experimental results have demonstrated the effectiveness of the method when compared to the existing state-of-the-art methods on three datasets while producing comparable results on one dataset.

Who does what to whom? graph representations of action-predication in speech relate to psychopathological dimensions of psychosis

Graphical representations of speech generate powerful computational measures related to psychosis. Previous studies have mostly relied on structural relations between words as the basis of graph formation, i.e., connecting each word to the next in a sequence of words. Here, we introduced a method of graph formation grounded in semantic relationships by identifying elements that act upon each other (action relation) and the contents of those actions (predication relation). Speech from picture descriptions and open-ended narrative tasks were collected from a cross-diagnostic group of healthy volunteers and people with psychotic or non-psychotic disorders. Recordings were transcribed and underwent automated language processing, including semantic role labeling to identify action and predication relations. Structural and semantic graph features were computed using static and dynamic (moving-window) techniques. Compared to structural graphs, semantic graphs were more strongly correlated with dimensional psychosis symptoms. Dynamic features also outperformed static features, and samples from picture descriptions yielded larger effect sizes than narrative responses for psychosis diagnoses and symptom dimensions. Overall, semantic graphs captured unique and clinically meaningful information about psychosis and related symptom dimensions. These features, particularly when derived from semi-structured tasks using dynamic measurement, are meaningful additions to the repertoire of computational linguistic methods in psychiatry.

Phonological delay of segmental sequences in a Greek child’s speech

ABSTRACT This paper investigates persistent elements of protracted phonological development (PPD) at ages 5;10 and 6;3 in a monolingual Greek girl’s speech following earlier articulation intervention (3;6) and post-intervention assessment (4;3). The re-assessment data examined here, five months apart, were elicited using the Phonological Assessment for Greek (PAel). Results reveal interesting idiosyncratic patterns in the production of segmental sequences (VV, CC) in complex syllables and longer words, with a striking imbalance between singletons and sequences, which holds especially true for inconsistency in the acquisition of /ɾ/ across CV, CC, and C.C contexts. Phonological delay surfaces as chronological mismatches, idiosyncratic forms and, most notably, disparity between segmental and structural development.

An end-to-end deep learning system for requirements classification using recurrent neural networks

Existing requirements classification approaches mainly use lexical and syntactical features to classify requirements using both traditional machine learning and deep learning approaches with promising results. However, the existing techniques depend on word and sentence structures and employ preprocessing and feature engineering techniques to classify requirements from textual natural language documents. Moreover, existing studies deal with requirements classification as binary or multiclass classification problems and not as multilabel classification, although a given requirement can belong to multiple classes at the same time. The objective of this study is to classify requirements into functional and different non-functional types with minimal preprocessing and to model the task as a multilabel classification problem. In this paper, we use Bidirectional Gated Recurrent Neural Networks (BiGRU) to classify requirements using raw text. We investigated two different approaches: (i) using word sequences as tokens and (ii) using character sequences as tokens. Experiments conducted on the publicly available PROMISE and EHR datasets show the effectiveness of the presented techniques. We achieve state-of-the-art results on most of the tasks using word sequences as tokens. Requirements can be effectively classified into functional and different non-functional categories using the presented recurrent neural networks-based deep learning system, which involves minimal text prepossessing and no feature engineering.

Effects of Noise and Chemical Exposure on Peripheral and Central Auditory Pathways in Normal-hearing Workers.

To assess the effects of noise and chemical exposure on peripheral and central auditory pathways in normal-hearing workers exposed to chemicals or high noise levels and compare the groups with each other and with workers not exposed to either of these agents. A total of 54 normal-hearing workers were divided into three groups (chemical, noise, control) and submitted to the following assessments: conventional and extended high-frequency pure-tone audiometry; transient and distortion-product otoacoustic emissions, the inhibitory effect of the efferent auditory pathway; and Staggered Spondaic Word (SSW) and Pitch Pattern Sequence (PPS) test. There were no significant differences between the groups in extended high-frequency hearing thresholds. Significantly lower amplitudes were observed in the noise group for otoacoustic emissions. There were significantly more absences of the inhibitory effect of the efferent system in the noise group. There was no difference between the groups in the SSW test, while in PPS, the noise group performed worse than the control group. These findings suggest that noise exposure produced deleterious effects on the workers' peripheral and central auditory systems, despite their normal hearing thresholds. The chemical group did not have significantly different results from those of the control group. It is important that individuals exposed to noise or chemicals have their auditory pathways monitored with complementary assessments.

Juxtaposing Urdu to English Captions of Facebook/ Metaverse: A Neural Machine Translation Study

Facebook/ Metaverse is a world-leading social networking platform that is equally popular in the world that is equally popular globally Due to the diversity of its users, Facebook is dedicated to improving its Machine Translation (MT) to refine the communication process and remove linguistic hurdles for efficient communication. Recently, Metaverse has started Neural Machine Translation (NMT) for the improvement of translation. NMT uses artificial neural networks to predict a more appropriate word choice and sequence. This research investigates the efficiency of Neural Machine Translation (NMT) in translating text from Urdu to the English language. The Oxford Urdu English Dictionary and FARHANG E AASFIA dictionaries have been used as validation parameters in translation. The semantic and syntactic errors have been segregated and categorized with Antoine Berman’s Twelve Deforming Tendencies for rectification. Major findings reveal that the most frequent errors result from transliteration that damages the semantic and syntactic structure of the text.