Search

This article addresses the issue of how nominal countability is grammatically encoded and how the counting function is realized in classifier languages by investigating classifier phrases in Donglan Zhuang, a Tai-Kadai language. According to the prevailing individuation account, classifiers are required to individuate nouns, which can then be counted by numerals. Under this approach, countability and counting are conflated. Donglan Zhuang has two syntactically distinct types of classifiers, namely, numeral classifiers CLnum and noun classifiers CLnom. CLnum performs the counting/measuring function, comparable to the cardinality function proposed in Scontras (The semantics of measurement, Harvard University, Cambridge, 2014), and CLnom encodes syntactic countability by singling out sortal nouns from the mass domain, whereby sortal nouns are, meanwhile, turned into (taxonomic) kind terms. Noun classifiers in Donglan Zhuang pose a challenge to Chierchia's (Nat Lang Semant 6(4):339–405, 1998) "bare argument hypothesis" and suggest that bare nouns in classifier languages are not uniform with respect to the [±argument] parametric setting. [ABSTRACT FROM AUTHOR]

Aim/Background: This research aims to develop an automated contextual classifier for scholarly papers by utilizing established algorithms and understanding the information retention of different parts of a scholarly article, such as the Abstract, Article Title, and Keywords. It also seeks to recommend a contextual classifier-based recommender system to help academics identify credible sources. Scholarly articles from various study fields often use similar terms in their titles and keywords. However, finding a publication venue can be challenging for researchers at the beginning of a scientific inquiry. Thus, it is crucial to classify information based on its context, especially when abstracts, keywords, and titles receive equal attention. Materials and Methods: An ensembled model was developed and trained using 114K instances from 38 classes of the Web of Science (WoS) dataset and 40 classes of the Dimensions dataset. The ensemble approach incorporated both machine learning and deep learning algorithms to build a diverse classifier. The model was evaluated by testing it with an 80:20 train-test split to assess performance. The classifier was further integrated into a recommender system designed to suggest probable publication sources based on given article information. Results: The ensemble classification approach demonstrated superior performance with faster inference and efficient training time. The balanced training model, tested on 114K instances, effectively categorized scholarly articles into one of 40 categories. The recommender system was capable of recommending up to 10 probable publication sources based on the article's Title, Keywords, and Abstract. Models utilizing abstractions yielded the best results and provided a better understanding of the context in every iteration of the experiment. Conclusion: This study successfully developed an ensemble-based contextual classifier for academic papers, which can also function as a recommender system. The system aids researchers in choosing the most appropriate sources to publish by categorizing articles into 40 categories and suggesting credible publication venues. This approach simplifies the decision-making process for academics, enabling them to identify relevant publications and suitable sources for their work more efficiently. [ABSTRACT FROM AUTHOR]

This paper establishes that Sorani Kurdish has a mass/count distinction that can be distinguished by looking at, which nouns occur with plural moprhology (-an), optional classifiers (e.g. dane, CLinainmate), unit denumerators (yek, =one'), fuzzy denumerators (nsikay panja =about fifty'), and other denumerators (herdû =both'). Moreover, properties of countability in Sorani Kurdish can be used to distinguish six classes of nouns that differ with respect to their acceptability with the aforementioned properties, examples of nouns in each class being setyare (=car'), polîs (=police'), miqes (=scissors'), manga (=cattle'), çiyakanî Himalaya (=Himalayas'), and xîn (=blood'). This data was gathered by comparing Sorani Kurdish near equivalents to English countability properties found in previous studies, and asking five speakers of Sorani Kurdish for their judgments, following semantic fieldwork methodology. This research expands what is found about countability in Sorani Kurdish grammar books. This paper shows that Sorani Kurdish has both a rich classifier system and mass/count distinction, a combination that has not yet been documented. Knowing these characteristics of Sorani Kurdish helps to create a clearer picture of the linguistic variation regarding countability and suggests a relationship between the number of countability properties and countability classes in a given language. [ABSTRACT FROM AUTHOR]

As there has been an increase in the number of mental illness cases, there is a need to curb this problem. Due to the complexity of traditional techniques based on clinical diagnosis, there should be an automated system for the detection and prevention of illness and hence there comes the need to develop a depression prediction system. The data is collected from twitter, followed by preprocessing and cleaning which includes removal of stop words, URL and HTML tags, expanding abbreviations etc. Following this, the process of feature extraction will be used to extract word count, pronouns, negations and other features from the comments made by the users following selection. Long Short Term Memory and Support Vector Machine classifiers are applied to obtain the results. Further, the tokenized words are embedded into a vector and passed to the encoder in and then to perform depression prediction, a classification layer is added on the top of the transformer output. Similarly the tokenized input is embedded into a vector of SVM, which is mapped to a class label, followed by classification into depressed and non-depressed classes. Both methods are compared for depression analysis. This helps in providing an early detection of depression in people. Social network and microblogging sites such as Twitter are widespread amongst all generations nowadays where people connect and share their feelings, emotions, pursuits etc. Depression, one of the most common mental disorder, is an acute state of sadness where person loses interest in all activities. [ABSTRACT FROM AUTHOR]

This study is concerned with the aperture of the mid vowel /E/ in nonfinal syllables in Quebec French. The hypothesis tested is that in underived disyllabic words, the aperture of /E/ would be determined via harmony with the following vowel. Based on predictions from a classifier trained on acoustic properties of word-final vowels, nonfinal vowels were labeled as mid-close or mid-open. Although distant coarticulatory effects were observed, the harmony hypothesis was not supported. The results revealed a bias toward a mid-open quality and a reduced acoustic distinction, which warrant further investigation. [ABSTRACT FROM AUTHOR]

The F-POD is designed for passive acoustic monitoring of odontocetes. The offline classifiers can identify and separate porpoise-like sounds from dolphin-like sounds. We show that these two classifiers are not working independently. Run together, virtually no detections of both species were reported within the same minute, whereas 10% of the detection positive minutes were reported positive for both species when the two classifiers were run sequentially. This has important implications for interpretation of data in areas containing both species groups, and we call for reporting all analysis details in such studies and for further description and analysis of the classifiers. [ABSTRACT FROM AUTHOR]

Most of the current day industries are suffering from nonlinear processes. Thus, both the stability and the process performance of high-degree nonlinear systems with dominating delay might be difficult to achieve. Adaptive and intelligent fuzzy classifiers and controllers have been more popular in recent years as a means of overcoming a significant number of difficulties faced by the industrial sector. A large number of dynamic process plants with a variety of orders and kinds have been represented heuristically and recognized. Fuzzy structures have also been employed for these interactive systems by making use of fuzzy and linguistic techniques. In view of all these initiatives, the purpose of this paper is to conduct an experimental investigation into the performance of a LabVIEW-based Type-1 Adaptive Mamdani Fuzzy Controller (AMFC) that has been designed and applied over a lag-dominant and a second-order nonlinear Dual Input Tank System (DITS) and Single Input Tank System (SITS). Compared to other type-1 approaches that were previously experimented with and are now in existence, the adaptability of AMFC demonstrates that it is quite effective. Performance indices such as the Integrated/Summated Absolute Error (IAE) and the Integrated/Summated Squared Error (ISE) are also computed for several variable set point profiles of DITS. These indices measure errors in integrated absolute value and integrated squared value, respectively. Adaptive Type-1 Intelligent Fuzzy Controller's response and error reduction efficiency have been found for several flow configurations of DITS, namely Multiple Input Multiple Output (MIMO) and Single Input Single Output (SISO). From the results, it can be concluded that the proposed experimental validation may be used for a wide variety of process challenges that are experienced in industrial systems to achieve robust and low error controller performances. [ABSTRACT FROM AUTHOR]

In the face of surging online education around the globe, it seems quite necessary and helpful for learners and teachers to have the plethora of online resources well sorted out beforehand. To some extent, the efficiency and accuracy of resource search and retrieval may determine the quality and influence of online education. In this research, based on the methodological framework of design science, the support vector machine (SVM) algorithm is chosen to optimise the design of an accurate resource classifier. The aim is to improve the unsatisfactory classification effect of traditional classification methods for online education resources, so that online learners can enjoy more accurate and convenient access to education resources they are seeking out of many more. For the purpose of performance evaluation, the proposed SVM-based classifier was compared with two other classification methods based on multiple neutral networks and deep learning respectively. Upon collection and pre-processing of online materials, the features of educational resources were extracted and output in the form of feature vectors. By calculating the similarity between the extracted feature vectors and the standard vectors of the set type, we obtained the classification results of online education resources for each of the three classifiers. It was found that, compared with those of the two traditional classification methods, the precision ratio and the recall ratio of the proposed classifier improved by 3.26% and 2.01% respectively. In the meantime, the proposed SVM-based classifier seems to more advantageous in performance balance with better F measurement. [ABSTRACT FROM AUTHOR]

Deep learning has received a growing interest in recent years for detecting different types of pavement distresses and automating pavement condition assessment. A proper choice of deep learning models is key for successful pavement assessment applications. In this study, we first present a comprehensive experimental comparison of state-of-the-art image classification models to evaluate their performances on 11 pavement objects classification. Our experiments are conducted in different dimensions of comparison, including deep classifier architecture, effects of network depth, and computational costs. Five convolutional neural network (CNN) classifiers widely used in transportation applications, including VGG16, VGG19, ResNet50, DenseNet121, and a generic CNN (as the control model), are tested with a comprehensive pixel-level annotated dataset for 11 different distress and non-distress classes (UCF-PAVE 2017). In addition, we investigate a simple yet effective approach of encoding contextual information with multi-scale input tiles to classify highly random pavement objects in size, shape, intensity, texture, and direction. Our comparison results show that the multi-scale approach significantly improves the classification accuracy for all compared deep classifiers at a negligible extra computational cost. Finally, we provide recommendations of how to improve the classification performance of deep CNNs for automated pavement condition assessment based on the comparison results. [ABSTRACT FROM AUTHOR]

An important cue to the meaning of a new noun is its accompanying classifier. For example, in English, X in "a sheet of X" should refer to a broad, flat object. A classifier is required in Chinese to quantify nouns. Using children's overt responses in an object/picture selection task, past research found reliable semantic knowledge of classifiers in Mandarin-reared children at around age three. However, it is unclear how children's semantic knowledge differs across different types of classifiers and how this difference develops with age. Here we use an arguably more sensitive measure of children's language knowledge (the intermodal preferential-looking paradigm) to examine Mandarin-reared three-, four-, and five-year-olds' semantic knowledge of four types of classifiers indicating animacy (human vs. animal distinction), configuration (how objects are arrayed), object shape, and vehicle function. Multiple factors were matched across classifier types: the number of classifiers, perceived familiarity and perceived typicality of the target, and the visual similarity of the two images paired together. Children's performances differed across classifier types, as they were better with animacy classifiers than with configuration and vehicle function classifiers. Their comprehension was reliable for animacy, object shape, and vehicle function classifiers but not for configuration classifiers. Furthermore, we did not find conclusive evidence for an age-dependent improvement in the child's performance. The analysis, including the oldest (five-year-olds) and youngest (three-year-olds) children, revealed a marginally significant age effect. [ABSTRACT FROM AUTHOR]

The scope of the Internet of Things (IoT) becomes inevitable in the communication and information-sharing routines of human life, similar to any technological architecture. The IoT is also not exempted from vulnerability to security issues and is even more vulnerable as the networks of IoT are built of non-smart devices. Though the few contributions endeavored to defend against the botnet's attacks on IoT, they partially or poorly performed to defend against the flash crowd or attacks by botnets on IoT networks. In this context, the method “Flash Attack Prognosis by Ensemble Supervised Learning for IoT Networks” derived in this manuscript is centric on defending the flash attacks by botnets. Unlike contemporary models, the proposed method uses the fusion of traditional network features and temporal features as input to train the classifiers. Also, the curse of dimensionality in the training corpus, which is often, appears in the corpus of flash attack transactions by a botnet, has addressed by the ensemble classification strategy. The comparative analysis of the statistics obtained from the experimental study has displayed the significance and robustness of the proposed model compared to contemporary models. [ABSTRACT FROM AUTHOR]

We investigate whether second language (L2) learners of English rely on first language (L1) transfer and atomicity in the acquisition of the count/mass distinction by examining L1-French and L1-Chinese learners of English. Atomicity encodes whether a noun contains 'atoms' or minimal elements that retain the property of the noun. As a semantic universal, atomicity holds across languages. However, the count/mass status of nouns may differ cross-linguistically. Our results, which show difficulty on atomic mass nouns in both learner groups, support an argument that atomicity is used as a semantic universal in the L2. Our results also suggest that both count/mass status in the L1 and word frequency in the L2 impact performance, suggesting roles for both L1 lexical transfer and lexical frequency. In addition, learners had better performance on abstract as opposed to concrete atomic mass nouns, providing evidence consistent with a theory of the accessibility of atoms. [ABSTRACT FROM AUTHOR]

Copyright of Polonica is the property of Instytut Jezyka Polskiego PAN and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This dataset provides detailed metadata on ca. 10.2 million works of fiction and nonfiction written after 1799 in 521 different languages available in the HathiTrust Digital Library. The dataset bolsters the May 2022 Hathifile by supplying missing predicted fiction tags with a bespoke BERT-based multilingual classifier. Our classifier completes the catalogue with an additional 400,000 non-English volumes predicted to be works of fiction, capturing 95% of all works presently provided by HathiTrust. We provide each work with metadata including the work's genre at the level of fiction or nonfiction, length in pages, original language, and the year the work was published. With a total page count of ca. 1.4 billion pages, our dataset provides researchers with a substantial source of non-English modern literature. We also present insight into how multilingual classifiers can be trained with monolingual data, itself a discovery with implications for the study of lower resource languages. We hope our provisions will accelerate empirical research into non-English prose and literature. [ABSTRACT FROM AUTHOR]

Korean and Japanese are typical classifier languages that classify a noun based on the semantic type of its referent with a counter word when plurality is involved. Their plural marking appears to be optional when the noun denotes general plurality, but obligatory when a noun is marked by the semantic feature, [+specific] (Ioni, Ko & Wexler, 2004). In this study, we characterize the so-called 'optional' plurality in Korean and Japanese as the manifestation of the grammaticalization. Drawing on actual data, we demonstrate that the plural suffix is increasingly used as a neutral plural marker. The grammaticalization is more prevalent when the nouns are higher in the Animacy Hierarchy (e.g., Comrie, 1989), although there are differences in acceptability between Korean and Japanese. We attribute the differences to the language-specific uses of the plural suffix, namely, Japanese associative reading and Korean event-plural reading (Song, 1997). [ABSTRACT FROM AUTHOR]

Vegetation mapping is an active research area in the domain of remote sensing. This study proposes a methodology for the mapping of vegetation by integrating several vegetation indices along with original spectral bands. The Land Use Land Cover classification was performed by two powerful Machine Learning techniques, namely Random Forest and AdaBoost. The Random Forest algorithm works on the concept of building multiple decision trees for the final prediction. The other Machine Learning technique selected for the classification is AdaBoost (adaptive boosting), converts a set of weak learners into strong learners. Here, multispectral satellite data of Dehradun, India, was utilised. The results demonstrate an increase of 3.87% and 4.32% after inclusion of selected vegetation indices by Random Forest and AdaBoost respectively. An Overall Accuracy (OA) of 91.23% (kappa value of 0.89) and 88.59% (kappa value of 0.86) was obtained by means of the Random Forest and AdaBoost classifi- ers respectively. Although Random Forest achieved greater OA as compared to AdaBoost, interestingly AdaBoost provided better class-specific accuracy for the Shrubland class compared to Random Forest. Furthermore, this study also evaluated the importance of each individual feature used in the classifi- cation. Results demonstrated that the NDRE, GNDVI, and RTVIcore vegetation indices, and spectral bands (NIR, and Red-Edge), obtained higher importance scores. [ABSTRACT FROM AUTHOR]

Despite achieving accuracies higher than 90% on recognizing upper-limb movements through sEMG (surface Electromyography) signal with the state of art classifiers in the laboratory environment, there are still issues to be addressed for a myo-controlled prosthesis achieve similar performance in real environment conditions. Thereby, the main goal of this review is to expose the latest researches in terms of strategies in each block of the system, giving a global view of the current state of academic research. A systematic review was conducted, and the retrieved papers were organized according to the system step related to the proposed method. Then, for each stage of the upper limb motion recognition system, the works were described and compared in terms of strategy, methodology and issue addressed. An additional section was destined for the description of works related to signal contamination that is often neglected in reviews focused on sEMG based motion classifiers. Therefore, this section is the main contribution of this paper. Deep learning methods are a current trend for classification stage, providing strategies based on time-series and transfer learning to address the issues related to limb position, temporal/inter-subject variation, and electrode displacement. Despite the promising strategies presented for contaminant detection, identification, and removal, there are still some factors to be considered, such as the occurrence of simultaneous contaminants. This review exposes the current scenario of the movement classification system, providing valuable information for new researchers and guiding future works towards myo-controlled devices. [ABSTRACT FROM AUTHOR]

One of the lessons learned from the global financial crisis in 2008 was raising attention to monitoring and maintaining household vulnerability, particularly household credit risk, by using the default rate as the indicator. The indicator would be worsening at the economic recession, likewise, recently happened caused by the pandemic. The default event has a complex nonlinearity relationship among the determinants. To tackle the complex relationship, this study suggests exploiting machine learning approach in modeling the probability of default, especially the individual and ensemble classifiers. Therefore, this study aims to investigate changes of the Indonesian household financial resilience before and during the pandemic, supported by the individual-level data of the Financial Information Service System. This study finds that the ensemble classifiers, notably extreme gradient boosting, have a more predominant performance than the individual classifiers. The best model, then has the feature importance analysis to identify the variable pattern in explaining the default event periodically which reveals the pattern changes before and during the pandemic. The cost of debt/repayment capability and the policy mix is significant in explaining the default event. At the same time, the project location feature weakens in discriminating the target class. [ABSTRACT FROM AUTHOR]

In recent years, deep learning algorithms have been popular in recognizing targets in synthetic aperture radar (SAR) images.However, due to the problem of overfitting, the performance of these models tends to worsen when just a small number of training data are available. In order to solve the problems of overfitting and an unsatisfied performance of the network model in the small sample remote sensing image target recognition, in this paper, we uses a deep residual network to autonomously acquire image features and proposes the Deep Feature Bayesian Classifier model (RBnet) for SAR image target recognition. In the RBnet, a Bayesian classifier is used to improve the effect of SAR image target recognition and improve the accuracy when the training data is limited. The experimental results on MSTAR dataset show that the RBnet can fully exploit effective information in limited samples and recognize the target of the SAR images more accurately. Compared with other state-of-the-art methods, our method offers significant recognition accuracy improvements under limited training data.Noted that theRBnet is moderately difficult to implement and has the value of popularization and application in engineering application scenarios in the field of small-sample remote sensing target recognition and recognition. [ABSTRACT FROM AUTHOR]

COVID-19 is a novel coronavirus that spread around the globe with the initial reports coming from Wuhan, China, turned into a pandemic and caused enormous casualties. Various countries have faced multiple COVID spikes which put the medical infrastructure of these countries under immense pressure with third-world countries being hit the hardest. It can be thus concluded that determining the likeliness of death of a patient helps in avoiding fatalities which inspired the authors to research the topic. There are various ways to approach the problem such as past medical records, chest X-rays, CT scans, and blood biomarkers. Since blood biomarkers are most easily available in emergency scenarios, blood biomarkers were used as the features for the model. The data was first imputed and the training data was oversampled to avoid class imbalance in the model training. The model is composed of a voting classifier that takes in outputs from multiple classifiers. The model was then compared to base models such as Random Forest, XGBoost, and Extra Trees Classifier on multiple evaluation criteria. The F1 score was the concerned evaluation criterion as it maximizes the use of the medical infrastructure with the minimum possible casualties by maximizing true positives and minimizing false negatives. [ABSTRACT FROM AUTHOR]

In recent years, the relationship between network graph parameters and the existence of specific H-factor has attracted the attention of researchers. The latest research has found that for factor deleted graphs and factor critical graphs, the tight bounds of graph parameters are changing as the low bound of connectivity changes. This work studies the existence of {K1,1,K1,2,C1}-factor from high-dimensional space perspective. Several surfaces reveal that the dynamic changing of tight parameter (toughness, isolated toughness, sun toughness, binding number) bounds are defined. The precise expression of these surfaces are presented. [ABSTRACT FROM AUTHOR]

Static authentication methods, like passwords, grow increasingly weak with advancements in technology and attack strategies. Continuous authentication has been proposed as a solution, in which users who have gained access to an account are still monitored in order to continuously verify that the user is not an imposter who had access to the user credentials. Mouse dynamics is the behavior of a user's mouse movements and is a biometric that has shown great promise for continuous authentication schemes. This article builds upon our previous published work by evaluating our dataset of 40 users using three machine learning and three deep learning algorithms. Two evaluation scenarios are considered: binary classifiers are used for user authentication, with the top performer being a 1-dimensional convolutional neural network (1D-CNN) with a peak average test accuracy of 85.73% across the top-10 users. Multi-class classification is also examined using an artificial neural network (ANN) which reaches an astounding peak accuracy of 92.48%, the highest accuracy we have seen for any classifier on this dataset. [ABSTRACT FROM AUTHOR]

Some languages around the Pacific have multiple possessive classes of alienable constructions using appositive nouns or classifiers. This pattern differs from the most common kind of alienable/inalienable distinction, which involves marking, usually affixal, on the possessum, and has only one class of alienables. The Japanese language isolate Ainu has possessive marking that is reminiscent of the Circum-Pacific pattern. It is distinctive, however, in that the possessor is coded not as a dependent in an NP but as an argument in a finite clause, and the appositive word is a verb. This paper gives a first comprehensive, typologically grounded description of Ainu possession and reconstructs the pattern that must have been standard when Ainu was still the daily language of a large speech community; Ainu then had multiple alienable class constructions. We report a cross-linguistic survey expanding previous coverage of the appositive type and show how Ainu fits in. We split alienable/inalienable into two different phenomena: argument structure (with types based on possessibility: optionally possessible, obligatorily possessed, and non-possessible) and valence (alienable, inalienable classes). Valence-changing operations are derived alienability and derived inalienability. Our survey classifies the possessive systems of languages in these terms. [ABSTRACT FROM AUTHOR]

The embedded, computational and cloud elements of industrial cyber physical systems (CPS) generate large volumes of data at high velocity to support the operations and functions of corresponding time-critical and mission-critical physical entities. Given the non-deterministic nature of these entities, the generated data streams are susceptible to dynamic and abrupt changes. Such changes, which are formally defined as concept drifts, leads to a decline in the accuracy and robustness of predicted CPS behaviors. Most existing work in concept drift detection are classifier dependent and require labeled data. However, CPS data streams are unlabeled, unstructured and change over time. In this paper, we propose an unsupervised machine learning algorithm for continuous concept drift detection in industrial CPS. This algorithm demonstrates three types of unsupervised learning, online, incremental and decremental. Furthermore, it distinguishes between abrupt and reoccurring drifts. We conducted experiments on SEA, a widely cited synthetic dataset of concept drift detection, and two industrial applications of CPS, task tracking in factory settings and smart energy consumption. The results of these experiments successfully validate the key features of the proposed algorithm and its utility of detecting change in non-deterministic CPS environments. [ABSTRACT FROM AUTHOR]

An information retrieval (IR) system is the core of many applications, including digital library management systems (DLMS). The IR-based DLMS depends on either the title with keywords or content as symbolic strings. In contrast, it ignores the meaning of the content or what it indicates. Many researchers tried to improve IR systems either using the named entity recognition (NER) technique or the words’ meaning (word sense) and implemented the improvements with a specific language. However, they did not test the IR system using NER and word sense disambiguation together to study the behavior of this system in the presence of these techniques. This paper aims to improve the information retrieval system used by the DLMS by adding the NER and word sense disambiguation (WSD) together for the English and Arabic languages. For NER, a voting technique was used among three completely different classifiers: rules-based, conditional random field (CRF), and bidirectional LSTM-CNN. For WSD, an examples-based method was used to implement it for the first time with the English language. For the IR system, a vector space model (VSM) was used to test the information retrieval system, and it was tested on samples from the library of the University of Kufa for the Arabic and English languages. The overall system results show that the precision, recall, and F-measures were increased from 70.9%, 74.2%, and 72.5% to 89.7%, 91.5%, and 90.6% for the English language and from 66.3%, 69.7%, and 68.0% to 89.3%, 87.1%, and 88.2% for the Arabic language. [ABSTRACT FROM AUTHOR]

Automatic speech-based severity level classification of Parkinson's disease (PD) enables objective assessment and earlier diagnosis. While many studies have been conducted on the binary classification task to distinguish speakers in PD from healthy controls (HCs), clearly fewer studies have addressed multi-class PD severity level classification problems. Furthermore, in studying the three main issues of speech-based classification systems – speaking tasks, features, and classifiers – previous investigations on the severity level classification have yielded inconclusive results due to the use of only a few, and sometimes just one, type of speaking task, feature, or classifier in each study. Hence, a systematic comparison is conducted in this study between different speaking tasks, features, and classifiers. Five speaking tasks (vowel task, sentence task, diadochokinetic (DDK) task, read text task, and monologue task), four features (phonation, articulation, prosody, and their fusion), and four classifier architectures (support vector machine (SVM), random forest (RF), multilayer perceptron (MLP), and AdaBoost) were compared. The classification task studied was a 3-class problem to classify PD severity level as healthy v s. mild v s. severe. Two MDS-UPDRS scales (MDS-UPDRS-III and MDS-UPDRS-S) were used for the ground truth severity level labels. The results showed that the use of the monologue task and the articulation and fusion of features improved classification accuracy significantly compared to the use of the other speaking tasks and features. The best classification systems resulted in a rate of accuracy of 58% (using the monologue task with the articulation features) for the MDS-UPDR-III scale and 56% (using the monologue task with fusion of features) for the MDS-UPDRS-S scale. • Speech-based classification of severity of Parkinson's disease was studied. • 5 speaking tasks, 4 feature sets, and 4 classifiers were compared. • Monologue task & articulation / fusion features improved classification accuracy. • Classifier had a minor effect on classification accuracy. • Classification differences between MDS-UPDRS-III and MDS-UPDRS-S were small. [ABSTRACT FROM AUTHOR]

• A novel voice conversion technique for body-conducted unvoiced speech is proposed based on Taylor-Boltzmann Machines. The proposed algorithm was the integration of Taylor Series with Boltzmann machine Deep Neural Networks. • A body conductive microphone referred as the non audible murmur (NAM) microphone is used to capture the non-audible murmur from the neck part behind the ear. • The classifier was trained with the features likes "pitch chroma, spectral centroid, and spectral skewness" that are derived from the incoming body conducted low voice verbal communication signal. • In the initial training phase, a Taylor series approximation DNN model is trained from a collection of stereo data which encompasses the NAM speech of same phonemes represented as log-power spectral features. • Moreover, in the recognition phase, a well trained Taylor-DNN model that is based on Boltzmann Machine intakes the speech features and generates the "log-power-spectra features" and it aids in recognizing the speech waveform. • The paper may gain high interest from the readers and experts of Speech processing, Voice Conversion Techniques, body-conducted unvoiced speech. Communication becomes effectual while the verbal communication signal reaches the destination with profound characteristics. An application based on human voice attracts the researcher's attention. It leads the researcher to proceed towards the speech conversion for low voice humans having murmured voices as their attitude. The researchers introduced a technique for improving the body-conducted unvoiced speech for silent speech communication using Taylor-Boltzmann machine-trained optimized Deep Neural Network (DNN). The projected model includes two main stages: the training stage as well as recognition stage. In the training stage, the following functionalities are undergone: (i) pre-processing, (ii) feature extraction, and (iii) deep learning-based speech enhancement. Initially, the collected input Non-Audible Murmur (NAM) speech signal is subjected to the pre-processing stage, wherein the improved wiener filtering is employed for suppressing the noise in the input signal. From the pre-processed speech signal, the most relevant characteristics such as "spectral skewness, spectral chroma, spectral centroid and improved Mel Frequency Cepstral Coefficients (MFCC)" are extracted, and they are fused together. Then, using these fused features, the Taylor series with optimized DNN and Boltzmann Machine (Taylor-DNN-BM) model is trained. The final enhanced speech is acquired from the Taylor- optimized DNN-BM. The recognition phase encapsulates the feature extraction and deep learning-based speech enhancement stage. The input speech data enters into the feature extraction phase, wherein the characteristics such as "spectral skewness, spectral chroma, spectral centroid and improved (MFCC)" are retrieved. After that, these features are fused and fed as input to the Taylor-optimized DNN-BM model. From Taylor- optimized DNN-BM, the enhanced speech signal is acquired. In DNN, the weight function is fine-tuned using the newly projected Self Improved Flower Pollination Algorithm (SI-FPA) algorithm. This SI-FPA model is the conceptual improvement of the Flower Pollination Algorithm (FPA). Finally, the projected model is compared with the existing models to validate its efficiency in terms of speech enhancement. [ABSTRACT FROM AUTHOR]

• A novel feature selection model is proposed for Parkinson's disease diagnosis. • Strongly relevant features are selected by the maximal information coefficient. • A dynamic feature selection strategy based on self-organizing map is proposed. • The model combines relevance, redundancy and complementarity to select features. • The effectiveness of the model is verified on two widely used datasets. Voice disorders are one of the incipient symptoms of Parkinson's disease (PD). The existing PD voice feature analysis methods generally reduce dimensionality by selecting features with high relevance and low redundancy, ignoring the complementarity between features. To fully explore the intrinsic connection between PD features, this paper proposes a local dynamic feature selection fusion method (LDFSF) for PD identification and symptom severity prediction. First, the maximal information coefficient is used to calculate the relevance between features and the target value, and removes low relevant features. Then, a dynamic feature selection strategy based on self-organizing map network clustering is designed. This strategy dynamically updates the feature subset in a forward manner, and combines the relevance, redundancy and complementarity for feature selection. Among them, self-organizing map network clustering reduces the computational cost by reducing the number of candidate features, and effectively improves the accuracy of the model. Finally, the selected features are fed into the appropriate classifier or regressor to achieve accurate PD identification and severity progress prediction. Experiments on public OPDD and PTDS datasets show that the classification accuracy, sensitivity and specificity of LDFSF method on OPDD are 98.20%, 92.00% and 99.50% respectively, and the mean absolute error of predicting PD severity (motor- and total-UPDRS) on PTDS are 1.62 and 1.99 respectively, which are better than the compared methods for the selection of five and eight features respectively, thus indicating that the proposed LDFSF method can be effectively used in PD voice diagnosis. [ABSTRACT FROM AUTHOR]

In recent years, a range of Neural Referring Expression Generation (REG) systems have been built and they have often achieved encouraging results. However, these models are often thought to lack transparency and generality. Firstly, it is hard to understand what these neural REG models can learn and to compare their performance with existing linguistic theories. Secondly, it is unclear whether they can generalise to data in different text genres and different languages. To answer these questions, we propose to focus on a sub-task of REG: Referential Form Selection (RFS). We introduce the task of RFS and a series of neural RFS models built on state-of-the-art neural REG models. To address the issue of interpretability, we probe these RFS models using probing classifiers that consider information known to impact the human choice of Referential Forms. To address the issue of generalisability, we assess the performance of RFS models on multiple datasets in multiple genres and two different languages, namely, English and Chinese. • The task of Referential Form Selection is introduced. • A number of Neural Referential Form Selection models are proposed. • Multigenre and multilingual datasets are used to test the models' generalisability. • To interpret the models' behaviour, we propose 8 probing tasks to probe each model. [ABSTRACT FROM AUTHOR]