Automated Processing Techniques Research Articles

Design and implementation of voice-command controller for fixed-wing unmanned aerial vehicles using automatic speech recognition and natural language processing techniques

This paper explores the development of a voice command controller leveraging the capabilities of an automatic speech recognition (ASR) system and natural language processing (NLP) technique to manage a fixed-wing unmanned aerial vehicle (UAV). The controller is designed to interpret voice commands for controlling fixed-wing UAVs. The implementation of the system involved two key stages: (1) implementation of a voice command controller using integrated ASR and NLP techniques deployed in a simulated plane in the SITL simulator followed by (2) deployment of the controller to an actual Sky Surfer plane fixed-wing aircraft. The results indicate that the algorithm achieved an average confidence rate of 91.86 % in transcribing voice commands to words, with a Word Error Rate (WER) of approximately 0.021. The developed system demonstrated the ability to interpret both low-level and high-level commands for UAV control interfaces. Such an interface offers greater intuitiveness compared to traditional RC controls, potentially requiring less training to operate effectively. Moreover, it reduces human workload, as once commands are issued, the system can execute them without the need for continuous supervision.

AI-based automatic identification and processing techniques for agricultural safety information

In the context of globalization, agricultural safety is directly linked to food security and human health. Modern agriculture’s challenge is effectively monitoring and processing vast agricultural safety information in the digital era. This study aims to achieve the automatic identification and intelligent processing of agricultural safety information within a digital media environment by applying artificial intelligence (AI) technologies. Initially, the background of AI applications in agriculture is explored, followed by an analysis of the urgent need for automated processing of agricultural safety information and an overview of current research in this field. It is demonstrated that, despite progress, existing methods still lack in-depth feature extraction and real-time capabilities of the data processing systems. A method based on multi-level feature fusion for identifying agricultural product safety is proposed to address these limitations alongside a decentralized AI Internet of Things (IoT) system for processing agricultural safety information. The multi-level feature fusion method can extract and integrate essential details on agricultural products from different dimensions and levels, thus enhancing the identification accuracy. Meanwhile, the decentralized AIoT system strengthens the efficiency and reliability of data processing, ensuring timely responses to agricultural safety information. Through deep neural networks, efficient recognition and classification of agricultural product images were achieved, enabling the automatic identification of agricultural safety information in a digital media environment. Utilizing deep learning algorithms, the system could learn and understand the characteristics of different agricultural products and accurately identify potential safety issues, such as pests, diseases, or spoilage, providing timely monitoring and management for agricultural production. The innovation of this research lies in the comprehensive utilization of multi-level data features and advanced Internet of Things (IoT) technology, significantly improving the level of intelligence in agricultural safety supervision.

Evaluating Automated Seismic Event Detection Approaches: An Application to Victoria Land, East Antarctica

AbstractAs seismic data collection continues to grow, advanced automated processing techniques for robust phase identification and event detection are becoming increasingly important. However, the performance, benefits, and limitations of different automated detection approaches have not been fully evaluated. Our study examines how the performance of conventional techniques, including the Short‐Term Average/Long‐Term Average method and cross‐correlation approaches, compares to that of various deep learning models. We also evaluate the added benefits that transfer learning may provide to machine learning‐based seismic applications. Each detection approach has been applied to 3 years of data recorded by stations in East Antarctica, which contain both cryospheric and tectonic‐related seismic events. Our results demonstrate that by integrating different event detection approaches with transfer learning, the strengths of each approach can maximize seismic detection accuracy and reliability while also enhancing the completeness of associated event catalogs. That said, model performance can vary, depending on the quality of the data set as well as the data augmentation schemes and training strategies employed. Our results in East Antarctica provide new insight into polar seismicity and demonstrate how automated event detection approaches can be optimized to investigate seismic activity in challenging environments.

Qualification and Quantification of Porosity at the Top of the Fuel Pins in Metallic Fuels Using Image Processing

Approximately 130,000 metal fuel pins were irradiated in the Experimental Breeder Reactor II (EBR-II) during its 30 years of operation to develop and characterize existing and prospective fuels. For many of the metal fuel irradiation experiments, neutron radiography imaging was performed to characterize fuel behavior, such as fuel axial expansion. While several fuel expansion results obtained from neutron radiography imaging have been published, the analysis of neutron radiography for the purpose of describing statistical properties of porous matter formed on top of the fuel pins, also referred to as fluff in previous publications, is significantly less represented in the literature with just a single paper so far. This study aims to validate and augment results reported in previous publications using automated image processing. The paper describes the statistical properties of the porous matter in terms of nine parameters derived from radiography images and correlates those parameters with such fuel properties as composition, expansion, temperature, and burnup. The reported results are based on 1097 fuel pins of eight different fuel compositions. For three major fuel types, U-10Zr, U-8Pu-10Zr, and U-19Pu-10Zr, a clear negative correlation is found between the Pu content and five parameters describing the amount of porous matter generated. The parameters describing granularity properties, however, showed either negative correlation or nonlinear dependency from fuel composition. The parameters describing the amount showed a positive correlation with fuel axial expansion, while granularity parameters showed a negative correlation with axial expansion. The dependency on cladding temperature was found to be weak. A positive correlation is demonstrated for volume parameters and fuel burnup. In general, reported results confirm and validate findings published in previous studies using a much larger number of pins and automated processing techniques, which easily lend themselves to reproducibility, thus avoiding subjective bias.

Predicting Transitions from Mild Cognitive Impairment to Dementia within 5‐years

AbstractBackgroundEarly detection of individuals at high risk of future progression toward dementia is in great need, particularly in the era of emerging treatments for Alzheimer’s disease. Given the easier administration of neuropsychological (NP) tests over invasive and expensive biomarker acquisition, we seek to predict the future cognitive status of an individual based on digital voice recordings of NP exams.MethodWe utilized automatic speech recognition and natural language processing techniques to create an automated tool that predicts if an individual with mild cognitive impairment (MCI) will develop dementia in a fixed follow‐up window (5 years). We characterize the predictive accuracy of this approach using data from the Framingham Heart Study. First, the automated transcription of the digital voice recordings were classified into 8 main NP sub‐tests including memory assessment, naming and language skill, verbal fluency, general questions, etc. Using a large language model, only the participants’ sentences were encoded into quantitative data. These data and the participants’ characteristics, such as age, sex, ApoE4 (+/‐) genotype, and education were employed to train and test a binary classifier, predicting dementia withing 5 years after a positive MCI diagnosis. We also provide a comparison between our approach and alternative models using demographics and ApoE genotype or the Mini‐Mental State Examination (MMSE) score.ResultWe evaluated the performance of the classification task on the data containing 111 stable MCI and 103 progressive MCI subjects. Our method achieved an average accuracy of 75.5% on the held‐out test data while a baseline model using age, sex, education, and ApoE has an average accuracy of 71.8%. The sensitivity and specificity of the proposed method versus the baseline model are 78.2% vs. 81.8% and 72.7% vs. 61.8%, respectively. Our approach also significantly outperforms the model based on the MMSE score by a margin of 14% in accuracy.ConclusionWe have analyzed digital voice recordings of NP exams for participants with MCI to predict their cognitive status in 5 years. The proposed approach offers a fully automated procedure with higher accuracy, providing an opportunity to develop an accessible and easy‐to‐administer screening tool that could be easily adapted to any language.

Fathers' unmodulated prosody in child-directed speech.

Mothers and fathers modify prosodic characteristics of child-directed speech relative to adult-directed speech. Evidence suggests that mothers and fathers may differ in how they use child-directed speech as communicative partners. Thus, fathers create communicative challenges during father-child interaction that facilitate the child's adaptation to a wider potential range of interlocutors. In this paper, speech production differences are examined between mothers and fathers in child-directed speech to toddlers as compared to adult-directed speech. Using a longitudinal, large-scale design of audio recordings in naturalistic environments and automatic speech processing techniques, it was found that mothers, but not fathers, increased their fundamental frequency when addressing their toddlers. The results suggest that fathers do not modulate the prosody of their speech in the same way as mothers when communicating with their toddlers. Findings have implications for emotional and communicative practices of fathers compared to mothers and the differential role each plays in child development.

A Novel Business Scheduling Approach for Enterprises via Vision Sensing-Based Automatic Documental Information Extraction

Currently, the prevalence of various Internet intrusion technologies has brought much challenge to the enterprise management. For many core documents, the information leakage may lead to the loss of secrets of enterprises. Therefore, some core official documents in enterprises are in the format of papers, rather than electronic format. As a consequence, it is of significance to develop automatic information processing techniques for official documents in the format of papers, so as to improve the working efficiency of enterprises. In this paper, a novel business scheduling approach for enterprises via vision sensing-based automatic documental information extraction is proposed. For the first stage, the vision sensing-based optical character recognition (OCR) technique is utilized to extract textual information from official documents in the format of papers. For the second stage, the deep neural network is utilized to output business scheduling results on the basis of digital recognition contents from the first stage. Finally, the experimental simulation is also carried out to verify efficiency of the proposal.

Automatic diagnosis of COVID-19 related respiratory diseases from speech.

In this work, an attempt is made to propose an intelligent and automatic system to recognize COVID-19 related illnesses from mere speech samples by using automatic speech processing techniques. We used a standard crowd-sourced dataset which was collected by the University of Cambridge through a web based application and an android/iPhone app. We worked on cough and breath datasets individually, and also with a combination of both the datasets. We trained the datasets on two sets of features, one consisting of only standard audio features such as spectral and prosodic features and one combining excitation source features with standard audio features extracted, and trained our model on shallow classifiers such as ensemble classifiers and SVM classification methods. Our model has shown better performance on both breath and cough datasets, but the best results in each of the cases was obtained through different combinations of features and classifiers. We got our best result when we used only standard audio features, and combined both cough and breath data. In this case, we achieved an accuracy of 84% and an Area Under Curve (AUC) score of 84%. Intelligent systems have already started to make a mark in medical diagnosis, and this type of study can help better the health system by providing much needed assistance to the health workers.

Lexical and syntactic deficits analyzed via automated natural language processing: the new monitoring tool in multiple sclerosis.

Impairment of higher language functions associated with natural spontaneous speech in multiple sclerosis (MS) remains underexplored. We presented a fully automated method for discriminating MS patients from healthy controls based on lexical and syntactic linguistic features. We enrolled 120 MS individuals with Expanded Disability Status Scale ranging from 1 to 6.5 and 120 age-, sex-, and education-matched healthy controls. Linguistic analysis was performed with fully automated methods based on automatic speech recognition and natural language processing techniques using eight lexical and syntactic features acquired from the spontaneous discourse. Fully automated annotations were compared with human annotations. Compared with healthy controls, lexical impairment in MS consisted of an increase in content words (p = 0.037), a decrease in function words (p = 0.007), and overuse of verbs at the expense of noun (p = 0.047), while syntactic impairment manifested as shorter utterance length (p = 0.002), and low number of coordinate clause (p < 0.001). A fully automated language analysis approach enabled discrimination between MS and controls with an area under the curve of 0.70. A significant relationship was detected between shorter utterance length and lower symbol digit modalities test score (r = 0.25, p = 0.008). Strong associations between a majority of automatically and manually computed features were observed (r > 0.88, p < 0.001). Automated discourse analysis has the potential to provide an easy-to-implement and low-cost language-based biomarker of cognitive decline in MS for future clinical trials.

Common‐midpoint cross‐correlation stacking tomography: A 3D approach for frequency‐dependent mapping of Rayleigh waves, group and phase velocity throughout an active seismic network

AbstractShear‐wave velocity constitutes an important characterization parameter in terms of engineering properties. Several techniques of surface wave analysis have been widely adopted for building near‐surface 1D, or even 2D, S‐wave velocity models, but comparable 3D approaches, such as the surface wave tomography (SWT) method, are usually limited to global or regional seismological studies. Moreover, a reliable 3D subsurface imaging in a complex and noisy near‐surface environment becomes difficult when using the fully automated processing techniques that are commonly suggested for the SWT method. Here, a new strategy that is well adapted to geotechnical investigations is proposed to accomplish automatically the frequency‐dependent mapping of Rayleigh waves, both for group and phase velocity, throughout an active seismic network. The specific methodology can be implemented in both 2D and 3D active seismic data acquisition schemes involving both uniform and non‐uniform layouts of receivers. It combines the advantages of a newly created common‐midpoint (CMP) cross‐correlation (CC) analysis technique, with the precision of the travel‐time tomography method that is usually used in regional seismological studies. The main analysis is based on partitioning the different wave propagation directions, weighting the CC frequency–velocity analysis around one‐wavelength and stacking the CMP dispersion images. The tomographic inversion is applied to frequency‐dependent virtual travel‐times, produced from the azimuth‐dependent results of the specific analysis, in order to generate, as output, the local dispersion curves. The output of the proposed processing method can be interpreted with any preferred inversion algorithm for group and phase velocity, either individually or jointly. Simple synthetic tests together with a 3D seismic experiment in well‐known conditions, using standard refraction and multichannel analysis of surface waves (MASW) equipment, confirmed the effectiveness of the proposed methodology in detecting both lateral and vertical S‐wave velocity variations. A ‘construction site’ case study finally highlighted the potential of the new tool in a very difficult and noisy environment.

Lightweight HI source finding for next generation radio surveys

Future deep HI surveys will be essential for understanding the nature of galaxies and the content of the Universe. However, the large volume of these data will require distributed and automated processing techniques. We introduce LiSA, a set of python modules for the denoising, detection and characterization of HI sources in 3D spectral data. LiSA was developed and tested on the Square Kilometer Array Science Data Challenge 2 dataset, and contains modules and pipelines for easy domain decomposition and parallel execution. LiSA contains algorithms for 2D-1D wavelet denoising using the starlet transform and flexible source finding using null-hypothesis testing. These algorithms are lightweight and portable, needing only a few user-defined parameters reflecting the resolution of the data. LiSA also includes two convolutional neural networks developed to analyze data cubes which separate HI sources from artifacts and predict the HI source properties. All of these components are designed to be as modular as possible, allowing users to mix and match different components to create their ideal pipeline. We demonstrate the performance of the different components of LiSA on the SDC2 dataset, which is able to find 95% of HI sources with SNR >3 and accurately predict their properties.

Towards Routine Clinical Use of Dosimetry in [177Lu]Lu-PSMA Prostate Cancer Radionuclide Therapy: Current Efforts and Future Perspectives

In light of widely expanding personalized medicine applications and their impact on clinical outcomes, it is naturally befitting to explore all the dimensional aspects of personalized radionuclide therapy (RNT). Adoption of absorbed radiation dose into clinical practice in the field of RNT has been hampered by difficulties such as evidence of dose-effect correlation, technical requirements in quantitative imaging of the radiopharmaceutical, heterogeneity of methods between not only centers, but also across software, hardware and radionuclides used. Additionally, standardized agreed upon definition of outcome measures is being debated whether it be solely related to toxicity, quality of life, survival or other measures. Many clinical RNT activity administrations are still based on empirical/fixed activities, or scaled based on parameters such as body surface area. Although still challenging, a tremendous amount of progress has been made to facilitate routine clinical dosimetry with discussions regarding standardization, harmonization and automated processing techniques. This has also been aided by the development and FDA approval of several companion diagnostics allowing within the theranostic paradigm not only a crude qualitative predictive biomarker but also an objective dosimetry based predictive therapeutic biomarker. This work aims to review the literature of [177Lu]Lu-PSMA RNT, focusing on clinical trials and studies, with the goal to summarize the range of dosimetry techniques and the range of doses calculated to organs and tissues of interest from these techniques. A dosimetry method for [177Lu]Lu-PSMA RNT should be reliable, reproducible and encompassing the knowledge gained from all clinical trials evaluating it. Its translation into clinical routine practice can be achieved with the confirmation that dose calculation represents good clinical efficacy and low treatment-related toxicity. Finally, some future perspectives on the future of [177Lu]Lu-PSMA RNT are made, especially in the rapidly emerging field of artificial intelligence (AI), where deep learning may be able to play a large role in the simplification of dosimetry calculations to aid in their clinical adoption.

An application of CNN to classify barchan dunes into asymmetry classes

• CNNs can be used to extract morphometric data from barchan dunes. • Convolutional Neural Networks (CNNs) can classify barchans based only on their outlines. • There is no clear pattern regarding the use of hyperparpameters to improve the performance of CNNs. • The asymmetry of a barchan can be described using geometric morphometrics. Barchan morphometric data have been used as proxies of meteorological and topographical data in environments where this data is lacking (such as other planetary bodies), gaining insights into barchan dune field dynamics such as barchan collision and sediment dynamics, and estimating migration speeds. However, manual extraction of this data is time-consuming which can impose limits on the spatial extent and temporal frequencies of observations. Combining remotely sensed big data with automated processing techniques such as convolutional neural networks (CNNs) can therefore increase the amount of data on barchan morphology. However, such techniques have not yet been applied to barchans and their efficacy remains unknown. This study addresses this issue by evaluating the classification performance (using the ACC, F 1 -score and MCC metrics) of CNNs on several different morphometric tasks: the side of horn elongation, the magnitude of elongation, the barchans a/c ratio, and a new metric, bilateral asymmetry, which takes a more holistic view of barchan asymmetry. Specifically, bilateral asymmetry offers a means by which the total points of variation on a barchan that is used in describing barchan morphology, can be expressed with a single measure. Twelve different CNN architectures, each with different hyperparameters, are trained and tested on a sample of 90 barchan dunes. Additionally, the potential of transfer learning is assessed using the VGG16 and ResNet50 architectures. The results show that the accuracy of the CNNs can exceed 80% in some cases and that “from scratch” CNNs can match the performance obtained using transfer learning approaches.

Representation of structured data of the text genre as a technique for automatic text processing

The present article was developed in the field of Natural Language Processing and Language Studies based on a corpus compiled by computational tools. This study is based on the assumption that it is helpful to trace a close relationship between corpus generation/annotation and the assessment of the constitutive elements of the text genre source. It aims to demonstrate, through specific studies of structured data from the text genre ‘scientific article’, alternatives to automatic text processing techniques. In order to reach the intended goal, the authors created a computational model for the compilation of a linguistic, specialized Corpus, representative of the genre Scientific Article - CorpACE. The object of study includes the constitutive elements of scientific articles, marked in XML, extracted and collected from the SciELO-Scientific Electronic Library On-line database. The final product was a database obtained with information extracted and structured in XML format, which designates and identifies the markups of the genre being analyzed and is available for many tools and applications. The results demonstrate how the representation of constitutive elements of the genre can condense available information with hierarchical and dynamic processes built during the compilation. At the end of the study, it is believed that more research will be required for bringing Language Science and Computer Science closer with emphasis on NLP in the attempt to represent and manipulate linguistic knowledge in its many levels – morphological, syntactic, semantic and discursive – in order to improve implementation and manipulation of automatic text processing.

Applications of TLS and ALS in Evaluating Forest Ecosystem Services: A Southern Carpathians Case Study

Forests play an important role in biodiversity conservation, being one of the main providers of ecosystem services, according to the Economics of Ecosystems and Biodiversity. The functions and ecosystem services provided by forests are various concerning the natural capital and the socio-economic systems. Past decades of remote-sensing advances make it possible to address a large set of variables, including both biophysical parameters and ecological indicators, that characterize forest ecosystems and their capacity to supply services. This research aims to identify and implement existing methods that can be used for evaluating ecosystem services by employing airborne and terrestrial stationary laser scanning on plots from the Southern Carpathian mountains. Moreover, this paper discusses the adaptation of field-based approaches for evaluating ecological indicators to automated processing techniques based on airborne and terrestrial stationary laser scanning (ALS and TLS). Forest ecosystem functions, such as provisioning, regulation, and support, and the overall forest condition were assessed through the measurement and analysis of stand-based biomass characteristics (e.g., trees’ heights, wood volume), horizontal structure indices (e.g., canopy cover), and recruitment-mortality processes as well as overall health status assessment (e.g., dead trees identification, deadwood volume). The paper, through the implementation of the above-mentioned analyses, facilitates the development of a complex multi-source monitoring approach as a potential solution for assessing ecosystem services provided by the forest, as well as a basis for further monetization approaches.

Quantification of Arterial and Venous Morphologic Markers in Pulmonary Arterial Hypertension Using CT Imaging

BackgroundPulmonary hypertension is a heterogeneous disease, and a significant portion of patients at risk for it have CT imaging available. Advanced automated processing techniques could be leveraged for early detection, screening, and development of quantitative phenotypes. Pruning and vascular tortuosity have been previously described in pulmonary arterial hypertension (PAH), but the extent of these phenomena in arterial vs venous pulmonary vasculature and in exercise pulmonary hypertension (ePH) have not been described.Research QuestionWhat are the arterial and venous manifestations of pruning and vascular tortuosity using CT imaging in PAH, and do they also occur in ePH?Study Design and MethodsA cohort of patients with PAH and ePH and control subjects with available CT angiograms were retrospectively identified to examine the differential arterial and venous presence of pruning and tortuosity in patients with precapillary pulmonary hypertension not confounded by lung or thromboembolic disease. The pulmonary vasculature was reconstructed, and an artificial intelligence method was used to separate arteries and veins and to compute arterial and venous vascular volumes and tortuosity.ResultsA total of 42 patients with PAH, 12 patients with ePH, and 37 control subjects were identified. There was relatively lower (median [interquartile range]) arterial small vessel volume in subjects with PAH (PAH 14.7 [11.7-16.5; P < .0001]) vs control subjects (16.9 [15.6-19.2]) and venous small vessel volume in subjects with PAH and ePH (PAH 8.0 [6.5-9.6; P < .0001]; ePH, 7.8 [7.5-11.4; P = .004]) vs control subjects (11.5 [10.6-12.2]). Higher large arterial volume, however, was only observed in the pulmonary arteries (PAH 17.1 [13.6-23.4; P < .0001] vs control subjects 11.4 [8.1-15.4]). Similarly, tortuosity was higher in the pulmonary arteries in the PAH group (PAH 3.5 [3.3-3.6; P = .0002] vs control 3.2 [3.2-3.3]).InterpretationLower small distal pulmonary vascular volume, higher proximal arterial volume, and higher arterial tortuosity were observed in PAH. These can be quantified by using automated techniques from clinically acquired CT scans of patients with ePH and resting PAH.

A method for automatic shale porosity quantification using an Edge-Threshold Automatic Processing (ETAP) technique

Scanning electron microscopy (SEM) is one of the most prevalent methods used to image and quantify the pore size distribution of shale rock, critical in understanding unconventional petroleum systems and production. Generally, digital greyscale SEM images of shale are currently processed for pore quantification either by a manual drawing method, manual threshold method, automatic threshold method, edge detection or watershed methods, all of which have some limitations that impact the quality of pore extraction results. A new, Edge-Threshold Automatic Processing (ETAP) method is reported here to enable robust extraction and quantification of pore data in shale images. Image pre-treatment makes the greyscale of regions brighter than that of kerogen set to the peak value of kerogen greyscale. The pore image is subsequently obtained using an edge detection method. A discriminant function has been designed to determine the best threshold of the greyscale image to obtain the pore image. Finally, combination of both processed pore images gives the final pore image. Our new method overcomes the impact of kerogen, mineral, roughness and artificial debris caused by pre-treatment of samples, which potentially introduce errors using alternative methods. We compare our new method to a systematic manual drawing method. The processing results through ETAP provide reliable results, and gets the highest value of 0.7466 using a discriminant function Qt, compared with the automatic threshold methods, the edge detection method and watershed method. The application of the ETAP method on shale samples of the Longmaxi Formation and Qiongzhusi Formatiosn in Sichuan basin shows that samples from the Longmaxi Formation have more organic pores than that of the Qiongzhusi Formation, however a larger size of inorganic pores develop in the Qiongzhusi shale. This indicates that shale of the Longmaxi Formation has better reservoir properties and reliable preservation conditions.

Machine learning based analysis of speech dimensions in functional oropharyngeal dysphagia

Background and objective: The normal swallowing process requires a complex coordination of anatomical structures driven by sensory and cranial nerves. Alterations in such coordination cause swallowing malfunctions, namely dysphagia. The dysphagia screening methods are quite subjective and experience dependent. Bearing in mind that the swallowing process and speech production share some anatomical structures and mechanisms of neurological control, this work aims to evaluate the suitability of automatic speech processing and machine learning techniques for screening of functional dysphagia.Methods: Speech recordings were collected from 46 patients with functional oropharyngeal dysphagia produced by neurological causes, and 46 healthy controls. The dimensions of speech including phonation, articulation, and prosody were considered through different speech tasks. Specific features per dimension were extracted and analyzed using statistical tests. Machine learning models were applied per dimension via nested cross-validation. Hyperparameters were selected using the AUC - ROC as optimization criterion.Results: The Random Forest in the articulation related speech tasks retrieved the highest performance measures (AUC=0.86±0.10, sensitivity=0.91±0.12) for individual analysis of dimensions. In addition, the combination of speech dimensions with a voting ensemble improved the results, which suggests a contribution of information from different feature sets extracted from speech signals in dysphagia conditions.Conclusions: The proposed approach based on speech related models is suitable for the automatic discrimination between dysphagic and healthy individuals. These findings seem to have potential use in the screening of functional oropharyngeal dysphagia in a non-invasive and inexpensive way.

ILSA: an automated language complexity analysis tool for French

Estimating language complexity is an important aspect of educational measurement and assessment that can be used, for instance, to control for unwanted variance due to language, or to provide students with texts that are conducive to learning. Automatic language processing techniques can be used to extract various linguistic features that reflect the complexity of vocabulary and sentence structure. In this paper, we present a new tool called ILSA (Integrated Lexico-Syntactic Analyzer), which we developed for research and educational applications. We summarize how the tool works and present the types of attributes it can extract. We then apply ILSA to 600 texts used in Quebec elementary and secondary schools and analyze the correlations between the attributes and the school grade associated with the text. The results show the potential of ILSA for modeling the complexity of French texts.

Sentiment analysis for customer relationship management: an incremental learning approach

In recent years there has been a significant rethinking of corporate management, which is increasingly based on customer orientation principles. As a matter of fact, customer relationship management processes and systems are ever more popular and crucial to facing today’s business challenges. However, the large number of available customer communication stimuli coming from different (direct and indirect) channels, require automatic language processing techniques to help filter and qualify such stimuli, determine priorities, facilitate the routing of requests and reduce the response times. In this scenario, sentiment analysis plays an important role in measuring customer satisfaction, tracking consumer opinion, interacting with consumers and building customer loyalty. The research described in this paper proposes an approach based on Hierarchical Attention Networks for detecting the sentiment polarity of customer communications. Unlike other existing approaches, after initial training, the defined model can improve over time during system operation using the feedback provided by CRM operators thanks to an integrated incremental learning mechanism. The paper also describes the developed prototype as well as the dataset used for training the model which includes over 30.000 annotated items. The results of two experiments aimed at measuring classifier performance and validating the retraining mechanism are also presented and discussed. In particular, the classifier accuracy turned out to be better than that of other algorithms for the supported languages (macro-averaged f1-score of 0.89 and 0.79 for Italian and English respectively) and the retraining mechanism was able to improve the classification accuracy on new samples without degrading the overall system performance.