Automatic Language Processing Research Articles

Advancing Accessibility through Automatic Speech Recognition and NLP Integration

Abstract: The integration of Automatic Speech Recognition (ASR) and Natural Language Processing (NLP) technologies has the potential to revolutionize accessibility and inclusive communication. This article explores the fundamentals of ASR, including acoustic modeling, language modeling, and speech signal processing techniques, and discusses the challenges posed by language diversity in developing accurate ASR systems. The advent of ASR technology has opened up numerous possibilities for enhancing accessibility across various domains, such as assistive technology, education, healthcare, and media. The integration of ASR with NLP techniques enables the processing and analysis of spoken language data, leading to the development of voice-enabled virtual assistants, conversational AI systems, and cross-lingual communication tools. However, several challenges remain, including the need for robust and accurate ASR systems, privacy and security concerns, and ethical considerations in the development and deployment of these technologies. The article also presents future directions, such as the integration of ASR and NLP with emotion recognition and sentiment analysis, advances in deep learning techniques, and the application of these technologies in healthcare and accessibility. Overall, the integration of ASR and NLP holds immense promise for creating more natural, empathetic, and inclusive communication systems, but their development and deployment must be approached with care to ensure fairness, transparency, and user privacy.

Experiences of Women With Medical Abortion Care Reflected in Social Media (VEILLE Study): Noninterventional Retrospective Exploratory Infodemiology Study.

Abortion (also known as termination of pregnancy) is an essential element of women's reproductive health care. Feedback from women who underwent medical termination of pregnancy about their experience is crucial to help practitioners identify women's needs and develop necessary tools to improve the abortion care process. However, the collection of this feedback is quite challenging. Social media offer anonymity for women who share their abortion experience. This exploratory infodemiology study aimed to analyze, through French social media posts, personal medical symptoms and the different experiences and information dynamics associated with the medical abortion process. A retrospective study was performed by analyzing posts geolocated in France and published from January 1, 2017, to November 30, 2021. Posts were extracted from all French-language general and specialized publicly available web forums using specific keywords. Extracted messages were cleaned and pseudonymized. Automatic natural language processing methods were used to identify posts from women having experienced medical abortion. Biterm topic modeling was used to identify the main discussion themes and the Medical Dictionary for Regulatory Activities was used to identify medical terms. Encountered difficulties were explored using qualitative research methods until the saturation of concepts was reached. Analysis of 5398 identified posts (3409 users) led to the identification of 9 major topics: personal experience (n=2413 posts, 44.7%), community support (n=1058, 19.6%), pain and bleeding (n=797, 14.8%), psychological experience (n=760, 14.1%), questioned efficacy (n=410, 7.6%), social pressure (n=373, 6.9%), positive experiences (n=257, 4.8%), menstrual cycle disorders (n=107, 2%), and reported inefficacy (n=104, 1.9%). Pain, which was mentioned in 1627 (30.1%) of the 5398 posts by 1024 (30.0%) of the 3409 users, was the most frequently reported medical term. Pain was considered severe to unbearable in 24.5% of the cases (399 of the 1627 posts). Lack of information was the most frequently reported difficulty during and after the process. Our findings suggest that French women used social media to share their experiences, offer and find support, and provide and receive information regarding medical abortion. Infodemiology appears to be a useful tool to obtain women's feedback, therefore offering the opportunity to enhance care in women undergoing medical abortion.

Mobile/VR/Robotics/IoT-Based Chatbots and Intelligent Personal Assistants for Social Inclusion

Chatbots are dialogue systems that utilize computational linguistics (CL), including automatic speech recognition (ASR) and natural language processing (NLP). Artificially intelligent conversational agents combine personalization, interoperability, and scalability with the aim of promoting safe information monitoring, management, and retrieval. AI chatbots offer user-friendly engagement for innovative e-learning, care assistance, and multilingual digital content creation, promoting inclusiveness and AI-enhanced communication. They are a breakthrough for future societies and economies as they can offer cost-effective, tailor-made, and instant exchange of knowledge and skills, including on-site assistance, health monitoring, and e-consultation. AI chatbots are constantly improving to the benefit of their users. Their economic and social impact is expected to rise as individuals, especially vulnerable populations, are trained to use them for lifelong learning, decision-making, and problem-solving.

Discontinuous Arabic frozen expressions modelization and implementation

Frozen expressions hold significant importance in the field of natural language processing, attracting considerable attention from researchers across various languages in recent years. The Arabic language, in particular, boasts a wealth of frozen expressions inherited from the pre-Islamic and early Islamic periods, with persistent usage to the present day. This linguistic richness has motivated researchers to systematically collect, classify, and elucidate these expressions. Various classifications have emerged, addressing aspects such as continuity, discontinuity, allowance for variations, and restriction from variations. Our aim is to produce lexicon-grammar tables of discontinuous Arabic frozen expressions and implement them. Our approach involves the meticulous collection and study of these expressions, followed by the transformation of their lexicon-grammar tables into dictionaries and syntactic grammars within the NooJ platform. This methodology allows us to recognize and annotate these expressions in texts and corpora, even when they exhibit discontinuity. Such recognition has the potential to address several challenges in automatic natural language processing, including the area of automatic translation.

Multilingual Framework for Risk Assessment and Symptom Tracking (MRAST).

The importance and value of real-world data in healthcare cannot be overstated because it offers a valuable source of insights into patient experiences. Traditional patient-reported experience and outcomes measures (PREMs/PROMs) often fall short in addressing the complexities of these experiences due to subjectivity and their inability to precisely target the questions asked. In contrast, diary recordings offer a promising solution. They can provide a comprehensive picture of psychological well-being, encompassing both psychological and physiological symptoms. This study explores how using advanced digital technologies, i.e., automatic speech recognition and natural language processing, can efficiently capture patient insights in oncology settings. We introduce the MRAST framework, a simplified way to collect, structure, and understand patient data using questionnaires and diary recordings. The framework was validated in a prospective study with 81 colorectal and 85 breast cancer survivors, of whom 37 were male and 129 were female. Overall, the patients evaluated the solution as well made; they found it easy to use and integrate into their daily routine. The majority (75.3%) of the cancer survivors participating in the study were willing to engage in health monitoring activities using digital wearable devices daily for an extended period. Throughout the study, there was a noticeable increase in the number of participants who perceived the system as having excellent usability. Despite some negative feedback, 44.44% of patients still rated the app's usability as above satisfactory (i.e., 7.9 on 1-10 scale) and the experience with diary recording as above satisfactory (i.e., 7.0 on 1-10 scale). Overall, these findings also underscore the significance of user testing and continuous improvement in enhancing the usability and user acceptance of solutions like the MRAST framework. Overall, the automated extraction of information from diaries represents a pivotal step toward a more patient-centered approach, where healthcare decisions are based on real-world experiences and tailored to individual needs. The potential usefulness of such data is enormous, as it enables better measurement of everyday experiences and opens new avenues for patient-centered care.

Lexicon-grammar tables standardization and implementation

<span>The lexicon-grammar approach is a very important linguistic approach in automatic natural language processing (NLP). It allows for the description of the lexicon of the language through readable and intuitive tables for human manual editing. However, the automatic use of the lexicon-grammar tables in the automatic NLP platforms remains difficult, given the incompatibility between the codes used to represent the properties in the lexicon-grammar tables and those used to represent the properties in the automatic NLP platforms. In this work, we present our method of standardizing the lexicon-grammar tables for the French language, since they constitute very rich lexical, syntactic, and semantic linguistic resources. First, we standardize their properties so that they can be compatible with those used in the NLP platforms. Then, to implement the standardized tables, we used a linguistic platform such as NooJ. For that, we describe the process of integrating these tables into this platform through the automatic generation of the dictionaries from these tables. Finally, to test the efficiency of the generated dictionaries, we create for some of them syntactic grammars that take into account all the grammatical, syntactic, and semantic information contained in the dictionaries.</span>

AlgBERT: Automatic Construction of Annotated Corpus for Sentiment Analysis in Algerian Dialect

Nowadays, sentiment analysis is one of the most crucial research fields of Natural Language Processing (NLP), and it is widely applied in a variety of applications such as marketing and politics. However, the Arabic language still lacks sufficient language resources to enable the tasks of opinion and emotion analysis comparing to other language such as English. Additionally, manual annotation requires a lot of effort and time. In this article, we address this problem and propose a novel automated annotation platform for sentiment analysis called AlgBERT by providing annotated corpus and using deep learning technology that includes many automatic natural language processing algorithms, which is the basis for text classification and opinion analysis. We suggest using BERT model as a method; it is the abbreviation of Bidirectional Encoder Representations from Transformers, as it is one of the most effective technologies in terms of results in different world languages. We used around of 54K comments collected from social networking (Twitter, YouTube) written in Arabic and Algerian dialects. Our AlgBERT system obtained excellent results with an accuracy of 91.04%, and this is considered as one of the best results for opinion analysis in Algerian dialect.

Machine learning-based donor permission extraction from informed consent documents

BackgroundWith more clinical trials are offering optional participation in the collection of bio-specimens for biobanking comes the increasing complexity of requirements of informed consent forms. The aim of this study is to develop an automatic natural language processing (NLP) tool to annotate informed consent documents to promote biorepository data regulation, sharing, and decision support. We collected informed consent documents from several publicly available sources, then manually annotated them, covering sentences containing permission information about the sharing of either bio-specimens or donor data, or conducting genetic research or future research using bio-specimens or donor data.ResultsWe evaluated a variety of machine learning algorithms including random forest (RF) and support vector machine (SVM) for the automatic identification of these sentences. 120 informed consent documents containing 29,204 sentences were annotated, of which 1250 sentences (4.28%) provide answers to a permission question. A support vector machine (SVM) model achieved a F-1 score of 0.95 on classifying the sentences when using a gold standard, which is a prefiltered corpus containing all relevant sentences.ConclusionsThis study provides the feasibility of using machine learning tools to classify permission-related sentences in informed consent documents.

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.

Twitch as a privileged locus to analyze young people’s attitudes in the climate change debate: a quantitative analysis

Social media platforms are becoming an increasingly important venue for public debate. Twitch, in particular, is a popular streaming platform that targets young adults. Originally created for online video game players, the platform also includes other types of content, such as politics, leisure, and science. Here, we present a study that examines climate change discourse on this understudied platform. Unlike previous studies, this work does not only focus on users’ reactions to streamer discourse but, using an ad-hoc methodology, also analyzes the content of the videos. Indeed, an added value of this research is the development of an objective evaluation based on automatic speech recognition (ASP) and natural language processing (NLP). We found that Twitch is an emergent locus for climate discussion with a thriving community of young users interested in the topic. Our findings challenge the understanding of social media discourse on climate change and suggest that platform architecture and intended commercial goals do not play a decisive role in shaping the topics circulating on them. In contrast, our findings support the idea that public discussion on climate change percolates through technology. In other words, the public debate finds its way across existing channels rather than being constrained by them. The research also contributes to the literature by expanding the empirical base for the study of online communication about climate change, especially among young audiences.

A State of the Art Review on Natural Language Processing applied to the Malagasy Language

Based on the growing mass of information of all kinds to be processed and to facilitate human/machine dialogue, research teams and language industries are still developing applications integrating automatic natural language processing techniques. Natural Language Processing (NLP) is a discipline on the border of linguistics and computer science that concerns the application of computer programs and techniques to all aspects of human language. NLP is an area of research that is still open. It has extended its applications in various fields and applied them in several languages worldwide. This paper will review the NLP methods and resources that have been performed on the Malagasy language only. Recently, work on the automatic processing of the Malagasy language occurs even if the language is part of the group of under-endowed languages. This will allow researchers who are interested in the Malagasy language to update themselves in terms of the work already carried out, and the challenges that may be envisaged in future.

DETERMINING THE MORPHOLOGICAL CLASS OF A WORD DURING THE AUTOMATIC NATURAL LANGUAGE PROCESSING

The article considers a mandatory component of the linguistic provision of any system of automatic natural language processing, i.e. automatic morphological analysis, the tasks of which include: determining for each text unit its place in the morphological system of the corresponding language; identification of word forms of the lexeme. As a result of automatic morphological analysis, each word form of the text is assigned a tag for the part of speech and the meaning of the grammatical categories (gender, number, case, aspect, tense, person, etc.). The nature of this information, its volume, and the methods used to establish morphological information depend on the purpose of the research, within which automatic analysis is carried out with the focus on the nature of the analyzed texts. Morphological analysis is present at all stages of text analysis, because neither morphemic, nor syntactic, nor semantic analysis can be performed without parts-of-speech tagging. With automatic syntactic analysis, only if lexical-grammatical and grammatical information is available for each word form, it is possible to syntactically bind word forms in a sentence. Morphological features of text units further become a tool for researching the relationship between vocabulary and grammar and the use in speech; between paradigmatics (in the aspect of consideration of case forms of declinable words) and syntagmatics (in the aspect of linear relationships of words, text coherence).
 The article examines the difficulties that prevent unifying the process of tagging text units, namely lexical-grammatical homonymy, ambiguity of grammatical forms, polysemy. The study considers approaches to resolving morphological ambiguity based on the context analysis of the ambiguous word, which can be divided into statistical and rule-based. Rules can be compiled manually or derived from marked-up corpora. Statistical methods are based on quantitative indicators in large labelled corpora. Morphological ambiguity resolution methods are usually applied after primary tagging, which is usually done using dictionaries.
 The article also provides a morphemic analysis algorithm for automatic morphological analysis.

"Mm-hm," "Uh-uh": are non-lexical conversational sounds deal breakers for the ambient clinical documentation technology?

Ambient clinical documentation technology uses automatic speech recognition (ASR) and natural language processing (NLP) to turn patient-clinician conversations into clinical documentation. It is a promising approach to reducing clinician burden and improving documentation quality. However, the performance of current-generation ASR remains inadequately validated. In this study, we investigated the impact of non-lexical conversational sounds (NLCS) on ASR performance. NLCS, such as Mm-hm and Uh-uh, are commonly used to convey important information in clinical conversations, for example, Mm-hm as a "yes" response from the patient to the clinician question "are you allergic to antibiotics?" In this study, we evaluated 2 contemporary ASR engines, Google Speech-to-Text Clinical Conversation ("Google ASR"), and Amazon Transcribe Medical ("Amazon ASR"), both of which have their language models specifically tailored to clinical conversations. The empirical data used were from 36 primary care encounters. We conducted a series of quantitative and qualitative analyses to examine the word error rate (WER) and the potential impact of misrecognized NLCS on the quality of clinical documentation. Out of a total of 135647 spoken words contained in the evaluation data, 3284 (2.4%) were NLCS. Among these NLCS, 76 (0.06% of total words, 2.3% of all NLCS) were used to convey clinically relevant information. The overall WER, of all spoken words, was 11.8% for Google ASR and 12.8% for Amazon ASR. However, both ASR engines demonstrated poor performance in recognizing NLCS: the WERs across frequently used NLCS were 40.8% (Google) and 57.2% (Amazon), respectively; and among the NLCS that conveyed clinically relevant information, 94.7% and 98.7%, respectively. Current ASR solutions are not capable of properly recognizing NLCS, particularly those that convey clinically relevant information. Although the volume of NLCS in our evaluation data was very small (2.4% of the total corpus; and for NLCS that conveyed clinically relevant information: 0.06%), incorrect recognition of them could result in inaccuracies in clinical documentation and introduce new patient safety risks.

Deep Learning Transformer Models for Building a Comprehensive and Real-time Trauma Observatory: Development and Validation Study.

Public health surveillance relies on the collection of data, often in near-real time. Recent advances in natural language processing make it possible to envisage an automated system for extracting information from electronic health records. To study the feasibility of setting up a national trauma observatory in France, we compared the performance of several automatic language processing methods in a multiclass classification task of unstructured clinical notes. A total of 69,110 free-text clinical notes related to visits to the emergency departments of the University Hospital of Bordeaux, France, between 2012 and 2019 were manually annotated. Among these clinical notes, 32.5% (22,481/69,110) were traumas. We trained 4 transformer models (deep learning models that encompass attention mechanism) and compared them with the term frequency-inverse document frequency associated with the support vector machine method. The transformer models consistently performed better than the term frequency-inverse document frequency and a support vector machine. Among the transformers, the GPTanam model pretrained with a French corpus with an additional autosupervised learning step on 306,368 unlabeled clinical notes showed the best performance with a micro F1-score of 0.969. The transformers proved efficient at the multiclass classification of narrative and medical data. Further steps for improvement should focus on the expansion of abbreviations and multioutput multiclass classification.

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.

An Efficient Hybrid Model for Arabic Text Recognition

In recent years, Deep Learning models have become indispensable in several fields such as computer vision, automatic object recognition, and automatic natural language processing. The implementation of a robust and efficient handwritten text recognition system remains a challenge for the research community in this field, especially for the Arabic language, which, compared to other languages, has a dearth of published works. In this work, we presented an efficient and new system for offline Arabic handwritten text recognition. Our new approach is based on the combination of a Convolutional Neural Network (CNN) and a Bidirectional Long-Term Memory (BLSTM) followed by a Connectionist Temporal Classification layer (CTC). Moreover, during the training phase of the model, we introduce an algorithm of data augmentation to increase the quality of data. Our proposed approach can recognize Arabic handwritten texts without the need to segment the characters, thus overcoming several problems related to this point. To train and test (evaluate) our approach, we used two Arabic handwritten text recognition databases, which are IFN/ENIT and KHATT. The Experimental results show that our new approach, compared to other methods in the literature, gives better results.

Towards Intelligent Aviation: The Integration of Voice Assistants in Autonomous Aircraft

This paper explores the integration of a virtual assistant into an autonomous aircraft, focusing on the utilization of cutting-edge technologies for automatic speech recognition, speech synthesis, and natural language processing. By leveraging the power of Edge AI, particularly NVIDIA Jetson devices, we demonstrate how this amalgamation of technologies creates a lightweight, mobile, and efficient voice assistant system. We introduce a novel approach to enhance the operational capabilities of autonomous aircraft, facilitating human-machine interaction and enhancing overall user experience. This research paves the way for the application of edge-based voice assistants in autonomous vehicles, offering a glimpse into the future of intelligent and responsive in-flight systems.

Квантитативная онтология и база данных русских предлогов

Automatic natural language processing that goes beyond the bag-of-words model requires an understanding of semantic and syntactic relationships between language components. This work attempts to address the following issues: 1) describe a set of Russian prepositional structures as an interconnected system; 2) collect corpus-based statistics reflecting preposition meanings and their hierarchy; and 3) describe meanings of prepositions acting as grammatical connectors for content words. Preposition meanings are described using categories based on G. A. Zolotova’s concept of syntaxeme.&#x0D; The results obtained are based on word frequencies extracted from corpora of contemporary Russian texts. This work describes the key objectives, approaches, methods, novelty, and findings of the research. The key findings include grammar and ontology of the main Russian prepositions, research into secondary prepositions (prepositions derived from content words), preposition database with a network interface, and a set of automated procedures aiding in ‘working out’ the meaning of a preposition.

Tool for Parsing Important Data from Web Pages

This paper discusses the tool for the main text and image extraction (extracting and parsing the important data) from a web document. This paper describes our proposed algorithm based on the Document Object Model (DOM) and natural language processing (NLP) techniques and other approaches for extracting information from web pages using various classification techniques such as support vector machine, decision tree techniques, naive Bayes, and K-nearest neighbor. The main aim of the developed algorithm was to identify and extract the main block of a web document that contains the text of the article and the relevant images. The algorithm on a sample of 45 web documents of different types was applied. In addition, the issue of web pages, from the structure of the document to the use of the Document Object Model (DOM) for their processing, was analyzed. The Document Object Model was used to load and navigation of the document. It also plays an important role in the correct identification of the main block of web documents. The paper also discusses the levels of natural language. These methods of automatic natural language processing help to identify the main block of the web document. In this way, the all-textual parts and images from the main content of the web document were extracted. The experimental results show that our method achieved a final classification accuracy of 88.18%.

A Systematic Review of User Studies as a Basis for the Design of Systems for Automatic Sign Language Processing

Deaf persons, whether or not they are sign language users, make up one of various existing marginalized populations that historically have been socially and politically underrepresented. Unfortunately, this also happens in technology design. Conducting user studies in which marginalized populations are represented is a step towards guaranteeing their right to participate in choices and decisions that are made for, with, and by them. This article presents and discusses results from a Systematic Literature Review (SLR) of user studies in the design of systems for Automatic Sign Language Processing (ASLP). Following our SLR protocol, from 2,486 papers initially found, we applied inclusion and exclusion criteria to finally select 37 papers in our review. We excluded publications that were not full papers, were not related to our main topic of interest, or that reported results that had been updated by more recent papers. All the selected papers focus on user studies as a basis for the design of three major aspects of ASLP: generation (ASLG), recognition (ASLR), and translation (ASLT). With regard to our specific area of interest, we analyzed four areas related to our research questions: goals and research methods, types of user involvement in the interaction design life cycle, cultural and collaborative aspects, and other lessons learned from the primary studies under review. Salient findings from our analysis show that numerical scale questionnaires are the most frequently used research instruments, co-designing ASLP systems with sign language users is not a common practice (as potential users are included mostly in the evaluation phase), and only seldom are Deaf persons who are sign language users included as members of research teams. These findings point to the need of conducting more inclusive and qualitative research for, with and by Deaf persons who are sign language users.