Search

The aim of this systematic review is to determine the current state of the art in the real-time classification of user-generated content from social media. Focus is on the identification of the main characteristics of data used for training and testing, the types of text processing and normalization that are required, the machine learning methods used most commonly, and how these methods compare to one another in terms of classification performance. Relevant studies were selected from subscription-based digital libraries, free-to-access bibliographies, and self-curated repositories and then screened for relevance with key information extracted and structured against the following facets: natural language processing (NLP) methods, data characteristics, classification methods, and evaluation results. A total of 25 studies published between 2014 and 2018 covering 15 types of classification algorithms were included in this review. Support vector machines (SVMs), Bayesian classifiers, and decision trees were the most commonly employed algorithms with recent emergence of neural network approaches. Domain-specific, application programming interface (API)-driven collection is the most prevalent origin of datasets. The reuse of previously published datasets as a means of benchmarking algorithms against other studies is also prevalent. In conclusion, there are consistent approaches taken when normalizing social media data for text mining and traditional text mining techniques are suited to the task of real-time analysis of social media.

Mammary cancer is one of the most common neoplasms of dogs, primarily bitches. While studies have been carried out identifying differing risk of mammary neoplasia in different dog breeds, few studies have reported associations between dog breeds and clinical features such as number of neoplastic lesions found in an individual case or the likelihood of lesions being benign or malignant. Such epidemiological studies are essential as a foundation for exploring potential genetic drivers of mammary tumour behaviour. Here, we have examined associations between breed, age and neuter status and the odds of a diagnosis of a mammary epithelial-origin neoplastic lesion (as opposed to any other histopathological diagnosis from a biopsied lesion) as well as the odds of a bitch presenting with either a single mammary lesion or multiple lesions, and the odds that those lesions are benign or malignant. The study population consisted of 129,258 samples from bitches, including 13,401 mammary epithelial neoplasms, submitted for histological assessment to a single histopathology laboratory between 2008 and 2021.In multivariable analysis, breed, age and neuter status were all significantly associated with the odds of a diagnosis of a mammary epithelial-origin neoplastic lesion. Smaller breeds were more likely to receive such a diagnosis. In cases diagnosed with a mammary epithelial neoplasm, these three factors were also significantly associated with the odds of diagnosis with a malignant lesion and of diagnosis with multiple lesions. Notably, while neutered animals were less likely to have a mammary epithelial neoplasm diagnosed, and were less likely to have multiple neoplasms, they were more likely to have malignant disease. Exploration of the patterns of risk of developing malignant disease, or multiple lesions, across individual breeds showed no breed with increased odds of both outcomes. Breeds with altered odds compared to the Crossbreed baseline were either at increased risk of malignant disease and decreased risk of multiple lesions, or vice versa, or they were at significantly altered odds of one outcome with no change in the other outcome. Our analysis supports the hypothesis that age, neuter status and intrinsic biological and genetic factors all combine to influence the biological heterogeneity of canine mammary neoplasia.

Idioms are multi-word expressions whose meaning cannot always be deduced from the literal meaning of constituent words. A key feature of idioms that is central to this paper is their peculiar mixture of fixedness and variability, which poses challenges for their retrieval from large corpora using traditional search approaches. These challenges hinder insights into idiom usage affecting users who are conducting linguistic research as well as those involved in language ed-ucation. To facilitate access to idioms examples taken from real-world contexts, we introduce an information retrieval system designed specifically for idioms. Given a search query that represents an idiom, typically in its canonical form, the system expands it automatically to account for the most common types of idiom variation including inflection, open slots, adjectival or adverbial modification, and passivisation. As a by-product of query expansion, other types of idiom varia-tion captured include derivation, compounding, negation, distribution across multiple clauses as well as other unforeseen types of variation. The system was implemented on top of Elasticsearch, an open-source, distributed, scalable, real-time search engine. Flexible retrieval of idioms is supported by a combination of linguistic pre-processing of the search queries, their translation into a set of query clauses written in a query language called Query DSL, and analysis, an indexing process that involves tokenisation and normalisation. Our system outperformed the phrase search in terms of recall and outperformed the keyword search in terms of precision. Out of the three, our approach was found to provide the best balance between precision and recall. By providing a fast and easy way of finding idioms in large corpora, our approach can facilitate further developments in fields such as linguistics, language education and natural language processing. Keywords: information retrieval; natural language processing; corpus linguistics; multi-word expressions; idioms

(1) Background: Aspect-based sentiment analysis (SA) is a natural language processing task, the aim of which is to classify the sentiment associated with a specific aspect of a written text. The performance of SA methods applied to texts related to health and well-being lags behind that of other domains. (2) Methods: In this study, we present an approach to aspect-based SA of drug reviews. Specifically, we analysed signs and symptoms, which were extracted automatically using the Unified Medical Language System. This information was then passed onto the BERT language model, which was extended by two layers to fine-tune the model for aspect-based SA. The interpretability of the model was analysed using an axiomatic attribution method. We performed a correlation analysis between the attribution scores and syntactic dependencies. (3) Results: Our fine-tuned model achieved accuracy of approximately 95% on a well-balanced test set. It outperformed our previous approach, which used syntactic information to guide the operation of a neural network and achieved an accuracy of approximately 82%. (4) Conclusions: We demonstrated that a BERT-based model of SA overcomes the negative bias associated with health-related aspects and closes the performance gap against the state-of-the-art in other domains.

This study systematically reviews existing approaches to unsupervised grammar induction in terms of their theoretical underpinnings, practical implementations and evaluation. Our motivation is to identify the influence of functional-cognitive schools of grammar on language processing models in computational linguistics. This is an effort to fill any gap between the theoretical school and the computational processing models of grammar induction. Specifically, the review aims to answer the following research questions: Which types of grammar theories have been the subjects of grammar induction? Which methods have been employed to support grammar induction? Which features have been used by these methods for learning? How were these methods evaluated? Finally, in terms of performance, how do these methods compare to one another? Forty-three studies were identified for systematic review out of which 33 described original implementations of grammar induction; three provided surveys and seven focused on theories and experiments related to acquisition and processing of grammar in humans. The data extracted from the 33 implementations were stratified into 7 different aspects of analysis: theory of grammar; output representation; how grammatical productivity is processed; how grammatical productivity is represented; features used for learning; evaluation strategy and implementation methodology. In most of the implementations considered, grammar was treated as a generative-formal system, autonomous and independent of meaning. The parser decoding was done in a non-incremental, head-driven fashion by assuming that all words are available for the parsing model and the output representation of the grammar learnt was hierarchical, typically a dependency or a constituency tree. However, the theoretical and experimental studies considered suggest that a usage-based, incremental, sequential system of grammar is more appropriate than the formal, non-incremental, hierarchical view of grammar. This gap between the theoretical as well as experimental studies on one hand and the computational implementations on the other hand should be addressed to enable further progress in computational grammar induction research.

In this paper we describe an automated approach to enriching sentiment analysis with idiom–based features. Specifically, we automated the development of the supporting lexico–semantic resources, which include (1) a set of rules used to identify idioms in text and (2) their sentiment polarity classifications. Our method demonstrates how idiom dictionaries, which are readily available general pedagogical resources, can be adapted into purpose–specific computational resources automatically. These resources were then used to replace the manually engineered counterparts in an existing system, which originally outperformed the baseline sentiment analysis approaches by 17 percentage points on average, taking the F–measure from 40s into 60s. The new fully automated approach outperformed the baselines by 8 percentage points on average taking the F–measure from 40s into 50s. Although the latter improvement is not as high as the one achieved with the manually engineered features, it has got the advantage of being more general in a sense that it can readily utilize an arbitrary list of idioms without the knowledge acquisition overhead previously associated with this task, thereby fully automating the original approach.

Self-supervised representation learning (SSRL) concerns the problem of learning a useful data representation without the requirement for labelled or annotated data. This representation can, in turn, be used to support solutions to downstream machine learning problems. SSRL has been demonstrated to be a useful tool in the field of geographical information science (GIS). In this article, we systematically review the existing research literature in this space to answer the following five research questions. What types of representations were learnt? What SSRL models were used? What downstream problems were the representations used to solve? What machine learning models were used to solve these problems? Finally, does using a learnt representation improve the overall performance?

Research on user experience (UX) has attracted much \ud attention from designers. Additionally, hedonic quality can help \ud designers understand user interaction (such as attractive, \ud original and innovative) when they experience a product. \ud Realising the user's interaction state is a significant step for \ud designers to optimise product design and service. Previous UX \ud modelling lacks exploration in user interaction state. Also, the \ud lack of user interaction state factor will reduce the accuracy of \ud the UX modelling. In this paper, we explore the interaction value \ud of online customer review and introduce a new approach to \ud integrating hedonic quality for UX modelling. Firstly, extracting \ud word list from online customer review; Secondly, hedonic quality\ud words are extracted from the word list and added as a hedonic \ud quality part to UX modelling; Thirdly, we compared the analysis \ud result with our previous study for the conclusion. This research \ud combines hedonic quality with UX modelling to enrich modelling \ud in the field of UX for the first time. The proposed data collection \ud method is superior to the traditional collection methods in \ud hedonic quality studies. Extracting hedonic quality factors from \ud online customer reviews can in-depth provide reflections for \ud designers to improve their product design. Furthermore, it also \ud explored the valuable relationship between UX and online \ud customer reviews to provide proactive thinking in user strategy \ud and design activities

CorCenCC (Corpws Cenedlaethol Cymraeg Cyfoes—National Corpus of Contemporary Welsh) is the first comprehensive corpus of Welsh designed to be reflective of language use across communication types, genres, speakers, language varieties (regional and social) and contexts. This article focuses on the computational infrastructure that we have designed to support data collection for CorCenCC, and the subsequent uses of the corpus which include lexicography, pedagogical research and corpus analysis. A grass-roots approach to design has been adopted, that has adapted and extended previous corpus-building and introduced new features as required for this specific context and language. The key pillars of the infrastructure include a framework that supports metadata collection, an innovative mobile application designed to collect spoken data (utilising a crowdsourcing approach), a backend database that stores curated data and a web-based interface that allows users to query the data online. A usability study was conducted to evaluate the user facing tools and to suggest directions for future improvements. Though the infrastructure was developed for Welsh language collection, its design can be re-used to support corpus development in other minority or major language contexts, broadening the potential utility and impact of this work.

Word embeddings are representations of words in a vector space that models semantic relationships between words by means of distance and direction. In this study, we adapted two existing methods, word2vec and fastText, to automatically learn Welsh word embeddings taking into account syntactic and morphological idiosyncrasies of this language. These methods exploit the principles of distributional semantics and, therefore, require a large corpus to be trained on. However, Welsh is a minoritised language, hence significantly less Welsh language data are publicly available in comparison to English. Consequently, assembling a sufficiently large text corpus is not a straightforward endeavour. Nonetheless, we compiled a corpus of 92,963,671 words from 11 sources, which represents the largest corpus of Welsh. The relative complexity of Welsh punctuation made the tokenisation of this corpus relatively challenging as punctuation could not be used for boundary detection. We considered several tokenisation methods including one designed specifically for Welsh. To account for rich inflection, we used a method for learning word embeddings that is based on subwords and, therefore, can more effectively relate different surface forms during the training phase. We conducted both qualitative and quantitative evaluation of the resulting word embeddings, which outperformed previously described word embeddings in Welsh as part of larger study including 157 languages. Our study was the first to focus specifically on Welsh word embeddings.

ObjectivesTo evaluate the feasibility of trialling taxonomy for the rehabilitation of knee conditions—ACL (TRAK-ACL), a digital health intervention that provides health information, personalised exercise plans and remote clinical support combined with treatment as usual (TAU), for people following ACL reconstruction.MethodsThe study design was a two-arm parallel randomised controlled trial (RCT). Eligible participants were English-speaking adults who had undergone ACL reconstruction within the last 12 weeks, had access to the internet and could provide informed consent. Recruitment took place at three sites in the UK. TRAK-ACL intervention was an interactive website informed by behaviour change technique combined with TAU. The comparator was TAU. Outcomes were: recruitment and retention; completeness of outcome measures at follow-up; fidelity of intervention delivery and engagement with the intervention. Individuals were randomised using a computer-generated random number sequence. Blinded assessors allocated groups and collected outcome measures.ResultsFifty-nine people were assessed for eligibility at two of the participating sites, and 51 were randomised; 26 were allocated to TRAK-ACL and 25 to TAU. Follow-up data were collected on 44 and 40 participants at 3 and 6 months, respectively. All outcome measures were completed fully at 6 months except the Client Service Receipt Inventory. Two patients in each arm did not receive the treatment they were randomised to. Engagement with TRAK-ACL intervention was a median of 5 logins (IQR 3–13 logins), over 18 weeks (SD 12.2 weeks).ConclusionTRAK-ACL would be suitable for evaluation of effectiveness in a fully powered RCT.

BACKGROUND Pharmacovigilance and safety reporting, which involve processes for monitoring the use of medicines in clinical trials, play a critical role in the identification of previously unrecognized adverse events or changes in the patterns of adverse events. OBJECTIVE This study aims to demonstrate the feasibility of automating the coding of adverse events described in the narrative section of the serious adverse event report forms to enable statistical analysis of the aforementioned patterns. METHODS We used the Unified Medical Language System (UMLS) as the coding scheme, which integrates 217 source vocabularies, thus enabling coding against other relevant terminologies such as the International Classification of Diseases–10th Revision, Medical Dictionary for Regulatory Activities, and Systematized Nomenclature of Medicine). We used MetaMap, a highly configurable dictionary lookup software, to identify the mentions of the UMLS concepts. We trained a binary classifier using Bidirectional Encoder Representations from Transformers (BERT), a transformer-based language model that captures contextual relationships, to differentiate between mentions of the UMLS concepts that represented adverse events and those that did not. RESULTS The model achieved a high F1 score of 0.8080, despite the class imbalance. This is 10.15 percent points lower than human-like performance but also 17.45 percent points higher than that of the baseline approach. CONCLUSIONS These results confirmed that automated coding of adverse events described in the narrative section of serious adverse event reports is feasible. Once coded, adverse events can be statistically analyzed so that any correlations with the trialed medicines can be estimated in a timely fashion.

Aspect-based sentiment analysis is a natural language processing task whose aim is to automatically classify the sentiment associated with a specific aspect of a written text. In this study, we propose a novel model for aspect-based sentiment analysis, which exploits the dependency parse tree of a sentence using graph convolution to classify the sentiment of a given aspect. To evaluate this model in the domain of health and well-being, where this task is biased towards negative sentiment, we used a corpus of drug reviews. Specific aspects were grounded in the Unified Medical Language System, a large repository of inter-related biomedical concepts and the corresponding terminology. Our experiments demonstrated that graph convolution approach outperforms standard deep learning architectures on the task of aspect-based sentiment analysis. Moreover, graph convolution over dependency parse trees (F-score of 0.8179) outperforms the same approach over a flat sequence representation of sentences (F-score of 0.7332). These results bring the performance of sentiment analysis\udin health and well-being in line with the state of the art in other domains. \ud\udKeywords: sentiment analysis, natural language processing, dependency parsing, neural network, graph convolutional network.

Background Musculoskeletal conditions are managed within primary care, but patients can be referred to secondary care if a specialist opinion is required. The ever-increasing demand for health care resources emphasizes the need to streamline care pathways with the ultimate aim of ensuring that patients receive timely and optimal care. Information contained in referral letters underpins the referral decision-making process but is yet to be explored systematically for the purposes of treatment prioritization for musculoskeletal conditions. Objective This study aims to explore the feasibility of using natural language processing and machine learning to automate the triage of patients with musculoskeletal conditions by analyzing information from referral letters. Specifically, we aim to determine whether referral letters can be automatically assorted into latent topics that are clinically relevant, that is, considered relevant when prescribing treatments. Here, clinical relevance is assessed by posing 2 research questions. Can latent topics be used to automatically predict treatment? Can clinicians interpret latent topics as cohorts of patients who share common characteristics or experiences such as medical history, demographics, and possible treatments? Methods We used latent Dirichlet allocation to model each referral letter as a finite mixture over an underlying set of topics and model each topic as an infinite mixture over an underlying set of topic probabilities. The topic model was evaluated in the context of automating patient triage. Given a set of treatment outcomes, a binary classifier was trained for each outcome using previously extracted topics as the input features of the machine learning algorithm. In addition, a qualitative evaluation was performed to assess the human interpretability of topics. Results The prediction accuracy of binary classifiers outperformed the stratified random classifier by a large margin, indicating that topic modeling could be used to predict the treatment, thus effectively supporting patient triage. The qualitative evaluation confirmed the high clinical interpretability of the topic model. Conclusions The results established the feasibility of using natural language processing and machine learning to automate triage of patients with knee or hip pain by analyzing information from their referral letters.

BACKGROUND Musculoskeletal conditions are managed within primary care, but patients can be referred to secondary care if a specialist opinion is required. The ever-increasing demand for health care resources emphasizes the need to streamline care pathways with the ultimate aim of ensuring that patients receive timely and optimal care. Information contained in referral letters underpins the referral decision-making process but is yet to be explored systematically for the purposes of treatment prioritization for musculoskeletal conditions. OBJECTIVE This study aims to explore the feasibility of using natural language processing and machine learning to automate the triage of patients with musculoskeletal conditions by analyzing information from referral letters. Specifically, we aim to determine whether referral letters can be automatically assorted into latent topics that are clinically relevant, that is, considered relevant when prescribing treatments. Here, clinical relevance is assessed by posing 2 research questions. Can latent topics be used to automatically predict treatment? Can clinicians interpret latent topics as cohorts of patients who share common characteristics or experiences such as medical history, demographics, and possible treatments? METHODS We used latent Dirichlet allocation to model each referral letter as a finite mixture over an underlying set of topics and model each topic as an infinite mixture over an underlying set of topic probabilities. The topic model was evaluated in the context of automating patient triage. Given a set of treatment outcomes, a binary classifier was trained for each outcome using previously extracted topics as the input features of the machine learning algorithm. In addition, a qualitative evaluation was performed to assess the human interpretability of topics. RESULTS The prediction accuracy of binary classifiers outperformed the stratified random classifier by a large margin, indicating that topic modeling could be used to predict the treatment, thus effectively supporting patient triage. The qualitative evaluation confirmed the high clinical interpretability of the topic model. CONCLUSIONS The results established the feasibility of using natural language processing and machine learning to automate triage of patients with knee or hip pain by analyzing information from their referral letters.

Mobile applications based on machine learning are reshaping and affecting many aspects of our lives. Implementing machine learning on mobile devices faces various challenges, including computational power, energy, latency, low memory, and privacy risks. In this article, we investigate the current state of implementing machine learning for mobile applications, providing an overview of five architectures commonly used for this purpose and the ways in which they address the given challenges. We also discuss their pros and cons, providing recommendations for each architecture. Additionally, we review recent studies, popular toolkits, cloud services, and platforms supporting machine learning as a service. This survey will, therefore, bring mobile developers up to speed on the latest trends in implementing machine learning for mobile applications.

Background Clinical narratives represent the main form of communication within health care, providing a personalized account of patient history and assessments, and offering rich information for clinical decision making. Natural language processing (NLP) has repeatedly demonstrated its feasibility to unlock evidence buried in clinical narratives. Machine learning can facilitate rapid development of NLP tools by leveraging large amounts of text data. Objective The main aim of this study was to provide systematic evidence on the properties of text data used to train machine learning approaches to clinical NLP. We also investigated the types of NLP tasks that have been supported by machine learning and how they can be applied in clinical practice. Methods Our methodology was based on the guidelines for performing systematic reviews. In August 2018, we used PubMed, a multifaceted interface, to perform a literature search against MEDLINE. We identified 110 relevant studies and extracted information about text data used to support machine learning, NLP tasks supported, and their clinical applications. The data properties considered included their size, provenance, collection methods, annotation, and any relevant statistics. Results The majority of datasets used to train machine learning models included only hundreds or thousands of documents. Only 10 studies used tens of thousands of documents, with a handful of studies utilizing more. Relatively small datasets were utilized for training even when much larger datasets were available. The main reason for such poor data utilization is the annotation bottleneck faced by supervised machine learning algorithms. Active learning was explored to iteratively sample a subset of data for manual annotation as a strategy for minimizing the annotation effort while maximizing the predictive performance of the model. Supervised learning was successfully used where clinical codes integrated with free-text notes into electronic health records were utilized as class labels. Similarly, distant supervision was used to utilize an existing knowledge base to automatically annotate raw text. Where manual annotation was unavoidable, crowdsourcing was explored, but it remains unsuitable because of the sensitive nature of data considered. Besides the small volume, training data were typically sourced from a small number of institutions, thus offering no hard evidence about the transferability of machine learning models. The majority of studies focused on text classification. Most commonly, the classification results were used to support phenotyping, prognosis, care improvement, resource management, and surveillance. Conclusions We identified the data annotation bottleneck as one of the key obstacles to machine learning approaches in clinical NLP. Active learning and distant supervision were explored as a way of saving the annotation efforts. Future research in this field would benefit from alternatives such as data augmentation and transfer learning, or unsupervised learning, which do not require data annotation.

Background Referral letters from primary care contain a large amount of information that could be used to improve the appropriateness of the referral pathway for individuals seeking specialist opinion for knee or hip pain. The primary aim of this study was to evaluate the content of the referral letters to identify information that can independently predict an optimal care pathway. Methods Using a prospective longitudinal design, a convenience sample of patients with hip or knee pain were recruited from orthopaedic, specialist general practice and advanced physiotherapy practitioner clinics. Individuals completed a Knee or hip Osteoarthritis Outcome Score at initial consultation and after 6 months. Participant demographics, body mass index, medication and co-morbidity data were extracted from the referral letters. Free text of the referral letters was mapped automatically onto the Unified Medical Language System to identify relevant clinical variables. Treatment outcomes were extracted from the consultation letters. Each outcome was classified as being an optimal or sub-optimal pathway, where an optimal pathway was defined as the one that results in the right treatment at the right time. Logistic regression was used to identify variables that were independently associated with an optimal pathway. Results A total of 643 participants were recruited, 419 (66.7%) were classified as having an optimal pathway. Variables independently associated with having an optimal care pathway were lower body mass index (OR 1.0, 95% CI 0.9 to 1.0 p = 0.004), named disease or syndromes (OR 1.8, 95% CI 1.1 to 2.8, p = 0.02) and taking pharmacologic substances (OR 1.8, 95% CI 1.0 to 3.3, p = 0.02). Having a single diagnostic procedure was associated with a suboptimal pathway (OR 0.5, 95% CI 0.3 to 0.9 p p Conclusion Over 30% of patients followed sub-optimal care pathway, which represents potential inefficiency and wasted healthcare resource. A core data set including body mass index should be considered as this was a predictor of optimal care and patient rated pain and function.