Data Brief. 2021 Feb 24;35:106841. doi: 10.1016/j.dib.2021.106841. eCollection 2021 Apr.

ABSTRACT

This article presents data that are further analyzed and interpreted in "Shouting at Each Other into the Void: A Semantic Network Analysis of Vaccine Hesitance and Support in Online Discourse Regarding California Law SB277" [1]. This research modified snowball sampling, a technique usually used to generate chains of informants that illuminate the structure of social networks, to collect digital documents following a chain of web links and recommendations, thus illuminating the underlying social, technical, and linguistic structure of online discourse. The resulting documents were manually coded according to the attitude towards vaccines they represented and/or the position they took with regard to California Senate Bill 277, a vaccine mandate policy that banned all nonmedical exemptions from school immunization requirements. Each attitude category, as well as the dataset as a whole, was subjected to quantitative linguistic analysis to identify key words and phrases in the data according to the frequency with which they appeared. A combination of that technique and semantic network analysis were used to generate clusters of related words that could be used for qualitative and narrative analysis, as detailed in the companion paper. The data collection and analysis processes described here will be of use to researchers conducting mixed-method analysis of online discourse who want their data to reflect the potential information and digital resources available to individuals who attempt to inform themselves about a particular topic using Internet searches. The data presented here could be useful for anyone seeking deeper insight into the linguistic and narrative patterns surrounding online debates about vaccination, controversial government policies, or both.

PMID:33748356 | PMC:PMC7966830 | DOI:10.1016/j.dib.2021.106841

Comput Ind Eng. 2021 Mar 15:107236. doi: 10.1016/j.cie.2021.107236. Online ahead of print.

ABSTRACT

The human suffering from diseases caused by life-threatening viruses such as SARS, Ebola, and COVID-19 motivated many of us to study and discover the best means to harness the potential of data integration to assist clinical researchers to curb these viruses. Integrating patients data with clinical trials data is enormously promising as it provides a comprehensive knowledge base that accelerates the clinical research response-ability to tackle emerging infectious disease outbreaks. This work introduces EMR2vec, a platform that customises advanced NLP, machine learning and semantic web techniques to link potential patients to suitable clinical trials. Linking these two different but complementary datasets allows clinicians and researchers to compare patients to clinical research opportunities or to automatically select patients for personalized clinical care. The platform derives a 'bag of medical terms' (BoMT) from eligibility criteria by normalizing extracted entities through SNOMED-CT ontology. With the usage of BoMT, an ontological reasoning method is proposed to represent EMR and clinical trials in a vector space model. The platform presents a matching process that reduces vector dimensionality using a neural network, then applies orthogonality projection to measure the similarity between vectors. Finally, the proposed EMR2vec platform is evaluated with an extendable prototype based on Big data tools.

PMID:33746344 | PMC:PMC7959675 | DOI:10.1016/j.cie.2021.107236

J Biomed Inform. 2021 Mar 15:103733. doi: 10.1016/j.jbi.2021.103733. Online ahead of print.

ABSTRACT

The context of medical conditions is an important feature to consider when processing clinical narratives. NegEx and its extension ConText became the most well-known rule-based systems that allow determining whether a medical condition is negated, historical or experienced by someone other than the patient in English clinical text. In this paper, we present a French adaptation and enrichment of FastContext which is the most recent, n-trie engine-based implementation of the ConText algorithm. We compiled an extensive list of French lexical cues by automatic and manual translation and enrichment. To evaluate French FastContext, we manually annotated the context of medical conditions present in two types of clinical narratives: (i)death certificates and (ii)electronic health records. Results show good performance across different context values on both types of clinical notes (on average 0.93 and 0.86 F1, respectively). Furthermore, French FastContext outperforms previously reported French systems for negation detection when compared on the same datasets and it is the first implementation of contextual temporality and experiencer identification reported for French. Finally, French FastContext has been implemented within the SIFR Annotator: a publicly accessible Web service to annotate French biomedical text data (http://bioportal.lirmm.fr/annotator). To our knowledge, this is the first implementation of a Web-based ConText-like system in a publicly accessible platform allowing non-natural-language-processing experts to both annotate and contextualize medical conditions in clinical notes.

PMID:33737205 | DOI:10.1016/j.jbi.2021.103733

PLoS One. 2021 Mar 18;16(3):e0248484. doi: 10.1371/journal.pone.0248484. eCollection 2021.

ABSTRACT

INTRODUCTION: People with ulnar, radial or median nerve injuries can present significant impairment of their sensory and motor functions. The prescribed treatment for these conditions often includes electrophysical therapies, whose effectiveness in improving symptoms and function is a source of debate. Therefore, this systematic review aims to provide an integrative overview of the efficacy of these modalities in sensorimotor rehabilitation compared to placebo, manual therapy, or between them.

METHODS: We conducted a systematic review according to PRISMA guidelines. We perform a literature review in the following databases: Biomed Central, Ebscohost, Lilacs, Ovid, PEDro, Sage, Scopus, Science Direct, Semantic Scholar, Taylor & Francis, and Web of Science, for the period 1980-2020. We include studies that discussed the sensorimotor rehabilitation of people with non-degenerative ulnar, radial, or median nerve injury. We assessed the quality of the included studies using the Risk of Bias Tool described in the Cochrane Handbook of Systematic Reviews of Interventions and the risk of bias across studies with the GRADE approach described in the GRADE Handbook.

RESULTS: Thirty-eight studies were included in the systematic review and 34 in the meta-analysis. The overall quality of evidence was rated as low or very low according to GRADE criteria. Low-level laser therapy and ultrasound showed favourable results in improving symptom severity and functional status compared to manual therapy. In addition, the low level laser showed improvements in pinch strength compared to placebo and pain (VAS) compared to manual therapy. Splints showed superior results to electrophysical modalities. The clinical significance of the results was assessed by effect size estimation and comparison with the minimum clinically important difference (MCID).

CONCLUSIONS: We found favourable results in pain relief, improvement of symptoms, functional status, and neurophysiological parameters for some electrophysical modalities, mainly when applied with a splint. Our results coincide with those obtained in some meta-analyses. However, none of these can be considered clinically significant.

TRIAL REGISTRATION: PROSPERO registration number CRD42020168792; https://www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=168792.

PMID:33735212 | DOI:10.1371/journal.pone.0248484

Front Artif Intell. 2021 Feb 23;3:621577. doi: 10.3389/frai.2020.621577. eCollection 2020.

ABSTRACT

In this paper, we discuss the use of natural language processing and artificial intelligence to analyze nutritional and sustainability aspects of recipes and food. We present the state-of-the-art and some use cases, followed by a discussion of challenges. Our perspective on addressing these is that while they typically have a technical nature, they nevertheless require an interdisciplinary approach combining natural language processing and artificial intelligence with expert domain knowledge to create practical tools and comprehensive analysis for the food domain.

PMID:33733227 | PMC:PMC7940824 | DOI:10.3389/frai.2020.621577

JMIR Form Res. 2021 Mar 16;5(3):e24738. doi: 10.2196/24738.

ABSTRACT

BACKGROUND: Traditionally, digital health data management has been based on electronic health record (EHR) systems and has been handled primarily by centralized health providers. New mechanisms are needed to give patients more control over their digital health data. Personal health libraries (PHLs) provide a single point of secure access to patients' digital health data and enable the integration of knowledge stored in their digital health profiles with other sources of global knowledge. PHLs can help empower caregivers and health care providers to make informed decisions about patients' health by understanding medical events in the context of their lives.

OBJECTIVE: This paper reports the implementation of a mobile health digital intervention that incorporates both digital health data stored in patients' PHLs and other sources of contextual knowledge to deliver tailored recommendations for improving self-care behaviors in diabetic adults.

METHODS: We conducted a thematic assessment of patient functional and nonfunctional requirements that are missing from current EHRs based on evidence from the literature. We used the results to identify the technologies needed to address those requirements. We describe the technological infrastructures used to construct, manage, and integrate the types of knowledge stored in the PHL. We leverage the Social Linked Data (Solid) platform to design a fully decentralized and privacy-aware platform that supports interoperability and care integration. We provided an initial prototype design of a PHL and drafted a use case scenario that involves four actors to demonstrate how the proposed prototype can be used to address user requirements, including the construction and management of the PHL and its utilization for developing a mobile app that queries the knowledge stored and integrated into the PHL in a private and fully decentralized manner to provide better recommendations.

RESULTS: To showcase the main features of the mobile health app and the PHL, we mapped those features onto a framework comprising the user requirements identified in a use case scenario that features a preventive intervention from the diabetes self-management domain. Ongoing development of the app requires a formative evaluation study and a clinical trial to assess the impact of the digital intervention on patient-users. We provide synopses of both study protocols.

CONCLUSIONS: The proposed PHL helps patients and their caregivers take a central role in making decisions regarding their health and equips their health care providers with informatics tools that support the collection and interpretation of the collected knowledge. By exposing the PHL functionality as an open service, we foster the development of third-party applications or services and provide motivational technological support in several projects crossing different domains of interest.

PMID:33724197 | DOI:10.2196/24738

J Korean Acad Nurs. 2021 Feb;51(1):68-79. doi: 10.4040/jkan.20213.

ABSTRACT

PURPOSE: This study aimed to analyze the mass and social media contents and structures related to particulate matter before and after the policy enforcement of the comprehensive countermeasures for particulate matter, derive nursing implications, and provide a basis for designing health policies.

METHODS: After crawling online news articles and posts on social networking sites before and after policy enforcement with particulate matter as keywords, we conducted topic and semantic network analysis using TEXTOM, R, and UCINET 6.

RESULTS: In topic analysis, behavior tips was the common main topic in both media before and after the policy enforcement. After the policy enforcement, influence on health disappeared from the main topics due to increased reports about reduction measures and government in mass media, whereas influence on health appeared as the main topic in social media. However semantic network analysis confirmed that social media had much number of nodes and links and lower centrality than mass media, leaving substantial information that was not organically connected and unstructured.

CONCLUSION: Understanding of particulate matter policy and implications influence health, as well as gaps in the needs and use of health information, should be integrated with leadership and supports in the nurses' care of vulnerable patients and public health promotion.

PMID:33706332 | DOI:10.4040/jkan.20213

Comput Struct Biotechnol J. 2021 Jan 29;19:1081-1091. doi: 10.1016/j.csbj.2021.01.037. eCollection 2021.

ABSTRACT

Diet is one of the main sources of exposure to toxic chemicals with carcinogenic potential, some of which are generated during food processing, depending on the type of food (primarily meat, fish, bread and potatoes), cooking methods and temperature. Although demonstrated in animal models at high doses, an unequivocal link between dietary exposure to these compounds with disease has not been proven in humans. A major difficulty in assessing the actual intake of these toxic compounds is the lack of standardised and harmonised protocols for collecting and analysing dietary information. The intestinal microbiota (IM) has a great influence on health and is altered in some diseases such as colorectal cancer (CRC). Diet influences the composition and activity of the IM, and the net exposure to genotoxicity of potential dietary carcinogens in the gut depends on the interaction among these compounds, IM and diet. This review analyses critically the difficulties and challenges in the study of interactions among these three actors on the onset of CRC. Machine Learning (ML) of data obtained in subclinical and precancerous stages would help to establish risk thresholds for the intake of toxic compounds generated during food processing as related to diet and IM profiles, whereas Semantic Web could improve data accessibility and usability from different studies, as well as helping to elucidate novel interactions among those chemicals, IM and diet.

PMID:33680352 | PMC:PMC7892627 | DOI:10.1016/j.csbj.2021.01.037

Glycobiology. 2021 Feb 23:cwab013. doi: 10.1093/glycob/cwab013. Online ahead of print.

ABSTRACT

Recent years have seen great advances in the development of glycoproteomics protocols and methods resulting in a sustainable increase in the reporting proteins, their attached glycans and glycosylation sites. However, only very few of these reports find their way into databases or data repositories. One of the major reasons are the absence of digital standard to represent glycoproteins and the challenging annotations with glycans. Depending on the experimental method, such a standard must be able to represent glycans as complete structures or as compositions, store not just single glycans but also represent glycoforms on a specific glycosylation side, deal with partially missing site information if no site mapping was performed, and store abundances or ratios of glycans within a glycoform of a specific site. In order to support the above, we have developed the GlycoConjugate Ontology (GlycoCoO) as a standard semantic framework to describe and represent glycoproteomics data. GlycoCoO can be used to represent glycoproteomics data in triplestores and can serve as a basis for data exchange formats. The ontology, database providers and supporting documentation are available online (https://github.com/glycoinfo/GlycoCoO).

PMID:33677548 | DOI:10.1093/glycob/cwab013

J Proteome Res. 2021 Mar 4. doi: 10.1021/acs.jproteome.0c00926. Online ahead of print.

ABSTRACT

The study of microbiomes has gained in importance over the past few years and has led to the emergence of the fields of metagenomics, metatranscriptomics, and metaproteomics. While initially focused on the study of biodiversity within these communities, the emphasis has increasingly shifted to the study of (changes in) the complete set of functions available in these communities. A key tool to study this functional complement of a microbiome is Gene Ontology (GO) term analysis. However, comparing large sets of GO terms is not an easy task due to the deeply branched nature of GO, which limits the utility of exact term matching. To solve this problem, we here present MegaGO, a user-friendly tool that relies on semantic similarity between GO terms to compute the functional similarity between multiple data sets. MegaGO is high performing: Each set can contain thousands of GO terms, and results are calculated in a matter of seconds. MegaGO is available as a web application at https://megago.ugent.be and is installable via pip as a standalone command line tool and reusable software library. All code is open source under the MIT license and is available at https://github.com/MEGA-GO/.

PMID:33661648 | DOI:10.1021/acs.jproteome.0c00926

Int J Environ Res Public Health. 2021 Feb 26;18(5):2309. doi: 10.3390/ijerph18052309.

ABSTRACT

Infants need sufficient nutrients even during disasters. Only qualitative descriptive analysis has been reported regarding nutritional problems of mothers and children after the Kumamoto earthquake, and non-subjective analysis is required. This study examined issues concerning maternal and child health, food and nutrition after the Kumamoto earthquake using automatic computer quantitative analysis from focus group interviews (FGIs). Study participants (n = 13) consisted of dietitians in charge of nutrition assistance of infants in affected areas. The content of the interviews was converted into text, nouns were extracted, and co-occurrence network diagram analysis was performed. In the severely damaged area, there were hygienic problems not only in the acute phase but also in the mid-to-long-term phase. "Allergy" was extracted in the surrounding area in the acute and the mid-to-long-term phase, but not in the severely damaged area as the acute phase issue. In the surrounding area, problems have shifted to health and the quality of diet in the mid-to-long-term phase. This objective analysis suggested that dietary problems for mothers and children after disaster occurred also in the mid-to-long-term phase. It will be necessary to combine the overall trends obtained in this study with the results of qualitative descriptive analysis.

PMID:33652781 | PMC:PMC7956302 | DOI:10.3390/ijerph18052309

Cognition. 2021 Jun;211:104631. doi: 10.1016/j.cognition.2021.104631. Epub 2021 Feb 24.

ABSTRACT

Older adults tend to have a broader vocabulary compared to younger adults - indicating a richer storage of semantic knowledge - but their retrieval abilities decline with age. Recent advances in quantitative methods based on network science have investigated the effect of aging on semantic memory structure. However, it is yet to be determined how this aging effect on semantic memory structure relates to its overall flexibility. Percolation analysis provides a quantitative measure of the flexibility of a semantic network, by examining how a semantic memory network is resistant to "attacks" or breaking apart. In this study, we incorporated percolation analyses to examine how semantic networks of younger and older adults break apart to investigate potential age-related differences in language production. We applied the percolation analysis to 3 independent sets of data (total N = 78 younger, 78 older adults) from which we generated semantic networks based on verbal fluency performance. Across all 3 datasets, the percolation integrals of the younger adults were larger than older adults, indicating that older adults' semantic networks were less flexible and broke down faster than the younger adults'. Our findings provide quantitative evidence for diminished flexibility in older adults' semantic networks, despite the stability of semantic knowledge across the lifespan. This may be one contributing factor to age-related differences in language production.

PMID:33639378 | PMC:PMC8058279 | DOI:10.1016/j.cognition.2021.104631

BMJ. 2021 Feb 25;372:n560. doi: 10.1136/bmj.n560.

NO ABSTRACT

PMID:33632849 | DOI:10.1136/bmj.n560

PLoS One. 2021 Feb 26;16(2):e0245847. doi: 10.1371/journal.pone.0245847. eCollection 2021.

ABSTRACT

Internet of Things (IoT) is the growing invention in the current development of different domains like industries, e-health, and education, etc. Semantic web of things (SWoT) is an extension of IoT that enhance the communication by behaving intelligently. SWoT comprises 7 layered architecture. The perception layer is an important layer for collecting data from devices and to communicate with its associated layer. The data loss at the perception layer is very common due to inadequate resources, unpredictable link, noise, collision, and unexpected damage. To address this problem, we propose a method based on Compressive Sensing which recovers and estimates sensory data from a low-rank structure. The contribution of this paper is three folds. Firstly, we determine the problem of data acquisition and data loss at semantic sensory nodes in SWoT. Secondly, we introduce a compressive sensing based framework for SWoT that recovers the data accurately using low-rank features. Thirdly, the data estimation method is utilized to reduce the volume of the data. Proposed Compressive Sensing based Data Recoverability and Estimation (CS-RE) method is evaluated and compared with the existing reconstruction methods. The simulation results on real sensory datasets depict that the proposed method significantly outperforms existing methods in terms of error ratio and data recoverability accuracy.

PMID:33635878 | DOI:10.1371/journal.pone.0245847

API Driven On-Demand Participant ID Pseudonymization in Heterogeneous Multi-Study Research.

Healthc Inform Res. 2021 Jan;27(1):39-47

Authors: Syed S, Syed M, Syeda HB, Garza M, Bennett W, Bona J, Begum S, Baghal A, Zozus M, Prior F

Abstract
OBJECTIVES: To facilitate clinical and translational research, imaging and non-imaging clinical data from multiple disparate systems must be aggregated for analysis. Study participant records from various sources are linked together and to patient records when possible to address research questions while ensuring patient privacy. This paper presents a novel tool that pseudonymizes participant identifiers (PIDs) using a researcher-driven automated process that takes advantage of application-programming interface (API) and the Perl Open-Source Digital Imaging and Communications in Medicine Archive (POSDA) to further de-identify PIDs. The tool, on-demand cohort and API participant identifier pseudonymization (O-CAPP), employs a pseudonymization method based on the type of incoming research data.
METHODS: For images, pseudonymization of PIDs is done using API calls that receive PIDs present in Digital Imaging and Communications in Medicine (DICOM) headers and returns the pseudonymized identifiers. For non-imaging clinical research data, PIDs provided by study principal investigators (PIs) are pseudonymized using a nightly automated process. The pseudonymized PIDs (P-PIDs) along with other protected health information is further de-identified using POSDA.
RESULTS: A sample of 250 PIDs pseudonymized by O-CAPP were selected and successfully validated. Of those, 125 PIDs that were pseudonymized by the nightly automated process were validated by multiple clinical trial investigators (CTIs). For the other 125, CTIs validated radiologic image pseudonymization by API request based on the provided PID and P-PID mappings.
CONCLUSIONS: We developed a novel approach of an ondemand pseudonymization process that will aide researchers in obtaining a comprehensive and holistic view of study participant data without compromising patient privacy.

PMID: 33611875 [PubMed]

Disparities in the hospital cost of cardiometabolic diseases among lesbian, gay, and bisexual Canadians: a population-based cohort study using linked data.

Can J Public Health. 2020 06;111(3):417-425

Authors: Gupta N, Sheng Z

Abstract
OBJECTIVES: Sexual identity has been recognized as a social determinant of health; however, evidence is limited on sexual minority status as a possible contributor to inequalities in cardiometabolic outcomes and the related hospital burden. This study aimed to investigate the association between sexual identity and hospital costs for cardiometabolic diseases among a cohort of Canadians using linked survey and administrative data.
METHODS: Data from the 2007-2011 Canadian Community Health Survey were linked to acute-care inpatient records from the 2005/2006-2012/2013 Discharge Abstract Database. Multiple linear regression was used to assess the association between self-reported sexual identity and inpatient resource use for cardiometabolic diseases.
RESULTS: Among the population ages 18-59, 2.1% (95% CI 1.9-2.2) identified as lesbian, gay, or bisexual (LGB). LGB individuals more often reported having diabetes or heart disease compared with heterosexuals. The mean inflation-adjusted cost for cardiometabolic-related hospitalizations was found to be significantly higher among LGB patients (CAD$26,702; 95% CI 26,166-60,365) than among their heterosexual counterparts ($10,137; 95% CI 8,639-11,635), in part a reflection of longer hospital stays (13.6 days versus 5.1 days). Inpatient costs remained 54% (95% CI 8-119) higher among LGB patients after controlling for socio-demographics, health status, and health behaviours.
CONCLUSION: This study revealed a disproportionate cost for potentially avoidable hospitalizations for cardiometabolic conditions among LGB patients, suggesting important unmet healthcare needs even in the Canadian context of universal coverage.

PMID: 32112310 [PubMed - indexed for MEDLINE]

Brain Behav. 2021 Apr;11(4):e02065. doi: 10.1002/brb3.2065. Epub 2021 Feb 18.

ABSTRACT

INTRODUCTION: The semantic network is an important mediator of language, enabling both speech production and the comprehension of multimodal stimuli. A major challenge in the field of neurosurgery is preventing semantic deficits. Multiple cortical areas have been linked to semantic processing, though knowledge of network connectivity has lacked anatomic specificity. Using attentional task-based fMRI studies, we built a neuroanatomical model of this network.

METHODS: One hundred and fifty-five task-based fMRI studies related to categorization of visual words and objects, and auditory words and stories were used to generate an activation likelihood estimation (ALE). Cortical parcellations overlapping the ALE were used to construct a preliminary model of the semantic network based on the cortical parcellation scheme previously published under the Human Connectome Project. Deterministic fiber tractography was performed on 25 randomly chosen subjects from the Human Connectome Project, to determine the connectivity of the cortical parcellations comprising the network.

RESULTS: The ALE analysis demonstrated fourteen left hemisphere cortical regions to be a part of the semantic network: 44, 45, 55b, IFJa, 8C, p32pr, SFL, SCEF, 8BM, STSdp, STSvp, TE1p, PHT, and PBelt. These regions showed consistent interconnections between parcellations. Notably, the anterior temporal pole, a region often implicated in semantic function, was absent from our model.

CONCLUSIONS: We describe a preliminary cortical model for the underlying structural connectivity of the semantic network. Future studies will further characterize the neurotractographic details of the semantic network in the context of medical application.

PMID:33599397 | PMC:PMC8035438 | DOI:10.1002/brb3.2065

A Novel Computerized Cognitive Stress Test to Detect Mild Cognitive Impairment.

J Prev Alzheimers Dis. 2021;8(2):135-141

Authors: Curiel Cid RE, Crocco EA, Kitaigorodsky M, Beaufils L, Peña PA, Grau G, Visser U, Loewenstein DA

Abstract
BACKGROUND: The Loewenstein Acevedo Scales of Semantic Interference and Learning (LASSI-L) is a novel and increasingly employed instrument that has outperformed widely used cognitive measures as an early correlate of elevated brain amyloid and neurodegeneration in prodromal Alzheimer's Disease (AD). The LASSI-L has distinguished those with amnestic mild cognitive impairment (aMCI) and high amyloid load from aMCI attributable to other non-AD conditions. The authors designed and implemented a web-based brief computerized version of the instrument, the LASSI-BC, to improve standardized administration, facilitate scoring accuracy, real-time data entry, and increase the accessibility of the measure.
OBJECTIVE: The psychometric properties and clinical utility of the brief computerized version of the LASSI-L was evaluated, together with its ability to differentiate older adults who are cognitively normal (CN) from those with amnestic Mild Cognitive Impairment (aMCI).
METHODS: After undergoing a comprehensive uniform clinical and neuropsychological evaluation using traditional measures, older adults were classified as cognitively normal or diagnosed with aMCI. All participants were administered the LASSI-BC, a computerized version of the LASSI-L. Test-retest and discriminant validity was assessed for each LASSI-BC subscale.
RESULTS: LASSI-BC subscales demonstrated high test-retest reliability, and discriminant validity was attained.
CONCLUSIONS: The LASSI-BC, a brief computerized version of the LASSI-L is a valid and useful cognitive tool for the detection of aMCI among older adults.

PMID: 33569559 [PubMed - in process]

Characterizing Tractability of Simple Well-Designed Pattern Trees with Projection.

Theory Comput Syst. 2021;65(1):3-41

Authors: Mengel S, Skritek S

Abstract
We study the complexity of evaluating well-designed pattern trees, a query language extending conjunctive queries with the possibility to define parts of the query to be optional. This possibility of optional parts is important for obtaining meaningful results over incomplete data sources as it is common in semantic web settings. Recently, a structural characterization of the classes of well-designed pattern trees that can be evaluated in polynomial time was shown. However, projection-a central feature of many query languages-was not considered in this study. We work towards closing this gap by giving a characterization of all tractable classes of simple well-designed pattern trees with projection (under some common complexity theoretic assumptions). Since well-designed pattern trees correspond to the fragment of well-designed {AND, OPTIONAL}-SPARQL queries this gives a complete description of the tractable classes of queries with projections in this fragment that can be characterized by the underlying graph structures of the queries. For non-simple pattern trees the tractability criteria for simple pattern trees do not capture all tractable classes. We thus extend the characterization for the non-simple case in order to capture some additional tractable cases.

PMID: 33568963 [PubMed]

Neural Netw. 2021 May;137:63-74. doi: 10.1016/j.neunet.2020.12.009. Epub 2021 Jan 22.

ABSTRACT

As deep neural net architectures minimize loss, they accumulate information in a hierarchy of learned representations that ultimately serve the network's final goal. Different architectures tackle this problem in slightly different ways, but all create intermediate representational spaces built to inform their final prediction. Here we show that very different neural networks trained on two very different tasks build knowledge representations that display similar underlying patterns. Namely, we show that the representational spaces of several distributional semantic models bear a remarkable resemblance to several Convolutional Neural Network (CNN) architectures (trained for image classification). We use this information to explore the network behavior of CNNs (1) in pretrained models, (2) during training, and (3) during adversarial attacks. We use these findings to motivate several applications aimed at improving future research on CNNs. Our work illustrates the power of using one model to explore another, gives new insights into the function of CNN models, and provides a framework for others to perform similar analyses when developing new architectures. We show that one neural network model can provide a window into understanding another.

PMID:33556802 | DOI:10.1016/j.neunet.2020.12.009