Learning Pattern Relation-Based Hyperbolic Embedding for Adverse Drug Reaction Extraction

2021 ◽  
Vol 12 (2) ◽  
pp. 69-87
Author(s):  
Siriwon Taewijit ◽  
Thanaruk Theeramunkong
Keyword(s):  
Hierarchical Structure ◽  
Medical Information ◽  
Hospital Admissions ◽  
Syntactic Structure ◽  
Semantic Representation ◽  
Cartesian Product ◽  
Natural Form ◽  
Relation Type ◽  
Vector Sum ◽  
Hyperbolic Embedding

Hyperbolic embedding has been recently developed to allow us to embed words in a Cartesian product of hyperbolic spaces, and its efficiency has been proved in several works of literature since the hierarchical structure is the natural form of texts. Such a hierarchical structure exhibits not only the syntactic structure but also semantic representation. This paper presents an approach to learn meaningful patterns by hyperbolic embedding and then extract adverse drug reactions from electronic medical records. In the experiments, the public source of data from MIMIC-III (Medical Information Mart for Intensive Care III) with over 58,000 observed hospital admissions of the brief hospital course section is used, and the result shows that the approach can construct a set of efficient word embeddings and also retrieve texts of the same relation type with the input. With the Poincaré embeddings model and its vector sum (PC-S), the authors obtain up to 82.3% in the precision at ten, 85.7% in the mean average precision, and 93.6% in the normalized discounted cumulative gain.

scholarly journals The Ka- Passive Form in Balinese

2019 ◽  
Vol 10 (4) ◽  
pp. 886
Author(s):  
Nyoman Sujaya ◽  
Ketut Artawa ◽  
I Nyoman Kardana ◽  
Made Sri Satyawati
Keyword(s):  
Syntactic Structure ◽  
Semantic Representation ◽  
Narrative Texts ◽  
Passive Voice ◽  
Base Form ◽  
The Voice

This paper accounts for the ka- passive form in Balinese. It focuses on its syntactic and semantic representation. Using the data taken from Balinese narrative texts issued in the Bali Orti of Bali Post newspaper, and applying the RRG theory by Van Valin and Randy (1999), it was found out that the ka- passive belongs to a morphological passive voice of Balinese where the the voice is marked on the verb (it is marked by prefix ka-). The ka- base form can be attached by applicative suffixes such as -ang, -in, and –an. These morphological verbs imply various syntactic structure and semantic representation.

The indivisibility of words

1979 ◽  
Vol 15 (1) ◽  
pp. 39-47 ◽  
Author(s):  
Geoffrey Sampson
Keyword(s):  
Natural Language ◽  
Syntactic Structure ◽  
Semantic Representation ◽  
Tree Structure ◽  
Breaking Point ◽  
Derivational Morphology ◽  
Adequate Description ◽  
Linguistic Representation ◽  
Level I

Many contemporary linguists hold that an adequate description of a natural language must represent many of its vocabulary items as syntactically and/or semantically complex. A sentence containing the word kill, for instance, will on this view be assigned a ‘deep syntactic structure’ or ‘semantic representation’ in which kill is represented by a portion or portions of tree-structure, the lowest nodes of which are labelled with ‘semantic primitives’ such as CAUSE and DIE, or CAUSE, BECOME, NOT and ALIVE. In the case of words such as cats or walked, which are formed in accordance with productive rules of ‘inflexional’ rather than ‘derivational’ morphology, there is little dispute that their composite status will be reflected at most or all levels of linguistic representation. (That is why I refer, above, to ‘vocabulary items’: cat and cats may be called different ‘words’, but not different elements of the English vocbulary.) When morphologically simple words such as kill are treated as composite at a ‘deeper’ level, I, for one, find my credulity strained to breaking point. (The case of words formed in accordance with productive or non-productive rules of derivational morphology, such as killer or kingly, is an intermediate one and I shall briefly return to it below.)

scholarly journals Linguistic processor integration for solving planimetric problems

2021 ◽  
Vol 15 (4) ◽  
pp. 0-0
Keyword(s):  
Natural Language ◽  
Syntactic Structure ◽  
Semantic Representation ◽  
Interactive Visualization ◽  
Subject Area ◽  
Syntactic Structures ◽  
The Subject ◽  
Language Description

The research deals with the original algorithms of the linguistic processor integration for solving planimetric problems. The linguistic processor translates the natural language description of the problem into a semantic representation based on the ontology that supports the axiomatics of geometry. The linguistic processor synthesizes natural-language comments to the solution and drawing objects. The method of interactive visualization of the linguistic processor functioning is proposed. The method provides a step-by-step dialog control of syntactic structure construction and its display in semantic representation. During the experiments, several dozens of standard syntactic structures correctly displayed in the semantic structures of the subject area were obtained. The direction of further research related to the development of the proposed approach is outlined.

scholarly journals An Automatic Approach to Extending the Consumer Health Vocabulary

2020 ◽  
Vol 0 (0) ◽  
Author(s):  
Michal Monselise ◽  
Jane Greenberg ◽  
Ou Stella Liang ◽  
Sonia Pascua ◽  
Heejun Kim ◽  
...  
Keyword(s):  
Research Design ◽  
Hierarchical Structure ◽  
Medical Information ◽  
Research Effort ◽  
The Internet ◽  
Consumer Health ◽  
Medical Conditions ◽  
Medical Subject Headings ◽  
Ontology Structure ◽  
Single Entry

AbstractPurposeGiven the ubiquitous presence of the internet in our lives, many individuals turn to the web for medical information. A challenge here is that many laypersons (as “consumers”) do not use professional terms found in the medical nomenclature when describing their conditions and searching the internet. The Consumer Health Vocabulary (CHV) ontology, initially developed in 2007, aimed to bridge this gap, although updates have been limited over the last decade. The purpose of this research is to implement a means of automatically creating a hierarchical consumer health vocabulary. This overall purpose is improving consumers’ ability to search for medical conditions and symptoms with an enhanced CHV and improving the search capabilities of our searching and indexing tool HIVE (Helping Interdisciplinary Vocabulary Engineering).Design/methodology/approachThe research design uses ontological fusion, an approach for automatically extracting and integrating the Medical Subject Headings (MeSH) ontology into CHV, and further convert CHV from a flat mapping to a hierarchical ontology. The additional relationships and parent terms from MeSH allow us to uncover relationships between existing terms in the CHV ontology as well. The research design also included improving the search capabilities of HIVE identifying alternate relationships and consolidating them to a single entry.FindingsThe key findings are an improved CHV with a hierarchical structure that enables consumers to search through the ontology and uncover more relationships.Research limitationsThere are some cases where the improved search results in HIVE return terms that are related but not completely synonymous. We present an example and discuss the implications of this result.Practical implicationsThis research makes available an updated and richer CHV ontology using the HIVE tool. Consumers may use this tool to search consumer terminology for medical conditions and symptoms. The HIVE tool will return results about the medical term linked with the consumer term as well as the hierarchy of other medical terms connected to the term.Originality/valueThis is a first attempt in over a decade to improve and enhance the CHV ontology with current terminology and the first research effort to convert CHV's original flat ontology structure to a hierarchical structure. This research also enhances the HIVE infrastructure and provides consumers with a simple, efficient mechanism for searching the CHV ontology and providing meaningful data to consumers.

Abstract 203: Adoptions of Smartphones, Tablets, and Mobile Applications Among Patients with Hypertension

Hypertension ◽  
2014 ◽  
Vol 64 (suppl_1) ◽  
Author(s):  
Victor Benvenuto ◽  
Giselle Baquero ◽  
David L Scher ◽  
Stacey LaPine ◽  
Jason Fragin ◽  
...  
Keyword(s):  
Medical Information ◽  
Hospital Admissions ◽  
Blood Pressure Monitoring ◽  
Limiting Factors ◽  
Ipod Touch ◽  
Health Related ◽  
A Cell ◽  
Therapy Compliance ◽  
Artery Disease ◽  
Develop Market

optimize management, monitoring, and therapy compliance in patients with hypertension (HTN). It could potentially prevent unnecessary emergency room visits and hospital admissions. Methods: We surveyed patients with HTN in cardiology and primary care clinics regarding their use of mobile technology. Results: 148 patients were included (79 female; age 16-64: 47%, >65: 53%, 70 male; 18-64:41%, >65: 59%). Associated diagnosis were coronary artery disease (21%), myocardial infarction (13%), arrhythmias (36%), heart failure (6%), or other forms of heart disease (3%). 83% own a cell phone, 29% are smartphones. 78% own a personal computer, laptop, iPod, or tablet. 35% report using APPS, of which 26% use health-related applications. 2% used APPS for blood pressure monitoring or management. 17% used APPS for diet or calorie counters, 13% for exercise, fitness or heart monitoring. 63% use APPS on a smartphone, 12% on iPod Touch, 62% on Tablet. Among all HTN patients surveyed, 64% report looking up medical information online on a computer, 53% more than once per month. When asked about willingness to pay, 44% were not willing to pay for a smartphone or tablet that would help monitor their condition. 61% were not willing to pay if an ER visit and its copayment could be prevented. 26% were not willing to use these devices if they were free or covered by insurance. Conclusions: A majority of HTN patients surveyed had access to technology with 22% using smartphones with APPS. However, there was limited use of health-specific APPS, especially those related to HTN. Limiting factors may include lack of outcomes-based research, structured programs that incorporate APPS and devices, or lack of familiarity with APPS and technological devices. A comprehensive strategy to develop, market, and demonstrate benefits of this rapidly growing technology is urgently needed.

scholarly journals Linguistic Processor Integration for Solving Planimetric Problems

2021 ◽  
Vol 15 (4) ◽  
pp. 1-14
Author(s):  
Sergeyi S. Kurbatov
Keyword(s):  
Natural Language ◽  
Syntactic Structure ◽  
Semantic Representation ◽  
Interactive Visualization ◽  
Subject Area ◽  
Syntactic Structures ◽  
The Subject ◽  
Language Description

The research deals with the original algorithms of the linguistic processor integration for solving planimetric problems. The linguistic processor translates the natural language description of the problem into a semantic representation based on the ontology that supports the axiomatics of geometry. The linguistic processor synthesizes natural-language comments to the solution and drawing objects. The method of interactive visualization of the linguistic processor functioning is proposed. The method provides a step-by-step dialog control of syntactic structure construction and its display in semantic representation. During the experiments, several dozens of standard syntactic structures correctly displayed in the semantic structures of the subject area were obtained. The direction of further research related to the development of the proposed approach is outlined.

scholarly journals Does Syntax Need to Grow on Trees? Sources of Hierarchical Inductive Bias in Sequence-to-Sequence Networks

2020 ◽  
Vol 8 ◽  
pp. 125-140
Author(s):  
R. Thomas McCoy ◽  
Robert Frank ◽  
Tal Linzen
Keyword(s):  
Neural Network ◽  
Hierarchical Structure ◽  
Syntactic Structure ◽  
Network Models ◽  
Training Data ◽  
Training Set ◽  
Neural Network Models ◽  
Structured Model ◽  
Inductive Bias ◽  
Question Formation

Learners that are exposed to the same training data might generalize differently due to differing inductive biases. In neural network models, inductive biases could in theory arise from any aspect of the model architecture. We investigate which architectural factors affect the generalization behavior of neural sequence-to-sequence models trained on two syntactic tasks, English question formation and English tense reinflection. For both tasks, the training set is consistent with a generalization based on hierarchical structure and a generalization based on linear order. All architectural factors that we investigated qualitatively affected how models generalized, including factors with no clear connection to hierarchical structure. For example, LSTMs and GRUs displayed qualitatively different inductive biases. However, the only factor that consistently contributed a hierarchical bias across tasks was the use of a tree-structured model rather than a model with sequential recurrence, suggesting that human-like syntactic generalization requires architectural syntactic structure.

The Notion of ‘Underlying Levels’ in the Arabic Grammatical Tradition

1994 ◽  
Vol 21 (3) ◽  
pp. 271-296 ◽  
Author(s):  
Kees Versteegh
Keyword(s):  
Early Stage ◽  
Syntactic Structure ◽  
Semantic Representation ◽  
Main Tool ◽  
Formal Approach ◽  
Syntactic Form ◽  
Islamic Tradition ◽  
Actual Speech ◽  
The Difference ◽  
Grammatical Tradition

Abstract Even ‘naive’ speakers use a distinction between actual, realized speech with its ‘literal’ meaning, and an underlying level of ‘what is actually meant’. Such a distinction is made because speakers instinctively feel that very often actual speech does not represent exactly what the speaker intends to say. In this paper it is claimed that this non-technical distinction lies at the basis of a technical distinction between a surface structure of speech and an underlying level. In the technical stage of Arabic grammar the emphasis shifts from an analysis of the underlying intention of the speaker towards an explanation of the syntactic form of actual speech, which is mapped onto an underlying representation. Both in the Classical Greek and the Arabic/Islamic tradition we find a development from an early stage of exegetical activity, in which the intention of the speaker or the text is elaborated by positing an underlying level of semantic representation, towards a technical distinction between a surface level and an underlying level. The difference between the two traditions lies in the fact that Greek linguistics was more semantically oriented, whereas in Arabic grammar the main tool of the grammarians, the taqdîr, was basically an instrument to explain the syntactic structure of speech, in line with the predominantly formal approach of the Arabic grammarians. Compared with modern linguistic theory, both traditions have in common that they do not look for an underlying level of meaning that is universal to all languages. The main reason for this difference is that neither Greek nor Arabic linguists were interested in the study of other languages.

scholarly journals Estimated number of seriously injured road users admitted to hospital in France between 2010 and 2017, based on medico-administrative data

2021 ◽  
Vol 21 (1) ◽  
Author(s):  
Anthony Zullo ◽  
Maxime Large ◽  
Emmanuelle Amoros ◽  
Jean-Louis Martin
Keyword(s):  
Cost Estimation ◽  
Injury Severity ◽  
Medical Information ◽  
Hospital Admissions ◽  
Estimation Method ◽  
Developed Countries ◽  
Road Accident ◽  
Data Bases ◽  
Serious Injury ◽  
Serious Injuries

Abstract Background In France, like in most developed countries, the number of road accident fatalities is estimated from police data. These estimates are considered to be good-quality, unlike estimates of road injuries admitted to hospital, and especially serious injuries. Methods The present study aimed to supply such data from French hospital medical information data-bases (PMSI). In the PMSI data-bases, road accident victims are identified by external causes of morbidity and mortality, which should be systematically recorded in case of injury, but are often missing. We therefore modeled presence/absence of external cause from the relevant subset of the medicine-surgery-obstetrics PMSI data-base using a logistic regression, and then weighting the results by inverse estimated probability. As ICD-10 coding does not include injury severity, we used the AAAM10 conversion instrument developed by the American Association for Automotive Medicine, originators of the Abbreviated Injury Scale, so as to conform to the European Commission’s definition of serious injury. Results The number of road-accident related hospital admissions is estimated to be about 100000 per year; serious injuries increased from about 18000 in 2010 to almost 20000 in 2017, with almost 17000 in 2012 and 2013, with a mean of one fatality per 5 serious injury admissions. Conclusions These serious injury estimates are close to those obtained by our team from other data and with a different estimation method. The present method has the advantage of using ICD codes for injured people admitted to hospital. This classification and data source (hospital discharge registry) are also used by most european countries reporting serious injury estimates to the Commission. It allows cost estimation of hospital care, and could be applied to other types of accidental injury.

Sign in / Sign up

Export Citation Format

Share Document