Blackwell reference online [electronic resource]. This link opens in a new window. With nearly reference volumes from the humanities and social sciences through to business and economics. Emerald insight [electronic resource]. Gale databases [electronic resource]. Gartner [electronic resource]. Web site provides access to the research and advisory firm's information technology news, research reports, and business and market analysis, and includes product descriptions and comparisons, information on trends, and a calendar of upcoming conferences and events for IT managers, specialists, and business interests.
JSTOR [electronic resource]. Provides fulltext online access to electronic archive of scholarly journal literature, with an emphasis on the retrospective conversion of selected journals in history, economics, political science, demography, mathematics and other fields of the humanities and social sciences.
The site also provides access to selected ebooks. Oxford reference online [electronic resource]. The database brings together language and subject dictionaries and reference works published by the Oxford University Press into a single cross-searchable resource. ProQuest [electronic resource]. The database includes periodicals and newspapers articles from over 5, publication, with many in full text.
It comprises of various databases. Sage knowledge [electronic resource]. Scopus [electronic resource].
- Just a New Beginning;
- Biblioteca Madre María Teresa Guevara.
- Big Data Analytics.
- Computers and Talk in the Primary Classroom (Language and Education Library , No 12)!
The database offers access to 14, peer-reviewed titles from more than 4, international publishers. SpringerLink [electronic resource]. Contains an alphabetical list of Springer electronic journals subscribed by the University of Hong Kong. Abstracts and full-text of the journal articles are available. Web of science [electronic resource] This link opens in a new window.
Covers over 12, journals from categories from and has direct links to full text and library holdings information. It also includes over , proceedings from the most significant conferences, symposia, seminars, colloquia, workshops and conventions worldwide.
Electronic Books Collection C - D
Wiley online library [electronic resource]. Such linking is important, because of the wide range of ways in which a given concept may be mentioned in text. For example, variant mentions of the disorder concept dyspnea may include spelling variations dyspnoea , completely different forms e. Such corpora can facilitate the development of normalisation methods e. However, successful normalisation makes it possible to develop search tools that can automatically locate all mentions of a concept of interest in large collections of text, regardless of how the concept is mentioned.
Binary relation annotations link together pairs of NEs whose textual contexts connect them in specific ways. S1 In vitro interaction of prostaglandin F2alpha and oxytocin in placental vessels. S2 Lupus - like syndrome caused by 5 - aminosalicylic acid. It has been shown that corpora annotated with binary relations can be used to train automated tools to recognise relations in text types with divergent characteristics [ 46 , 67 , 68 , 69 , 70 , 71 , 72 , 73 ], thus allowing the discovery of high-accuracy evidence from multiple information sources that can satisfy a particular information need, e.
Despite this, binary relations are limited in terms of the complexity of the information that they can encode. For example, a binary relation representing an adverse drug reaction can only encode the fact that a single drug adversely affects or causes the occurrence of a particular disorder. However, additional information in the text may provide important or even critical details regarding the safe usage of the drug. For instance:. S3 We hypothesize that a pharmacodynamic or pharmacokinetic drug interaction between venlafaxine and trimipramine involving the CYP2D6 isoenzyme may have played a role in inducing the seizures.
The occurrence of a reaction may be restricted to certain types of medical subjects, who may be characterised in various ways, e. S5 We describe a life threatening side effect of acute epoprostenol infusion pulmonary edema in a patient with pulmonary hypertension associated with limited scleroderma and discuss its management and potential etiology.
Depending on the type and volume of evidence available, a reaction may be stated with varying degrees of confidence see sentence S6. S6 Marked elevation of serum CK may be a possible complication of olanzapine therapy.
New / Trial Databases
Uniquely within the field of PV, the abstracts in the corpus include annotations that go beyond binary relations, to encode more complex information, such as the cases exemplified above, in a structured manner. Ultimately, it is hoped that such tools will lead to the provision of curator-oriented applications that provide sophisticated, efficient and flexible means to explore and pinpoint relevant information in different textual sources, and thus help to increase the coverage, consistency and completeness of information in PV resources.
To ensure the utility of our corpus, all levels of annotation are inspired by, and comparable in scope to, annotation efforts in other domains, for which successful automated recognition has been demonstrated. The five levels are as follows:. Three types of NEs that correspond to the important details about drug effects, i. We apply an automated normalisation method [ 59 ], to link drug and disorder NEs with concept IDs in terminological resources, i.
Four types of complex relation annotations, called events [ 75 ], which link together an arbitrary number of participants either NEs or other events , to encode detailed information about the behaviour and effects of drugs.
Events have previously been used to encode a range of information in texts with divergent characteristics [ 42 , 76 , 77 , 78 , 79 , 80 ], and can be recognised automatically by a number of configurable tools [ 81 , 82 , 83 , 84 , 85 , 86 , 87 , 88 , 89 , 90 , 91 , 92 ]. Events alleviate several issues with binary relations. Our DDI and Combination event types, which encode cases in which multiple drugs are administered together or interact with each other, can themselves act as participants of AE and PTE events.
This makes it possible to encode complex causes of drug effects, which involve multiple drugs as in sentence S3. Three types of interpretative attributes assigned to events, which encode whether the events are negated, speculated sentence S6 and their manner, i. This latter information may be relevant in assessing the balance between the risks and benefits of taking particular combinations of drugs. For example, the phrase life - threatening in sentence S5 denotes a severe adverse reaction. Two types of static binary relations between NEs.
These are used, as in other event-annotated corpora [ 76 , 99 , ], to encode more detailed information about event participants, and have been shown to improve event extraction results [ ]. An example of such a relation is shown in sentence S7; structured relationships such as the one shown between 60 - year old woman and diabetes mellitus would make it possible to explore, e. These relations can supplement our automatically-added associations between NEs and concept IDs in allowing links to be established between events that refer to a common concept in different ways. Co-reference annotations between anaphoric event participants and their co-referent NEs in nearby sentences.
Our annotation scheme requires that all participants of a given event occur within a single sentence. However, some participants may correspond to anaphoric expressions, such as it, the drug, this disease, etc. We allow event participants corresponding to anaphoric expressions to be annotated, but require such expressions to be linked to their co-referent NEs in other sentences, to allow events to be correctly interpreted [ 81 ].
Co-reference annotation efforts in other biomedical corpora [ , ] have been used to demonstrate the feasibility of developing novel approaches to co-reference resolution [ , , ].
The creation of PHAEDRA was driven by detailed annotation guidelines, developed in consultation with domain experts, and used to ensure consistent annotation. Such consistency is quantitatively demonstrated by generally high levels of inter-annotator agreement, reaching as high as Subsequently, we provide further details about each of the five different annotation levels.
Finally, we explain our approach to normalising NE annotations to concept IDs in appropriate terminological resources.
- Ciba Foundation Symposium - Water and Electrolyte Metabolism in Relation to Age and Sex (Colloquia on Ageing), Volume 4.
- Paranoid schizophrenia - Wikipedia.
- The Bonhoeffer Legacy: Post-holocaust Perspectives.
- The Cambridge History of Seventeenth-Century Philosophy (vol. 1).
- Post navigation?
- Marxism, Socialism, Indian Politics: A View from the Left.
- Children in Changing Families: A Study of Adoption and Illegitimacy.
PHAEDRA is comprised of documents that are gathered based on: 1 publicly available resources relevant to the PV domain and 2 our own search—conducted in October —for more recently published scientific abstracts. In searching for further documents, two strategies were taken. The PubMed eUtilities [ ] were then used to download the content of corresponding abstracts. Overall, abstracts were gathered.
In order to select the subset that would form our corpus, the abstracts were ranked according to the number of unique drug names that they contain. To this end, a drug NE recognition tool [ ] was employed to annotate drug names in the set of otherwise un-annotated abstracts retrieved based on information from Drugs. By selecting the documents containing the highest number of drug names, we finally formed a corpus comprising , and 52 abstracts from the ADE, PK and DDI corpora, respectively, and abstracts from our own search.
The design of each annotation level was guided by an examination of a wide range of relevant MEDLINE abstracts, to determine the types of information that are most frequently specified, combined with discussions with domain experts, which helped us to focus on the most important types of information to be encoded in our scheme. To our knowledge, PHAEDRA is the first freely available corpus focussing on PV that simultaneously includes annotations corresponding to all of our three chosen NE types, as a basis for linking the effects of drugs with information about the medical subjects in which they occur.
Our annotators were encouraged to consult the Medical Dictionary for Regulatory Activities MedDRA [ ] as a guide to helping to determine when a phrase should be considered as a disorder. Most Subject phrases have a similar scope to [ 42 ], although our annotations extend beyond human subjects to cover different organisms and their sub-structures. Annotated NE spans may either be continuous or discontinuous.
Discontinuous annotations consist of two or more disconnected spans of text, which denote a complete NE when concatenated. Examples are shown in Fig. Each event annotation is anchored to a word or phrase known as the trigger that denotes the occurrence of the event in the sentence. As can be seen in the examples in Fig.
Each event participant is assigned a semantic role label to characterise its contribution towards the event description, and is linked to the trigger. Following most other event annotated corpora, and according to the current capabilities of available automated event extraction tools e. Different event types use different sets of semantic roles according to their semantics.
The semantic roles that we have defined for each event type e. Examples of different event types. For each event type, certain participant types are obligatory i.
For each event type, it is permitted to annotate multiple participants with the same role label. The DDI event type is annotated when an interaction is explicitly mentioned, while the Combination type is used in cases where it is simply stated that two or more drugs are administered together.
The Library: Faculty of Medicine, Chiang Mai University
The Combination event type makes it possible to differentiate between cases where a number of different drugs taken in isolation are listed as common causes of a given adverse effect see second AE event example in Fig. For AE events, it may be useful to detect statements that a given medication has a potentially harmful effect, even if it is not stated which disorder s are affected. This additional information may be present elsewhere in the abstract, or else such underspecified information may help to flag drugs for which further research is needed.
Although PTE events can be considered to be the opposite of AE events, we impose a different set of restrictions. Since most pharmacological substances are mentioned in the context of treatments, a simple statement that a drug has been administered is not particularly informative, and hence is not annotated as an event.