Citation Context

...based approaches. Rule-based approaches consist of a set of rules that are manually defined or automatically learned from training data (Bui & Sloot, 2011; Cohen et al,. 2009; Kaljurand et al., 2009; =-=Kilicoglu & Bergler, 2011-=-). To extract events from text, first event triggers are detected using a dictionary, then the defined rules are applied to the output of the NLP tools e.g., dependency parse trees, to find their argu...

Citation Context

...nd task 3 adds the detection of whether 1 http://uima.apache.org/ events are stated in a negated or speculative context. In event extraction, common approaches use Rule-based (Kaljurand et al., 2009; =-=Kilicoglu and Bergler, 2011-=-), Machine Learning (ML)-based (Björne at al., 2009; Miwa et al., 2010) and hybrid methods (Ahmed et al., 2009; Riedel, McClosky et al., 2011). Recently, (Riedel et al., 2011) present an approach base...

Citation Context

...; that triggers may be expressed in diverse ways (two event of type Negative Regulation are defined with different trigger words); and, finally, that the same trigger word (expression) may indicate different types of event, depending on the context. The various approaches proposed for trigger detection can be roughly categorized in three types: rulebased, dictionary-based, and machine learning-based. These approaches are summarized in Table 2 and presented in the remainder of this section. 3.4.1. Patterns and Matching Rules for Trigger Detection. There are several strategies based on patterns [70, 93] and matching rules. Rule-based methods commonly follow some manually defined linguistic patterns, which are then augmented with additional constraints based on word forms and Computational and Mathematical Methods in Medicine 7 syntactic categories to generate better matching precision. The main advantage of this kind of approach is that they usually require little computational effort. Rule-based event extraction systems consist of a set of rules that are manually defined or generated from training data. For instance, Casillas et al. [88] present a strategy based on Kybots (Knowledge Yieldin...

Citation Context

... used are the same as those employed by systems operating on continuous text spans, such as syntactic and token features, and dictionaries of negation and speculation clues. The approach described in =-=[44,77]-=- adapted existing modules [78] developed using a different corpus [52] that aim to find dependency relations between dictionaries of negation/speculation clues and event triggers. In [79], a model is ...

Citation Context

... and p65” Adjectival triggers English has a general morphological process of adjective conversion (Adjtr), which enables verbs to be used as adjectives. The pattern template for adjective triggers is =-=(13)-=- Template: [ NP ADJ NP1 ] 〈ADJ head〉 = Adjtr 〈example〉 = “expressed pseudogenes” In this framework, adjectival derivations can be the present participle (14a), the past participle (14b), and the adjec...

Citation Context

...utilized the similarity of the ID and GE datasets. The three machine learning systems [24,27,28] were trained for the ID task on a combination of ID and GE data, while the rule-based Concordia system =-=[29]-=- was developed to have mostly a single rule set for the GE, EPI and ID tasks. Following these approaches, we added the GE corpus into the training data of the ID task trigger and edge detectors, furth...

Citation Context

...d Table 4 Teams who participated in the GE and CO tasks Team ’09 Task Background reference FAUST √ 12- - 3C [18] UMASS √ 12- - 1C [19] UTurku √ 123 C 1BI [20] MSR-NLP √ 1– - 4C [21] ConcordU 1-3 C 2C =-=[22]-=- UWMadison √ 1– - 2C [23] Stanford √ 1– - 3C+1.5L [24] BMI@ASU 12- - 3C [25] CCP-BTMG √ 1– - 3BI [26] TM-SCS 1– - 1C [27] XABioNLP 1– - 4C [28] HCMUS 1– - 6L [29] UUtah — C 1C [30] UZurich — C 1C [31]...

Citation Context

...ipants made use of supporting syntactic analyses provided by the organizers [98], none applied the analyses for supporting tasks such as coreference or REL - at least in cases due to time constraints =-=[99]-=-. There may thus remain further opportunities for improvement through combinations of supporting analyses with main task extraction systems. We find a remarkable number of similarities between the app...