Process Fragment Recognition in Clinical Documents

Abstract

Abstract. We describe a first experiment on automated activity and relation identification, and more in general, on the automated identification and extraction of computer-interpretable guideline fragments from clinical documents. We rely on clinical entity and relation (activities, actors, artifacts and their relations) recognition techniques and use MetaMap and the UMLS Metathesaurus to provide lexical information. In particular, we study the impact of clinical document syntax and semantics on the precision of activity and temporal relation recognition.

Clinical information extraction applications: A literature review

“…Later, MetaMap’s ability was improved to process clinical text [63], which is reflected by the large number of studies using MetaMap for clinical IE tasks. In the included studies, MetaMap has been used for phenotype extraction [31, 64–69], assessment of emergency department use [27, 70], drug-disease treatment relationships [71], fragment recognition in clinical documents [72], and extraction of patient-related attributes [73]. MedLEE is one of the earliest clinical NLP systems developed and is mostly used for pharmacovigilance [26, 74, 75] and pharmacoepidemiology [76, 77].…”

Automatic process model discovery from textual methodologies

“…In addition to the later m other approaches to disco documents [13], emails collaboration [15] and busin 17]. Compared to our techn and fully-supervised, relyin corpus.…”

“…TED WORKS ext for discovering models has s. Extensive work done in this ents engineering where textual reated before the design and tion system. Years ago, Rolland of strategies to address the d conceptual modelling: support from text, support the model eral understanding of texts and ork is related to the first strategy process models from textual ork, Rolland [42] classifies the ort the discovery of conceptual ng the following model-related namic, rule-based or ontologyspects of the conceptual models, and generate object models or requirements specifications [43, he dynamic aspects, Rolland t use cases and scenarios from lly, regarding the rule-based or re are approaches looking for models from business rules [51, mentioned works, we have found over processes from: clinical [14], stories extracted from ness process documentation [16, ique, [13] is domain dependent ng on prior annotation of the ique similar to ours but the key heir case are the actors and the mplate for activity extraction that e passive-voice sentences and n the articles there is no fineobjects. [16] first defines a list for identifying process-related xtract activities.…”

“…The technique we created, TextProcessMiner, generates the log of activities from text and then discovers the process instance model. Compared to other related works [13,14,15,16,17], TextProcessMiner is fully unsupervised and uses natural language processing techniques with a focus on the verb semantics. The technique has emerged as a result of technical action research [19].…”

From Natural Language Text to Visual Models: A survey of Issues and Approaches

“…Medical field [78] reveals a new approach based on the identification of four concepts: activities, resources, actors, and control flows.…”