本文主要是对在线问诊中产生的医疗文本进行命名实体识别的研究.使用在线医疗问答网站的数据,采用{B, I, O}标注体系构建数据集,抽取疾病、治疗、检查和症状四个医疗实体.以BiLSTM-CRF为基准模型,提出两种深度学习模型IndRNN-CRF和IDCNN-BiLSTM-CRF,并在自构建数据集上验证模型的有效性.将新提出的两种模型与基准模型通过实验对比得出:模型IDCNN-BiLSTM-CRF的F1值0.8116,超过了BiLSTM-CRF的F1值0.8009, IDCNN-BiLSTM-CRF整体性能好于BiLSTM-CRF模型;模型IndRNN-CRF的精确率0.8427,但该模型在召回率上低于基准模型BiLSTM-CRF.
The paper gives an overview of a recently started research project addressing the question:Can context be exploited to add value to IT-services?Beyond today's way-finding services based on GPS and maps,we believe there is a new class of context-aware pervasive services making good use of a broad range of context information(e.g.identity,time of day,temperature,history as well as relative location of friends and belongings).Last but not least we believe one can find ways of adding value to IT-services by using disembodied conversational agents based on voice,sounds and visual signals integrated into the environment, instead of being guided by explicit directions shown on a display.The project aims at giving both new fundamental insight about the theory behind IT-services,their architecture and user interfaces and at building a number of prototypes demonstrating innovative context-aware pervasive IT-services.
Currently a large number of web services as well as other kinds of services such as EJBs,COM,and even Java Classes are made available to the general public.Facilitating the SOA based system development by leveraging such kinds of services becomes a challenge.A framework for service repository,ontology based service discovery and service composition is put forward.The service repository can maintain the web services,EJBs,and Java Classes with the functions such as service registration,publishing,discovery, matching,versioning,and monitoring.The details of service description are analyzed.A domain ontology for Procurement,Selling,and Inventory is also given.Based on the domain ontology and the service repository, the semantic enhanced service composition algorithm is discussed.
