Several recent works explore the synergy between stream processing and reasoning to fully capture the requirements of modern data intensive applications.
5th International Symposium, RuleML 2011 – Europe, Barcelona, Spain, July 19-21, 2011. Proceedings
This gave birth to the research domain of stream reasoning. This tutorial offers a detailed presentation of the theoretical and technological achievements in stream reasoning, highlighting the key benefits and limitations of existing approaches, and discussing the open issues and opportunities for future research. The tutorial is conceived for an audience that is familiar with the models and systems for event and stream processing. It aims to further promote the integration of reasoning and event and stream processing in two ways: i it presents an active research domain, where researchers on event and stream processing can apply their expertise; ii it overviews reasoning techniques and technologies that can help advancing the state of the art in event and stream processing.
We do not assume prior knowledge of Semantic Web technologies and we dedicate the first part of the tutorial to introduce the terminology and concepts in the domain of Semantic Web and reasoning. We assume the audience to be familiar with the fundamental models and technologies for event and stream processing.
For PoC purposes, a new ontology has been designed for this. This is an OWL 2 DL ontology designed to represent different aspects of patient care in continuous care settings [ 55 ]. For this use case, the ACCIO ontology allows to represent hospital departments, rooms, sensors, BLE bracelets, observations, nurses, patients, actions, nurse calls, etc. These concepts can be represented, as well as the relations between them. Moreover, the ontology contains some concepts for this use case that allow the inference of certain situations and hence easier query writing.
The existing ACCIO ontology has been further extended for this work with some key concepts and relations specific for the current use case.
This extension is the CareRoomMonitoring. Here, all possible medical symptoms, diagnoses, symptoms and faults are defined. For this use case, the Concussion class is defined as being equivalent to a Diagnosis that has two medical symptoms, being light sensitiveness and sound sensitiveness:. Two conditions need to be fulfilled for an Observation individual to also be a SoundAboveThresholdFault individual. First, it needs to have a SoundAboveThresholdSymptom , indicating the possibly unpleasant situation. Second, the observation needs to be made by a sensor system situated at the same location as a sound sensitive patient, i.
If that is also the case, the possibly unpleasant situation is alarming. The approach of using medical symptoms, diagnoses, symptoms and faults allows complete separation of diagnosis registration and fault detection. For an observation to be a fault, the exact diagnosis of the patient located at the corresponding room is unimportant; only the medical symptom, e. Because the diagnosis is already defined in the system according to medical domain knowledge, its corresponding medical symptoms are implicitly known. By design, the ACCIO ontology contains different patterns to represent the logic related to this use case.
Each such Action has exactly one Status via hasStatus , indicating the status in the life cycle of the action. To this end, a general overview of the most important ontology patterns and classes is presented in Figure 4. Overview of the most important ontology patterns and related classes of the Proof-of-Concept use case. The example for light intensity is given; the classes and patterns for sound are similar. Both are use case specific. For this use case, the knowledge base consists of the continuous care ontology, described in Section 4.
On the BRS, the full knowledge base is available.
- Introduction to Reasoning in Event-Based Distributed Systems.
- From the Ground Up: Reasoning About Distributed Systems in the Real World.
- Rule-Based Reasoning, Programming, and Applications.
- Constraint Event-Driven Automated Reasoning – CEDAR.
- Gender Trouble: Feminism and the Subversion of Identity (2nd Edition);
Context data can be considered as static data: although it can change over time, the number of updates is low compared to the number of times a query evaluation is performed on the data before it changes. Changes to this context data are less frequent but do occur. For example, a newly diagnosed person is being accommodated in a room, or a new nurse starts working at a department.
In these cases, the knowledge base of each relevant component needs to be updated.
This is coordinated from the central knowledge base at the BRS. On each RSPS, only the patient information of its associated patient is available.www.viktorialovasklub.hu/wp-includes/zyvosyz/498-last-minute-dla.php
Constraint Event-Driven Automated Reasoning | ANR
Similarly, patient data of all patients in the room is present in the knowledge base of each LRS. Bob is lying in room of department A. According to the modeled diagnosis, he is suffering from a concussion. Moreover, each person is assigned a BLE bracelet. Moreover, a threshold value is modeled for the light and sound sensor. These thresholds are identical to the corresponding threshold values of Bob for exposure to light intensity and sound.
Directly linking these thresholds to the sensors themselves is crucial for the filtering at the RSP engine, as will be explained in Section 5. This is achieved by running an appropriate query, inserting the sensor thresholds into the different knowledge bases. For this use case, the streaming data is semantically annotated and pushed to different streams by an OBU. The semantic annotation is an important task.
Donate to arXiv
In this mapping process, the observations are modeled according to the continuous care ontology. The name of each observation is unique due to the observation identifier A - B. For a light intensity or BLE tag observation, the template is similar. Figure 5 gives an overview of the main components of the PoC. Overview of the components of the Proof-of-Concept.
The additional components E X and E Y represent the components taking local action. C Z represents the component taking care of the actual nurse call. Feedback loops are omitted from the overview. In every hospital room, one OBU per patient is present to monitor the environment. As explained in Section 4.
As a BLE beacon is a different type of sensor, it is not part of this board. Recall that the architecture is push-based, i. In the given use case, locally means within the hospital room. However, reasoning is time-consuming and may take too much time when fast reevaluation of the continuous queries is required. In the cascading reasoning approach, RSP engines are used to filter the change frequency of the data streams. Only interesting information is retained. Therefore, appropriate continuous queries are constructed that intelligently aggregate and filter the data streams for the use case at hand.
As explained before, a window needs to be placed on top of the continuous data streams. For these queries, a logical window is used, which is a window extracting all triples occurring during a certain time interval. In concrete, both queries are executed every 5 s, on a logical sliding window containing all light, sound and BLE tag observations of the previous 6 s.
The window size of 6 s is chosen as such to avoid situations where certain observations fall between two windows. Theoretically, this should not be possible when the window size and sliding step are both equal, but, in practice, a real implementation of the system may exhibit a lag of a few milliseconds between two query executions. The FilterLightIntensity query and its motivation are completely similar.
Lines 14—16 of the FilterSound query define its inputs: the stream with sensor board observations, the stream with BLE tag observations, and the context data available in the local knowledge base. Considering the first part, its first section lines 19—25 extracts any sound sensor observation in the input window.
About this book
The second section lines 27—41 contains an optional pattern that looks at BLE tag observations in the window corresponding to the BLE bracelet of a nurse. Note that this pattern only matches BLE tag observations of nurses that are working on the department the hospital room belongs to. This is because context information about the BLE bracelets of other nurses is not available in the local knowledge base. The rationale for this is that it is assumed that nurses of other departments know too little about the patients of the current department.
Hence, their presence in the room should not affect the outcome of the query. Line 43 of the query contains an important filter.