** Abstract: **NAIE has built an online operation and maintenance expert system based on knowledge mapping.
The trend of AI has brought us many new terms: deep learning, neural network, knowledge graph….
Are they all familiar to you? Do you all know what applications they have?
Huawei Network Artificial Intelligence Engine (NAIE) is dedicated to self-optimizing, self-healing, and self-maintenance of networks. One of the key components is the ability to automatically analyze and locate the root cause of a network failure, and automatically take appropriate recovery measures and maintenance strategies. NAIE has built an online operation and maintenance expert system with knowledge and reasoning using knowledge maps.
This system in addition to the data pattern in the automation interface docking machine, report processing network system to collect the failure data of automatic fault identification and returning for positioning, also can through the human-computer interaction interface intelligent network operations staff query response and feedback precise fault auxiliary information, like the model of fault equipment, processing steps, such as the details of the fault information, The feedback results are presented in a clear visual atlas, so that network operation and maintenance personnel can fully control the fault and trust the system to deal with it automatically.
“Expert system for online operations” covers the use of knowledge map with proper knowledge representation structure describing failure events, objects and relationships, through the map generation techniques to realize automatic docking failure data and the said structure, through static knowledge automatic acquisition, integration, support semantic understanding and human-computer interaction and a series of key technologies.
In order to help you better understand the application of knowledge Graph in “Root cause of failure”, we specially invited Liu Ruihong, huawei product management expert and director of Knowledge Graph TIG, to answer your questions.
Q: What do you think are the high value application scenarios for the Knowledge Graph?
A: You can comprehensively select high-value scenarios from various dimensions such as high network complexity, long time consuming, high labor cost, low processing efficiency, and large service impact range. In the network of automatic driving “rules, building, d, superior, camp, province” and so on each link, the link of network maintenance is a knowledge map application scenarios of high value, because the network maintenance needs both in-depth knowledge of basic knowledge of networking and fault maintenance experience, also need to failure occurs system to produce a reflection of the failure phenomenon, the information such as location, status of various machine data, Only the combination of fault knowledge and machine data can assist in locating faults and reasoning the root cause.
Q: What are the technical forms of knowledge reasoning?
A: In the process of failure of returning for automatic reasoning, reasoning based on expert rules is the most commonly used is also one of the most effective way at present, the difficulty lies in the representation of inference rules to be able to reflect the failure mechanism of conduction relationship and network, and the rules of the object should be based on the network object classes and class event ontology, should not be aimed at specific machine data instances, To ensure that every rule is atomic rule, every rule can be reused.
In addition, reasoning based on description logic, representation learning reasoning based on distributed representation, ontology reasoning and composite reasoning are all important technical means.
Q: What do you think should be stored in a knowledge base?
A: Network structure of fault knowledge, namely schema is the key to the knowledge base to store the core, in addition, docking machine data adaptation rules, the bearing expert experience of inference rules, describe the failure of static knowledge and semantic knowledge, reflect fault field of machine data, such as topology, alarm, performance indicators, all kinds of log, etc., Are the key content that the knowledge base needs to store.
Q: What technologies do you think are key to the automated construction of knowledge graphs?
A: Although there are some bottom-up automatic modeling techniques and methods in schema construction, it is difficult to apply in the professional field. Knowledge modeling is the process of building A cognitive network system, which is the key foundation and prerequisite for instilling knowledge into the system. Knowledge modeling is like teaching a child to know Chinese characters, which requires basic cognitive knowledge one word at a time. When the cognition system of the basic structure of knowledge is established, automatic technology can be adopted in automatic extraction, knowledge fusion, knowledge query, knowledge base inspection and other links.
If you are interested in this topic, welcome to “Huawei 2020 Fully Connected” to communicate with experts face to face.
Click to follow, the first time to learn about Huawei cloud fresh technology ~