Network intelligent performance of knowledge maps

Now that you have known about the Salary of Network Security Engineer, you must be attracted to achieving it. If you wish to have it, you must have to do lots and lots of studies, unless you have a good and reliable CCNP 300–825 Dumps provider like that of the EveDumps.

Let Ai smarter, Google should use knowledge map to make AI to understand the world like people.

Let AI are more intelligent, we must use knowledge map to make AI to understand the network like network experts.

Knowledge spectrum leads artificial intelligence to evolve into a cognitive stage in a perceptual phase, and become one of the current hotspot technology, focusing on the ICT industry.

Why is people attach great importance to knowledge map technology?

Because knowledge maps can not only build a full range of links to all-round links, support search needs based on common sense knowledge and conceptual knowledge, spawning search technologies such as Google, Baidu, Amazon Go, Microsoft Bing. Intelligent upgrade, and make various industry applications to get new progress in knowledge map blessings, have born all kinds of knowledge map applications, such as intelligent question and answer, financial credit, medical research, public security technology, Internet + living service, etc.

Different scenarios have put forward a variety of appeals in different knowledge fields, giving birth to explosive growth of knowledge map engineering and NLP various technologies, and puts forward various kinds of knowledge extraction and data processing technology. The technical needs.

Huawei network artificial intelligence in the field of knowledge map

Huawei network artificial intelligence hopes to use knowledge map technology to solve the intelligent operation and maintenance problem under the typical scene in the network field / B>, also to build a map, application map proposes various appeals: From knowledge content, different from encyclopedia knowledge maps, network domain knowledge map more attention to knowledge depth and completeness in the network field, from human machine interaction technology Anglerance, differently than open chat interactions, the network field is more concerned about the goal-oriented question and answer experience for solving problems. For example, a telecom core network operation and maintenance expert can answer and solve some of the expertise, whether the machine can do, even more profound understanding and reasoning interpretation, and then let the machine can assist the operation efficiency, Reduce operation and maintenance costs and time-saving purposes; future evolution to the advanced stage of network automatic driving, can reduce the operation and maintenance pressure of network operations engineers and network experts and night rate, providing more precise and more user-friendly intelligent services , Good.

Constructing knowledge maps can be divided into knowledge acquisition, knowledge fusion, knowledge verification, knowledge calculation, knowledge application, etc. , Huawei network artificial intelligence based on NaIE platform needs The knowledge map engineering system establishes a domain knowledge standard specification, refining knowledge processing technology chain, and improve operational operation and trusted ability. The overall idea is as shown in the following figure:

open to see this piece, the piece, the square side has:

knowledge Source:

from the source form Look, the knowledge is structured (for example: alarm, indicator, etc.), semi-structured (for example: configuration, log, standardized product document), non-structured (eg practical manual, fault case, sharing post) data, even In the brain of an expert. These network knowledge comes from the product documentation of the Support website, the maintenance document of the operation and maintenance expert, the current network capture package data in the event of alarm fault, the extensive network environment configuration document data, operation and maintenance expert’s experience of the sedimentation document or fault propagation knowledge collection, etc. Accordingly, we need to match the tools to facilitate access to these data, you can multiplex the data acquisition tools for the existing Naie platform, but also need to supplement such as captain data acquisition tools, interface, document data acquisition link channels and management tools, etc. Different sources, data sources of different structures are obtained.

Knowledge modeling

has a corpus, the first important issue we face is what kind of knowledge we need? Or do we need to extract what valuable knowledge in the data can solve the problem of faulty operations we face? This requires a valid knowledge organization structure. We need to design the knowledge model to create a data mode of the knowledge map before data acquisition. There are two ways to design: one is from the top-down method, network expert and modeling expert to use knowledge map modeling tool to edit Schema; the other is from the end Up to , based on the structure of source data, the corpus specification standard, the design of the graph technology, including: entity (point) modeling, attribute modeling, relationship (side) modeling, data The intellectual organizational form is established in the way in the figure, and mapping from the existing high quality data source. The importance of data modeling is that this work is the basis of all work of knowledge map project, so the standard specification SCHEMA design can effectively reduce the overall cost of the domain knowledge extraction and docking.

For example, we do fault communication knowledge maps, you need to define where the fault occurs (product object), what happened (alarm, abnormal indicator, fault phenomenon, log exception), what happened There is any transmission or dependence between the fault (the relationship between the alarm, the relationship between the alarm and abnormal abnormality, the relationship between the abnormal abnormalities, the relationship between faults, etc.). It is important to note that the classification criteria defines important knowledge that need to be embodied in the design. In addition, only this business knowledge is still not enough, to support a good human-machine interaction, it is also required to make sense knowledge in the network field. For example, when Ne, when the abbreviation occurs, you must know that “NE” is “NE”, it is not “Northeast”; when “POD does not come”, it is said that a process failed, not called Pod guy. Sleep.

Knowledge Storage:

With knowledge model, the organization and placement of knowledge is the shelf, how knowledge is needed according to the shelf, it is necessary to use storage It is necessary to store the key to knowledge storage technology. It is important to choose what kind of database is stored in Schema. Do you want to select Relational Database or NOSQL Database ? What kind of diagram database ? These need to be chosen according to the data scene. Wikidata selected Virtuso, CN-DBPEDIA is actually based on a Mongo database. Generally, the knowledge maps based on specific areas may be used to use a map database. Select RDF Store or Property Graph, need to consider knowledge source, use methods, Application characteristics. Network failure knowledge requires not only the picture query, the graph calculation, but also needs to understand the replica, carrying the answer to the fault problem, and the most ideal figure database is to parallelize, support the relationship storage, support graph computing, but also effectively store RDF form. Knowledge, support semantic understanding, three-way group, symbolicization knowledge, currently limited to physical list, we can only choose in the compliance open source diagram database and self-research map database, this is also We have spawned some of the key appeals to the self-research map database — multi-capacity fusion, and when the version of this demand is met, it is believed that it is a look to the developer.

Knowledge Extract:

We know that the knowledge distributed online often exists in the form of dispersion, heterogeneous , and traditional data cleaning is not necessarily Suitable for knowledge extraction, many problems cannot be resolved, therefore requires targeted design to extract tool capabilities for knowledge source formats and knowledge extraction. At present, we use self-developed regular expressible tool TIE as machine data knowledge extraction tool; for document knowledge extraction, the situation is slightly more complicated, first we need to retain the chapter section classification layering in the product document organization structure, use Document metadata parses the XML tag, obtains a particle size knowledge of the paragraph sentence sub-level, then requires the use of neural network model and NLP tool needle to extract the word level fine-grained knowledge, including classification, relationships between entity words and feature words. Usually the results need to be selected iteration and verifying to improve new words discovery accuracy, this way to integrate data from different sources into different particle size, save In the knowledge base.

Knowledge representation and knowledge integration

Lean the knowledge obtained, the expression level is often sparse, it is insufficient, and it is often necessary to automatically pass through various algorithms. Mining, discover new relationships, doing knowledge to make up. The knowledge of knowledge we need includes not only the relationship between the entities, but also the completion of various fault feature conduction relationships. For example, the possible cause of failure A may be a meaning that may affect the possible influence of failure B, which requires a similar relationship between “reasons” and “influence”. Such knowledge is complementary to complement the fine granular level knowledge extraction.

Pure text data is obtained entity identification, physical link, physical link, relationship, concept abstract , etc., need to use many natural language processing, including but not only Limited to words, sample labels, mean labels, synonyms extraction, etc.

Do a good job in knowledge processing, just completing a part of the development of AI application development, how to use the obtained knowledge, the most important thing is to resolve key application scenarios, realize business value, can reflect the value of technology .

If you wish to make your career in network, the Certifications is considered to be the best certification, to jump-start your career. But gaining this certification isn’t considered to be that much easy. You have to go through lots and lots of study process unless you have the help of the CCNP 300–835 Dumps offered at the EveDumps.