Sophon base 3.1 launched mlops function to provide wings for the operation of enterprise AI capabilities

Knowledge map as AI From perceptual intelligence to cognitive intelligence , It is always the key technology to organically integrate multiple concepts and relationships and infer new knowledge . In the past, Star Ring Technology 2022 Spring new product launch week , Star ring technology knowledge map platform Sophon KG Launched 3.1 edition .

Sophon KG It provides enterprise level users with a model that supports multi model data processing , It combines distributed graph storage and graph computing engine 、 Basic capabilities and cutting-edge technologies of multiple self-developed product lines such as full-text search , Set knowledge extraction 、 The fusion 、 Storage 、 It is a basic software product integrating the whole life cycle of computational reasoning and application , Help enterprise users gather information “ Quantitative change ” Leap into a burst of wisdom “ qualitative change ”.

Star Ring Technology

be based on Sophon KG Completeness and business orientation , Users can create a one-stop solution for the industry's full stack of knowledge maps . The knowledge map platform supports low code map construction 、 Intelligent knowledge extraction 、 Multimodal knowledge storage and fusion 、 Multi form knowledge calculation and reasoning and multi-dimensional map analysis . In addition to the above link integrity , The platform also starts from the business scenario , A graph data model that precipitates several scenarios 、 Rule model and algorithm model , It can help customers quickly solve business problems in similar scenarios .

Take the financial industry as an example ,Sophon KG built-in NLP Model , Support entities in financial scenarios 、 Automatic extraction of relationships , It supports manual modification . It also supports automatic entity linking , Realize the one click graph of model annotation results . Take the entity extraction task as an example , adopt Bert Model to extract the underlying entities , use BiLSTM Sequence tagging , And pass CRF The model constrains the sequence ; Simultaneous adoption FLAT/Simple-Lexicon And other ways to do vocabulary enhanced entity recognition , So as to realize 、 Company and organization name 、 Address 、 The person's name 、 product 、 Entity extraction of time, etc .

be based on Sophon KG Build an intelligent investment research knowledge map 、 Policy knowledge map 、 Bond knowledge map 、 Industry knowledge maps such as upstream and downstream analysis of corporate banking and risk maps , It can help regulators quickly identify suspicious transactions , More intuitively discover money laundering gangs and high-risk customers ; It can also help the bank monitor the cash out merchants in the acquiring system 、 Identify the lender 、 The high-risk guarantee chain behind the guarantor 、 Guarantee Circle , Give early warning to financial institutions in a timely manner .

this , The new version Sophon KG 3.1 Based on the original product functions , The following features have been added ：

• newly added Text processing modular , Support entity annotation 、 Semantic relation annotation 、 Single text classification 、 Multi text classification 、 The five text annotation tasks of emotion classification ;
• preset Financial scenario Entity extraction and semantic relation extraction model , Support model-based pre annotation , It also supports manual modification of annotation results ;
• Entity annotation and semantic relationship annotation tasks support One click map , Also support Point edge table export ;
• all Natural language tagging Tasks can be exported to model training data format after approval , It can be used as the input of subsequent model training ;
• Algorithms such as community clustering support edge Weight configuration .

In the applied scenario ： Anti money laundering 、 Anti fraud 、 Epidemic prevention and control 、 Public safety and enterprise level marketing , This update adds a new application scenario —— Insurance knowledge intelligent Q & A .Sophon KG Support product 、 type 、 Terms and other data are modeled through the atlas , Build a professional knowledge base , The answer is clearly shown in the map , Be clear at a glance . At the same time, on the basis of knowledge base , Through natural language processing technology , Realize automatic semantic retrieval of product terms and other issues 、 Question and answer , Provide an employee oriented 、 An encyclopedia of insurance knowledge for insurance agents . This solution directly hits the pain point of the industry ： It is difficult for the external manual customer service to master all the details of the company's insurance products , It is difficult to query efficiently without knowledge base , Reduce service efficiency ; And it is difficult for the insurance marketing team to master all the insurance products on the market , It may be difficult to find product information of other companies when facing customers , It is difficult to highlight the advantages of their own products .

Xinghuan science and technology knowledge map adheres to low code 、 Interactive 、 Visual product design principles , Let target users get started quickly , To improve the efficiency of problem solving , Bring users a more intelligent application experience .