使用 Amazon Neptune 的无服务器知识图谱进行混合搜索.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.D A T 4 3 2Use serverless knowledge graphs for hybrid search with Amazon NeptuneDave Bechberger(he/him)Principal Graph Architect,Amazon NeptuneMelissa Kwok(she/her)S

2、r.Neptune Specialist SA 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.What is hybrid search?Search systemKeyword searchVector searchKnowledge graphUser asks questionUser gets answer 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Graphs allow you to find relev

3、ant information that are connected/related concepts.Comparable concepts will be closely connected in graph space.Vectors allow you to find relevant information that is similar in language.Comparable sentences will be highly similar in vector space.Why combine graphs and vectors?2025,Amazon Web Servi

4、ces,Inc.or its affiliates.All rights reserved.Vector searchInformation access Primarily through vector similarity search Information similar to the question is structurally available Information dissimilar to question is structurally unavailable Unable to retrieve all relevant information 2025,Amazo

5、n Web Services,Inc.or its affiliates.All rights reserved.Vector searchSimilarUnrelatedOpposite 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Finding relevant but dissimilar informationRelevant information Related to the topic at handMultiple ways in which things are related Pare

6、nt-child Cause-effect Contiguity in time/space Social,organizational,legal,taxonomic relations Semantic similarity 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Graph search enhances results with topological knowledgeProvenance and lineageConnections within and between sourcesAu