AWS AIP-C01 开源练习 · 基础 · 第 02 套

英文原题与中文机器翻译。本批暂不提供日文题文;日语界面显示英文原文。

开源练习,非 AWS 官方试题。按作者参考答案判分,解析附官方文档链接。

← 题库目录

已答 0/50正确 0错误 0正确率 —
Question 1Domain 1 · Task 1.3 · Select 2

Where can Amazon Comprehend add value in a foundation-model pipeline? (Select TWO.)

Amazon Comprehend 可以在基础模型管道的哪些方面增加价值? (选择两项。)

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. Comprehend can clean and enrich inputs before inference.

B. Comprehend can analyze model outputs after generation.

C. That is Provisioned Throughput, not Comprehend.

D. That is Transcribe's job, not Comprehend's.

E. Vector storage is handled by vector stores, not Comprehend.

A. Comprehend 可以在推理之前清理和丰富输入。

B. Comprehend 可以分析生成后的模型输出。

C. 那是预置吞吐量,而不是理解。

D. 这是转录的工作,而不是理解的工作。

E. 向量存储由向量存储处理,而不是 Comprehend。

AWS Docs

Question 2Domain 1 · Task 1.3

A multimodal ingestion pipeline must take audio recordings, transcribe them, then detect and redact PII before the transcripts are embedded for RAG. Which ordered combination of services is correct?

多模式摄取管道必须获取音频录音,对其进行转录,然后在为 RAG 嵌入转录文本之前检测并编辑 PII。哪种有序的服务组合是正确的?

解析 / Explanation

正确答案 / Answer: A

A. Transcribe converts audio to text, Comprehend redacts PII, then the cleaned text is embedded.

B. Textract is for documents and Polly is text-to-speech; neither fits audio transcription + redaction.

C. You must transcribe audio to text before Comprehend can analyze it; the order is reversed.

D. Rekognition analyzes images/video, not audio transcription.

A. Transcribe 将音频转换为文本,Comprehend 编辑 PII,然后嵌入清理后的文本。

B. Textract 用于文档,Polly 用于文本转语音;都不适合音频转录+编辑。

C. 您必须先将音频转录为文本,然后 Comprehend 才能对其进行分析;顺序颠倒了。

D. Rekognition 分析图像/视频,而不是音频转录。

AWS Docs

Question 3Domain 1 · Task 1.3

A team wants to scan an Amazon S3 data lake to discover and classify where sensitive data (like PII) resides for security and compliance posture. Which service is the right fit?

一个团队想要扫描 Amazon S3 数据湖,以发现敏感数据(如 PII)所在的位置并对其进行分类,以确保安全性和合规性。哪种服务最合适?

解析 / Explanation

正确答案 / Answer: A

A. Macie discovers and classifies sensitive data in S3 for security/compliance visibility.

B. Comprehend detects/redacts PII in text you send it, but discovery/classification across an S3 lake is Macie's role.

C. Textract extracts text from documents; it is not a data-discovery tool.

D. Glue Data Quality scores dataset quality, not sensitive-data discovery.

A. Macie 发现 S3 中的敏感数据并对其进行分类,以实现安全/合规性可见性。

B. Comprehend 会检测/编辑您发送的文本中的 PII,但跨 S3 湖的发现/分类是 Macie 的职责。

C. Textract 从文档中提取文本;它不是数据发现工具。

D. Glue 数据质量对数据集质量进行评分,而不是对敏感数据发现进行评分。

AWS Docs

Question 4Domain 1 · Task 1.3

Which statement best distinguishes Amazon Textract from Amazon Comprehend?

哪种说法最能区分 Amazon Textract 和 Amazon Comprehend?

解析 / Explanation

正确答案 / Answer: A

A. Textract is OCR/extraction; Comprehend is NLP analysis of text.

B. Neither does speech-to-text; that is Transcribe.

C. This reverses their roles.

D. Neither is a vector store.

A. Textract是OCR/提取; Comprehend 是对文本的 NLP 分析。

B. 语音转文本也不行。那就是转录。

C. 这颠倒了他们的角色。

D. 两者都不是矢量存储。

AWS Docs

Question 5Domain 1 · Task 1.3

A team building a chatbot wants automated, low-code intelligent document processing to extract content from mixed documents and media (text, tables, images, audio, video) to feed Bedrock. Which AWS capability is designed for this multimodal extraction?

构建聊天机器人的团队需要自动化、低代码智能文档处理,以从混合文档和媒体(文本、表格、图像、音频、视频)中提取内容以提供给 Bedrock。哪种 AWS 功能是专为这种多模式提取而设计的?

解析 / Explanation

正确答案 / Answer: A

A. Bedrock Data Automation extracts and normalizes content from unstructured multimodal documents/media for GenAI use.

B. Polly is text-to-speech only.

C. Kendra provides search over data; it is not a multimodal extraction pipeline.

D. Glue Data Quality scores datasets; it does not extract multimodal content.

A. Bedrock Data Automation 从非结构化多模式文档/媒体中提取内容并对其进行规范化,以供 GenAI 使用。

B. Polly 只能进行文本转语音。

C. Kendra 提供数据搜索;它不是多模式提取管道。

D. Glue 数据质量评分数据集;它不提取多模式内容。

AWS Docs

Question 6Domain 1 · Task 1.3

A team needs repeatable, scheduled, large-scale feature engineering on terabytes of tabular data with custom Python/Spark code as part of an ML pipeline. Which is most appropriate?

团队需要使用自定义 Python/Spark 代码作为 ML 管道的一部分,对 TB 级的表格数据进行可重复、预定的大规模特征工程。哪个最合适?

解析 / Explanation

正确答案 / Answer: A

A. Processing jobs run custom containerized code at scale on managed infrastructure, ideal for repeatable pipelines.

B. Data Wrangler is for interactive exploration; large-scale repeatable jobs fit Processing.

C. That is for text classification, not general tabular feature engineering.

D. Transcribe handles audio, not tabular feature engineering.

A. 处理作业在托管基础设施上大规模运行自定义容器化代码,非常适合可重复的管道。

B. Data Wrangler 用于交互式探索;适合大规模可重复作业的加工。

C. 这是用于文本分类,而不是一般的表格特征工程。

D. Transcribe 处理音频,而不是表格特征工程。

AWS Docs

Question 7Domain 1 · Task 1.3

A team wants to add automated data-quality rules (completeness, uniqueness, freshness) that fail a pipeline run when thresholds are breached, before data is used to fine-tune a model. Which service provides built-in rule evaluation and anomaly detection?

团队希望添加自动化的数据质量规则(完整性、唯一性、新鲜度),在使用数据微调模型之前,当违反阈值时,管道运行将失败。哪种服务提供内置规则评估和异常检测?

解析 / Explanation

正确答案 / Answer: A

A. Glue Data Quality defines and evaluates rules and detects anomalies inside Glue pipelines.

B. Synthetics tests endpoints/canaries, not dataset quality.

C. Comprehend analyzes text content, not data-quality rule thresholds.

D. Athena queries data; it does not provide built-in data-quality rules/anomaly detection like Glue Data Quality.

A. Glue 数据质量定义和评估规则并检测 Glue 管道内的异常情况。

B. 综合测试端点/金丝雀,而不是数据集质量。

C. Comprehend 分析文本内容,而不是数据质量规则阈值。

D. 雅典娜查询数据;它不提供像 Glue Data Quality 那样的内置数据质量规则/异常检测。

AWS Docs

Question 8Domain 1 · Task 1.3

A team must redact PII from free-text fields, but keep the text otherwise intact and usable for retrieval. Which is the best approach?

团队必须从自由文本字段中编辑 PII,但保持文本完整并可用于检索。哪种方法是最好的?

解析 / Explanation

正确答案 / Answer: A

A. Comprehend reliably detects and redacts PII entities while preserving the remaining content.

B. Prompt instructions are not reliable and can be bypassed; redact before indexing.

C. Encryption protects at rest but does not selectively redact PII for usable retrieval.

D. Discarding the data destroys its usefulness for retrieval.

A. Comprehend 可靠地检测和编辑 PII 实体,同时保留剩余内容。

B. 提示指令不可靠,可以绕过;在索引之前进行编辑。

C. 加密可提供静态保护,但不会有选择地编辑 PII 以供检索使用。

D. 丢弃数据会破坏其检索的有用性。

AWS Docs

Question 9Domain 1 · Task 1.3

Which service should transform raw, messy tabular data into ML-ready features using a point-and-click flow that can then be exported to a SageMaker pipeline?

哪种服务应使用点击式流程将原始、混乱的表格数据转换为 ML 就绪功能,然后将其导出到 SageMaker 管道?

解析 / Explanation

正确答案 / Answer: A

A. Data Wrangler offers visual transforms and can export the flow into a pipeline/processing job.

B. Textract extracts document text, not tabular feature transforms.

C. Knowledge Bases handle RAG ingestion, not tabular feature prep.

D. Macie discovers sensitive data; it is not a data-prep canvas.

A. Data Wrangler 提供可视化转换,并可以将流导出到管道/处理作业中。

B. Textract 提取文档文本,而不是表格特征转换。

C. 知识库处理 RAG 摄取,而不是表格特征准备。

D. Macie 发现敏感数据;它不是数据准备画布。

AWS Docs

Question 10Domain 1 · Task 1.3

An ingestion flow processes scanned contracts. The team needs both the raw text AND the key-value pairs from form fields. Which service extracts both?

摄取流程处理扫描的合同。团队需要原始文本和表单字段中的键值对。哪个服务提取两者?

解析 / Explanation

正确答案 / Answer: A

A. Textract returns raw text and structured form key-value pairs and tables.

B. Comprehend analyzes text but does not OCR documents or pull form fields.

C. Transcribe handles audio, not document forms.

D. Polly is text-to-speech.

A. Textract 返回原始文本和结构化表单键值对和表格。

B. Comprehend 分析文本,但不会 OCR 文档或提取表单字段。

C. Transcribe 处理音频,而不是文档表单。

D. Polly 是文本转语音。

AWS Docs

Question 11Domain 1 · Task 1.3

A team wants sentiment, key phrases, and entity recognition on millions of product reviews to enrich metadata before RAG indexing. Which service is the best fit?

团队希望对数百万条产品评论进行情感、关键短语和实体识别,以在 RAG 索引之前丰富元数据。哪种服务最适合?

解析 / Explanation

正确答案 / Answer: A

A. Comprehend provides sentiment, key-phrase, and entity extraction at scale for text enrichment.

B. Textract extracts text from documents; it does not perform sentiment/entity NLP.

C. Transcribe converts speech to text; not NLP enrichment.

D. Glue Data Quality scores dataset quality, not text NLP.

A. Comprehend 提供大规模的情感、关键短语和实体提取,以丰富文本。

B. Textract 从文档中提取文本;它不执行情感/实体 NLP。

C. Transcribe 将语音转换为文本;不是 NLP 强化。

D. Glue 数据质量对数据集质量进行评分,而不是文本 NLP。

AWS Docs

Question 12Domain 1 · Task 1.3

Which pairing correctly maps a service to its primary role in data preparation?

哪种配对正确地将服务映射到其在数据准备中的主要角色?

解析 / Explanation

正确答案 / Answer: A

A. This correctly summarizes each service's core role.

B. These roles are scrambled.

C. None of them is a vector store.

D. Capacity reservation is Provisioned Throughput, not these services.

A. 这正确地概括了每个服务的核心角色。

B. 这些角色都是乱七八糟的。

C. 它们都不是矢量存储。

D. 容量预留是预配置吞吐量,而不是这些服务。

AWS Docs

Question 13Domain 1 · Task 1.3

A team is choosing between Amazon Comprehend and Amazon Macie for handling PII. Which guidance is correct?

一个团队正在 Amazon Comprehend 和 Amazon Macie 之间进行选择来处理 PII。哪个指导是正确的?

解析 / Explanation

正确答案 / Answer: A

A. Comprehend redacts text content; Macie is for S3-wide sensitive-data discovery and posture.

B. Macie is a discovery/classification service for S3, not a real-time text redactor.

C. Lake-wide discovery/classification is Macie's role.

D. They serve different purposes and are not interchangeable.

A. 理解编辑文本内容; Macie 用于 S3 范围内的敏感数据发现和状态。

B. Macie 是 S3 的发现/分类服务,而不是实时文本编辑器。

C. Macie 的职责是全湖范围内的发现/分类。

D. 它们有不同的用途并且不能互换。

AWS Docs

Question 14Domain 1 · Task 1.3

A GenAI ingestion pipeline must handle PDFs, images, audio, and video and output normalized text/insights for a knowledge base with the least custom code. Which is the most fitting first choice?

GenAI 摄取管道必须处理 PDF、图像、音频和视频,并使用最少的自定义代码输出知识库的规范化文本/见解。哪个是最合适的首选?

解析 / Explanation

正确答案 / Answer: A

A. Bedrock Data Automation is built to ingest multimodal content and produce normalized output for GenAI with minimal code.

B. Hand-built parsers per format add significant operational overhead.

C. Transcribe only covers audio; it cannot handle PDFs, images, and video extraction.

D. Textract only covers documents/images, not audio/video.

A. Bedrock Data Automation 旨在摄取多模式内容并以最少的代码为 GenAI 生成标准化输出。

B. 每种格式手工构建的解析器会增加大量的操作开销。

C. 转录仅涵盖音频;它无法处理 PDF、图像和视频提取。

D. Textract 仅涵盖文档/图像,不涵盖音频/视频。

AWS Docs

Question 15Domain 1 · Task 1.4

A team wants managed Retrieval-Augmented Generation that automatically chunks documents, generates embeddings, and stores them in a vector store, exposing retrieval to a foundation model. Which Amazon Bedrock feature provides this?

一个团队需要托管的检索增强生成,该生成可以自动对文档进行分块、生成嵌入并将其存储在向量存储中,从而将检索公开给基础模型。哪个 Amazon Bedrock 功能可以提供此功能?

解析 / Explanation

正确答案 / Answer: A

A. Knowledge Bases handle ingestion (chunk, embed, index) and provide retrieval APIs for RAG.

B. Guardrails filter content; they do not build a RAG pipeline.

C. Prompt Flows orchestrate prompt steps; they are not a managed RAG ingestion pipeline.

D. That reserves capacity; it is unrelated to RAG ingestion.

A. 知识库处理摄取(块、嵌入、索引)并为 RAG 提供检索 API。

B. 护栏过滤内容;他们不构建 RAG 管道。

C. 提示流程编排提示步骤;它们不是托管 RAG 摄取管道。

D. 即储备能力;它与 RAG 摄入无关。

AWS Docs

Question 16Domain 1 · Task 1.4

When you create an Amazon Bedrock Knowledge Base and accept the defaults for the vector store, which vector database is provisioned by default?

当您创建 Amazon Bedrock 知识库并接受矢量存储的默认设置时,默认情况下会预置哪个矢量数据库?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch Serverless is the default vector store created for a Knowledge Base.

B. DynamoDB is not a Knowledge Base vector store option.

C. Redshift is a data warehouse, not the default KB vector store.

D. MySQL/RDS is not the default KB vector store.

A. OpenSearch Serverless 是为知识库创建的默认矢量存储。

B. DynamoDB 不是知识库矢量存储选项。

C. Redshift 是一个数据仓库,而不是默认的知识库矢量存储。

D. MySQL/RDS 不是默认的 KB 向量存储。

AWS Docs

Question 17Domain 1 · Task 1.4 · Select 2

Which TWO are supported vector stores you can use with an Amazon Bedrock Knowledge Base?

您可以将哪两个受支持的矢量存储与 Amazon Bedrock 知识库一起使用?

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. Aurora PostgreSQL (pgvector) is a supported KB vector store.

B. Neptune Analytics is a supported KB vector store option.

C. SQS is a message queue, not a vector store.

D. CloudFront is a CDN, not a vector store.

E. SNS is pub/sub messaging, not a vector store.

A. Aurora PostgreSQL (pgvector) 是受支持的 KB 矢量存储。

B. Neptune Analytics 是受支持的 KB 矢量存储选项。

C. SQS 是一个消息队列,而不是向量存储。

D. CloudFront 是 CDN,而不是矢量存储。

E. SNS 是发布/订阅消息传递,而不是矢量存储。

AWS Docs

Question 18Domain 1 · Task 1.4

A developer needs only the relevant retrieved document chunks back from a Knowledge Base so they can build a fully custom prompt themselves. Which API should they call?

开发人员只需要从知识库检索相关的文档块,这样他们就可以自己构建完全自定义的提示。他们应该调用哪个 API?

解析 / Explanation

正确答案 / Answer: A

A. Retrieve returns matching chunks only, leaving generation to the caller.

B. RetrieveAndGenerate also calls an FM to produce an answer; the developer wants chunks only.

C. InvokeModel calls a model directly without KB retrieval.

D. That ingests data into the KB; it does not retrieve chunks.

A. 检索仅返回匹配的块,将生成留给调用者。

B. RetrieveAndGenerate 还调用 FM 来生成答案;开发人员只想要块。

C. InvokeModel 直接调用模型,无需检索 KB。

D. 将数据摄取到知识库中;它不检索块。

AWS Docs

Question 19Domain 1 · Task 1.4

A team wants the Knowledge Base to both retrieve relevant chunks AND have a foundation model generate a grounded answer in one managed call. Which API fits?

团队希望知识库既能检索相关块,又能让基础模型在一次托管调用中生成有依据的答案。哪个 API 适合?

解析 / Explanation

正确答案 / Answer: A

A. RetrieveAndGenerate performs retrieval and then FM generation in a single call.

B. Retrieve only returns chunks; it does not generate an answer.

C. That provisions a KB; it is not a query API.

D. Converse is a chat API without built-in KB retrieval orchestration.

A. RetrieveAndGenerate 在一次调用中执行检索,然后执行 FM 生成。

B. 检索仅返回块;它不会生成答案。

C. 这提供了一个知识库;它不是一个查询 API。

D. Converse 是一个没有内置知识库检索编排的聊天 API。

AWS Docs

Question 20Domain 1 · Task 1.4 · Select 2

Which TWO data sources can an Amazon Bedrock Knowledge Base ingest from for RAG?

Amazon Bedrock 知识库可以从哪两个数据源提取 RAG?

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. S3 is a primary supported data source for Knowledge Bases.

B. Knowledge Bases can crawl websites as a data source.

C. SQS is not a KB data source.

D. Instance store is ephemeral local disk, not a KB data source.

E. IAM manages access; it is not a data source.

A. S3 是知识库的主要支持数据源。

B. 知识库可以抓取网站作为数据源。

C. SQS 不是 KB 数据源。

D. 实例存储是临时本地磁盘,而不是 KB 数据源。

E. IAM 管理访问;它不是数据源。

AWS Docs

Question 21Domain 1 · Task 1.4

For a Bedrock Knowledge Base to read from and write to its OpenSearch Serverless vector store, what must be configured?

对于要读取和写入 OpenSearch Serverless 矢量存储的 Bedrock 知识库,必须配置哪些内容?

解析 / Explanation

正确答案 / Answer: A

A. Bedrock assumes an IAM role with permissions to the vector collection to ingest and query.

B. Public exposure is not required and is a security risk.

C. Network plumbing alone does not grant Bedrock access; the IAM role does.

D. OpenSearch capacity is OCU-based; the access requirement is an IAM role.

A. Bedrock 承担 IAM 角色,有权摄取和查询矢量集合。

B. 公开曝光不是必需的,并且存在安全风险。

C. 网络管道本身并不能授予 Bedrock 访问权限; IAM 角色可以。

D. OpenSearch容量基于OCU;访问要求是 IAM 角色。

AWS Docs

Question 22Domain 1 · Task 1.4

A company runs frequent, high-throughput, low-latency vector searches and wants a managed vector store that autoscales without capacity planning, also used as the default for Bedrock Knowledge Bases. Which is the best fit?

一家公司运行频繁、高吞吐量、低延迟的向量搜索,并希望有一个无需容量规划即可自动扩展的托管向量存储,也用作 Bedrock 知识库的默认值。哪个最合适?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch Serverless autoscales for frequent low-latency vector search and is the KB default.

B. S3 Vectors targets cost-effective, infrequent queries, not high-throughput low-latency workloads.

C. DocumentDB supports vectors but is not the autoscaling default for KB high-throughput search.

D. DynamoDB has no native vector search.

A. OpenSearch Serverless 可自动缩放以进行频繁的低延迟矢量搜索,并且是 KB 默认设置。

B. S3 Vectors 的目标是经济高效、不频繁的查询,而不是高吞吐量、低延迟的工作负载。

C. DocumentDB 支持向量,但不是 KB 高吞吐量搜索的默认自动缩放。

D. DynamoDB 没有本机矢量搜索。

AWS Docs

Question 23Domain 1 · Task 1.4

In Amazon OpenSearch Serverless, capacity is measured in OpenSearch Compute Units (OCUs). Which statement is correct?

在 Amazon OpenSearch Serverless 中,容量以 OpenSearch 计算单元 (OCU) 来衡量。哪种说法是正确的?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch Serverless scales indexing and search capacity separately, matching workloads that read a lot but index rarely.

B. OCUs are split into search and indexing, which scale independently.

C. OCUs measure compute capacity, not just storage.

D. OpenSearch Serverless supports time-series, search, and vector workloads.

A. OpenSearch Serverless 分别扩展索引和搜索容量,匹配读取大量但索引很少的工作负载。

B. OCU 分为搜索和索引两部分,两者可独立扩展。

C. OCU 衡量计算能力,而不仅仅是存储。

D. OpenSearch Serverless 支持时间序列、搜索和矢量工作负载。

AWS Docs

Question 24Domain 1 · Task 1.4

A team uses a self-managed Amazon OpenSearch Service domain for vector search. Which plugin enables k-nearest-neighbor vector similarity search?

团队使用自我管理的 Amazon OpenSearch Service 域进行矢量搜索。哪个插件支持 k 最近邻向量相似度搜索?

解析 / Explanation

正确答案 / Answer: A

A. The k-NN plugin provides approximate/exact nearest-neighbor vector search in OpenSearch.

B. The SQL plugin enables SQL queries, not vector kNN search.

C. Alerting triggers notifications; it is not for vector search.

D. ISM manages index lifecycle, not vector similarity.

A. k-NN 插件在 OpenSearch 中提供近似/精确最近邻向量搜索。

B. SQL 插件支持 SQL 查询,而不是向量 kNN 搜索。

C. 警报触发通知;它不适用于矢量搜索。

D. ISM 管理索引生命周期,而不是向量相似性。

AWS Docs

Question 25Domain 1 · Task 1.4

Within OpenSearch, which plugin enables neural/semantic search and retrieval that can integrate with Bedrock for RAG-style queries?

在 OpenSearch 中,哪个插件支持神经/语义搜索和检索,可以与 Bedrock 集成以进行 RAG 式查询?

解析 / Explanation

正确答案 / Answer: A

A. Neural Search enables semantic vector retrieval and can integrate with Bedrock models.

B. The Security plugin manages access control, not semantic search.

C. Anomaly Detection finds outliers in data, not semantic retrieval.

D. That replicates indices across clusters; it is not semantic search.

A. 神经搜索支持语义向量检索,并可以与基岩模型集成。

B. 安全插件管理访问控制,而不是语义搜索。

C. 异常检测发现数据中的异常值,而不是语义检索。

D. 跨集群复制索引;这不是语义搜索。

AWS Docs

Question 26Domain 1 · Task 1.4

An engineer increases the number of shards in an OpenSearch vector index. What is a likely trade-off to be aware of?

工程师增加了 OpenSearch 矢量索引中的分片数量。需要注意的可能权衡是什么?

解析 / Explanation

正确答案 / Answer: A

A. Sharding scales capacity but adds coordination overhead that can raise latency.

B. Over-sharding can hurt latency; it is a trade-off, not a free win.

C. Sharding directly affects how much data can be distributed/stored.

D. Sharding affects vector indices too.

A. 分片可以扩展容量,但会增加协调开销,从而增加延迟。

B. 过度分片会损害延迟;这是一种权衡,而不是免费的胜利。

C. 分片直接影响可以分发/存储的数据量。

D. 分片也会影响向量索引。

AWS Docs

Question 27Domain 1 · Task 1.4

A team tunes an HNSW vector index in OpenSearch. What does the "m" parameter control?

团队在 OpenSearch 中调整 HNSW 矢量索引。 “m”参数控制什么?

解析 / Explanation

正确答案 / Answer: A

A. m sets graph connectivity; raising it boosts accuracy at the cost of memory and speed.

B. That is shard sizing, not the HNSW m parameter.

C. Dimension is a property of the vectors, not the HNSW m parameter.

D. Temperature is an inference parameter, unrelated to HNSW.

A. m 设置图连通性;提高它会以内存和速度为代价提高准确性。

B. 这是分片大小,而不是 HNSW m 参数。

C. 维度是向量的属性,而不是 HNSW m 参数。

D. 温度是一个推断参数,与 HNSW 无关。

AWS Docs

Question 28Domain 1 · Task 1.4

In an HNSW index, what does increasing "ef_construction" do?

在 HNSW 索引中,增加“ef_construction”有什么作用?

解析 / Explanation

正确答案 / Answer: A

A. ef_construction sets the candidate list size during graph construction — higher is more accurate but slower to build.

B. It does not change vector dimensionality.

C. It is unrelated to sharding.

D. It is an index parameter, not an inference setting.

A. ef_construction 在图构建期间设置候选列表大小 - 越高,越准确,但构建速度较慢。

B. 它不会改变向量维度。

C. 它与分片无关。

D. 它是一个索引参数,而不是一个推理设置。

AWS Docs

Question 29Domain 1 · Task 1.4

Beyond vector search, what else is Amazon OpenSearch Service commonly used for?

除了矢量搜索之外,Amazon OpenSearch Service 还常用于哪些用途?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch is a search and analytics suite covering full-text search, logs, and time-series in addition to vectors.

B. Relational OLTP is for databases like Aurora, not OpenSearch.

C. OpenSearch does not train FMs.

D. Queuing is SQS; OpenSearch is search/analytics.

A. OpenSearch 是一个搜索和分析套件,除了向量之外还涵盖全文搜索、日志和时间序列。

B. 关系型 OLTP 适用于 Aurora 等数据库,而不适用于 OpenSearch。

C. OpenSearch 不训练 FM。

D. 排队是SQS; OpenSearch 是搜索/分析。

AWS Docs

Question 30Domain 1 · Task 1.4

A company stores JSON documents in Amazon DocumentDB (MongoDB-compatible) and wants to add vector similarity search to the same datastore. Which is true?

一家公司将 JSON 文档存储在 Amazon DocumentDB(与 MongoDB 兼容)中,并希望将向量相似性搜索添加到同一数据存储中。哪个是真的?

解析 / Explanation

正确答案 / Answer: A

A. DocumentDB adds vector search indexes such as HNSW and IVFFlat for its JSON data.

B. DocumentDB is for JSON documents, not arbitrary blob objects.

C. It does support vector search.

D. DocumentDB performs vector search itself; DynamoDB is unrelated.

A. DocumentDB为其JSON数据添加了向量搜索索引,例如HNSW和IVFFlat。

B. DocumentDB 适用于 JSON 文档,而不是任意 blob 对象。

C. 它确实支持矢量搜索。

D. DocumentDB 本身执行矢量搜索; DynamoDB 不相关。

AWS Docs

Question 31Domain 1 · Task 1.4

A team needs a relational database that can store transactional data AND vector embeddings together for a RAG feature. Which option fits best?

团队需要一个关系数据库,可以将事务数据和向量嵌入存储在一起以实现 RAG 功能。哪个选项最适合?

解析 / Explanation

正确答案 / Answer: A

A. Aurora PostgreSQL + pgvector combines relational data with vector search in one store.

B. DynamoDB is NoSQL and lacks native vector search.

C. S3 Vectors stores vectors but is not a relational transactional database.

D. Standard Neptune is a graph DB; relational + vectors is Aurora pgvector.

A. Aurora PostgreSQL + pgvector 将关系数据与矢量搜索结合在一个存储中。

B. DynamoDB 是 NoSQL,缺乏原生向量搜索。

C. S3 Vectors 存储向量,但不是关系事务数据库。

D. 标准Neptune是一个图数据库;关系+向量是Aurora pgvector。

AWS Docs

Question 32Domain 1 · Task 1.4

A company wants enterprise search across internal content with built-in connectors to sources like Salesforce, SharePoint, and ServiceNow, and no custom ETL to keep data fresh. Which service fits best?

A company wants enterprise search across internal content with built-in connectors to sources like Salesforce, SharePoint, and ServiceNow, and no custom ETL to keep data fresh.哪种服务最适合?

解析 / Explanation

正确答案 / Answer: A

A. Kendra provides intelligent enterprise search with managed connectors to third-party sources, refreshing on a schedule without custom ETL.

B. Athena queries data in S3 via SQL; it has no enterprise-search connectors.

C. QuickSight is BI dashboards, not enterprise search.

D. Comprehend is NLP analysis, not enterprise search with connectors.

A. Kendra 提供智能企业搜索以及与第三方源的托管连接器,无需自定义 ETL 即可按计划进行刷新。

B. Athena 通过 SQL 查询 S3 中的数据;它没有企业搜索连接器。

C. QuickSight 是 BI 仪表板,而不是企业搜索。

D. Comprehend 是 NLP 分析,而不是带有连接器的企业搜索。

AWS Docs

Question 33Domain 1 · Task 1.4 · Select 2

What are TWO advantages of using Amazon Kendra's managed connectors for sources like SharePoint or ServiceNow?

对于 SharePoint 或 ServiceNow 等源使用 Amazon Kendra 的托管连接器有哪两个优势?

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. Kendra ships managed connectors, removing custom ingestion code.

B. Connectors can sync on a schedule so content stays current.

C. Search quality helps grounding but does not guarantee zero hallucination.

D. Access still requires properly scoped permissions.

E. Kendra is search; it does not train FMs.

A. Kendra 提供了托管连接器,删除了自定义摄取代码。

B. 连接器可以按计划同步,以便内容保持最新。

C. 搜索质量有助于扎根,但不能保证零幻觉。

D. 访问仍然需要适当范围的权限。

E. 肯德拉是搜索;它不训练 FM。

AWS Docs

Question 34Domain 1 · Task 1.4

A diagnostic app stores 50 million image embeddings and runs similarity searches only occasionally, wanting the most cost-effective, serverless option with no infrastructure to manage. Which vector store is best?

诊断应用程序存储 5000 万个图像嵌入,仅偶尔运行相似性搜索,需要最具成本效益的无服务器选项,无需管理基础设施。哪个矢量商店最好?

解析 / Explanation

正确答案 / Answer: A

A. S3 Vectors is serverless, scales to billions of vectors, and is cost-optimized for infrequent queries — ideal here.

B. OpenSearch Serverless is built for frequent low-latency search; provisioned capacity sits idle for rare queries.

C. Aurora requires managing a database instance and is not cost-optimal for infrequent workloads.

D. DynamoDB has no native vector search; custom logic adds latency and cost.

A. S3 Vectors 是无服务器的,可扩展到数十亿个向量,并且针对不频繁的查询进行了成本优化 - 在这里非常理想。

B. OpenSearch Serverless 专为频繁的低延迟搜索而构建;预配置容量闲置以用于罕见的查询。

C. Aurora 需要管理数据库实例,对于不频繁的工作负载来说并不是成本最优的。

D. DynamoDB 没有原生矢量搜索;自定义逻辑会增加延迟和成本。

AWS Docs

Question 35Domain 1 · Task 1.4

In Amazon S3 Vectors, how is vector data organized?

在 Amazon S3 Vectors 中,矢量数据是如何组织的?

解析 / Explanation

正确答案 / Answer: A

A. S3 Vectors uses vector buckets containing vector indexes; an index fixes dimension/metric and is the unit you query.

B. S3 Vectors uses a dedicated vector bucket type with vector indexes, not ordinary objects.

C. S3 Vectors stores vectors natively; DynamoDB is not involved.

D. Many vectors and many indexes live within one vector bucket.

A. S3 Vectors使用包含向量索引的向量桶;索引固定维度/指标,是您查询的单位。

B. S3 Vectors 使用带有向量索引的专用向量桶类型,而不是普通对象。

C. S3 Vectors 原生存储向量; DynamoDB 不涉及。

D. 许多向量和许多索引存在于一个向量桶中。

AWS Docs

Question 36Domain 1 · Task 1.4

A team wants S3 Vectors as the vector store for a Bedrock Knowledge Base. Which statement is accurate?

一个团队希望 S3 Vectors 作为基岩知识库的矢量存储。哪种说法是准确的?

解析 / Explanation

正确答案 / Answer: A

A. S3 Vectors is a supported KB vector store; Bedrock ingests and writes embeddings into the S3 Vectors index.

B. It is supported as a KB vector store.

C. S3 Vectors is independent of OpenSearch.

D. S3 Vectors is purpose-built to store and query embeddings.

A. S3 Vectors 是受支持的 KB 向量存储; Bedrock 摄取嵌入并将其写入 S3 矢量索引。

B. 它作为 KB 向量存储受到支持。

C. S3 Vectors 独立于 OpenSearch。

D. S3 Vectors 专为存储和查询嵌入而构建。

AWS Docs

Question 37Domain 1 · Task 1.4

Which guidance best contrasts DynamoDB, OpenSearch, and Aurora pgvector for vector workloads?

对于矢量工作负载,哪项指南最能对比 DynamoDB、OpenSearch 和 Aurora pgvector?

解析 / Explanation

正确答案 / Answer: A

A. This matches each option's strengths for vector use cases.

B. DynamoDB lacks native vector search; OpenSearch is the strong vector choice.

C. Aurora is relational and adds vectors via pgvector.

D. OpenSearch is a leading vector-search option.

A. 这与矢量用例的每个选项的优势相匹配。

B. DynamoDB 缺乏原生矢量搜索; OpenSearch 是强有力的矢量选择。

C. Aurora 是关系型的,通过 pgvector 添加向量。

D. OpenSearch 是领先的矢量搜索选项。

AWS Docs

Question 38Domain 1 · Task 1.4

A near-real-time, resilient ingestion pipeline must add newly uploaded S3 documents to a Knowledge Base ASAP and remove deleted ones ASAP, tolerating spikes. Which architecture is best?

近乎实时、弹性的摄取管道必须尽快将新上传的 S3 文档添加到知识库,并尽快删除已删除的文档,以承受峰值。哪种架构最好?

解析 / Explanation

正确答案 / Answer: A

A. SQS buffers spikes and provides retries between S3 events and the Lambda, making ingestion resilient and event-driven.

B. Polling on a schedule is time-based, not near-real-time, and is inefficient.

C. Direct invocation lacks buffering/retries and can fail under load.

D. Manual steps do not meet the automated, near-real-time requirement.

A. SQS 缓冲峰值并在 S3 事件和 Lambda 之间提供重试,从而使摄取具有弹性和事件驱动。

B. 按计划进行的轮询是基于时间的,不是近乎实时的,而且效率低下。

C. 直接调用缺乏缓冲/重试,并且可能在负载下失败。

D. 手动步骤无法满足自动化、近实时的要求。

AWS Docs

Question 39Domain 1 · Task 1.4

A team needs to monitor Knowledge Base data ingestion to troubleshoot why some documents fail to process (e.g., ignored, embedding failed, indexing failed). Which solution is correct?

团队需要监控知识库数据摄取,以解决某些文档无法处理的原因(例如,被忽略、嵌入失败、索引失败)。哪个解决方案是正确的?

解析 / Explanation

正确答案 / Answer: A

A. KB logging records per-file ingestion status (e.g., RESOURCE_IGNORED, EMBEDDING_FAILED, INDEXING_FAILED) to CloudWatch.

B. CloudTrail logs API calls, not per-document ingestion status.

C. That logs inference calls, not KB ingestion.

D. Application Signals monitors app performance, not KB ingestion.

A. KB 日志记录将每个文件的摄取状态(例如 RESOURCE_IGNORED、EMBEDDING_FAILED、INDEXING_FAILED)记录到 CloudWatch。

B. CloudTrail 记录 API 调用,而不是每个文档的摄取状态。

C. 它记录推理调用,而不是知识库摄取。

D. 应用程序信号监视应用程序性能,而不是 KB 摄取。

AWS Docs

Question 40Domain 1 · Task 1.4

For exam purposes, which vector store is generally the "default/expected" answer for building RAG with Amazon Bedrock Knowledge Bases when no special cost or relational requirement is stated?

出于考试目的,在没有说明特殊成本或关系要求的情况下,哪个向量存储通常是使用 Amazon Bedrock 知识库构建 RAG 的“默认/预期”答案?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch Serverless is the default and most commonly expected KB vector store for RAG.

B. Redshift is a warehouse, not the KB vector store.

C. Memcached is a cache without vector search.

D. Timestream is a time-series DB, not a KB vector store.

A. OpenSearch Serverless 是 RAG 的默认且最常见的 KB 向量存储。

B. Redshift是一个仓库,而不是KB向量存储。

C. Memcached 是一个没有矢量搜索的缓存。

D. Timestream 是一个时间序列数据库,而不是 KB 向量存储。

AWS Docs

Question 41Domain 1 · Task 1.4 · Select 2

A Knowledge Base must pull content from a managed SaaS source. Which TWO are valid Bedrock Knowledge Base data-source connectors?

知识库必须从托管 SaaS 源中提取内容。哪两个是有效的基岩知识库数据源连接器?

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. Salesforce is a supported KB data source connector.

B. Confluence is a supported KB data source connector.

C. Lightsail is simplified compute hosting, not a data source connector.

D. Step Functions orchestrates workflows; it is not a KB data source.

E. Inspector is vulnerability scanning, not a data source.

A. Salesforce 是受支持的知识库数据源连接器。

B. Confluence 是受支持的 KB 数据源连接器。

C. Lightsail 是简化的计算托管,而不是数据源连接器。

D. Step Functions 协调工作流程;它不是 KB 数据源。

E. Inspector是漏洞扫描,而不是数据源。

AWS Docs

Question 42Domain 1 · Task 1.4

Which statement about the Retrieve vs RetrieveAndGenerate APIs is correct?

关于 Retrieve 与 RetrieveAndGenerate API 的哪种说法是正确的?

解析 / Explanation

正确答案 / Answer: A

A. Retrieve = retrieval only; RetrieveAndGenerate = retrieval + generation.

B. Retrieve does not call an FM.

C. It also generates an answer.

D. Both are core RAG APIs for Knowledge Bases.

A. 检索 = 仅检索;检索并生成 = 检索 + 生成。

B. Retrieve 不会呼叫 FM。

C. 它还会生成一个答案。

D. 两者都是知识库的核心 RAG API。

AWS Docs

Question 43Domain 1 · Task 1.4

A team must store binary (quantized) vectors to save memory in their vector store for a Knowledge Base. Which store supports binary vectors?

团队必须存储二进制(量化)向量,以节省知识库向量存储中的内存。哪个商店支持二进制向量?

解析 / Explanation

正确答案 / Answer: A

A. OpenSearch Serverless and managed clusters are the KB vector stores that support binary vectors.

B. S3 Vectors is cost-optimized but is not the binary-vector option among these stores.

C. DynamoDB has no native vector support.

D. Timestream is time-series, not a vector store.

A. OpenSearch Serverless 和托管集群是支持二进制向量的知识库向量存储。

B. S3 Vectors 进行了成本优化,但不是这些商店中的二进制向量选项。

C. DynamoDB 没有本机矢量支持。

D. 时间流是时间序列,而不是向量存储。

AWS Docs

Question 44Domain 1 · Task 1.4

An OpenSearch Serverless workload ingests new data rarely but serves a very high volume of vector queries. How should capacity be approached?

OpenSearch Serverless 工作负载很少摄取新数据,但提供大量矢量查询。应该如何接近容量?

解析 / Explanation

正确答案 / Answer: A

A. Because search and indexing scale separately, you can right-size each to the workload.

B. Indexing OCUs do not serve query throughput; search OCUs do.

C. Model Units are a Bedrock model-capacity concept, not OpenSearch query capacity.

D. Networking does not change OCU-based search capacity.

A. 由于搜索和索引是分开扩展的,因此您可以根据工作负载调整各自的大小。

B. 索引 OCU 不提供查询吞吐量; search OCUs do.

C. 模型单元是基岩模型容量概念,而不是 OpenSearch 查询容量。

D. 联网不会改变基于 OCU 的搜索能力。

AWS Docs

Question 45Domain 1 · Task 1.4

A team chose Amazon Kendra and wants to combine it with a Bedrock FM for grounded answers. What is the typical role split?

一个团队选择了 Amazon Kendra,并希望将其与 Bedrock FM 结合起来以获得可靠的答案。典型的角色划分是什么?

解析 / Explanation

正确答案 / Answer: A

A. Kendra supplies retrieval/grounding; the FM composes the final response.

B. This reverses their roles.

C. Neither trains a model per query.

D. Capacity reservation is Provisioned Throughput, unrelated to Kendra.

A. Kendra 提供检索/接地; FM 构成最终响应。

B. 这颠倒了他们的角色。

C. 两者都不根据查询训练模型。

D. 容量预留是预置吞吐量,与 Kendra 无关。

AWS Docs

Question 46Domain 1 · Task 1.4

Which statement about IVFFlat vs HNSW indexes (as used in stores like DocumentDB) is most accurate?

关于 IVFFlat 与 HNSW 索引(在 DocumentDB 等商店中使用)的哪种说法最准确?

解析 / Explanation

正确答案 / Answer: A

A. HNSW (graph-based) and IVFFlat (inverted-file) are both ANN index types with different performance/recall trade-offs.

B. They are index structures, not encryption.

C. They are vector index types, not models.

D. They are for vector similarity search, not relational joins.

A. HNSW(基于图)和 IVFFlat(倒排文件)都是 ANN 索引类型,具有不同的性能/召回权衡。

B. 它们是索引结构,而不是加密。

C. 它们是向量索引类型,而不是模型。

D. 它们用于向量相似性搜索,而不是关系连接。

AWS Docs

Question 47Domain 1 · Task 1.4

A retrieval system over highly connected/graph-structured knowledge wants vector search integrated with graph relationships in a managed store usable by a Knowledge Base. Which option fits?

基于高度连接/图形结构知识的检索系统需要将矢量搜索与知识库可用的托管存储中的图形关系集成。哪个选项适合?

解析 / Explanation

正确答案 / Answer: A

A. Neptune Analytics combines graph analytics with vector search and is a supported KB vector store.

B. DynamoDB is key-value/document without graph+vector capability.

C. A standard bucket is object storage, not a graph+vector store.

D. CloudSearch is legacy text search, not a graph+vector KB store.

A. Neptune Analytics 将图形分析与矢量搜索相结合,是受支持的知识库矢量存储。

B. DynamoDB 是键值/文档,没有图+向量功能。

C. 标准存储桶是对象存储,而不是图+向量存储。

D. CloudSearch 是传统文本搜索,而不是图形+矢量知识库存储。

AWS Docs

Question 48Domain 1 · Task 1.5

In a Bedrock Knowledge Base, why is the source content split into "chunks" before embedding?

在基岩知识库中,为什么在嵌入之前将源内容分成“块”?

解析 / Explanation

正确答案 / Answer: A

A. Chunking keeps units small enough to embed and retrieve precisely, improving relevance and fitting context windows.

B. Chunking is about segmentation for retrieval, not encryption.

C. That is Provisioned Throughput, not chunking.

D. That is transcription, unrelated to chunking.

A. 分块使单元足够小,可以精确地嵌入和检索,从而提高相关性并适合上下文窗口。

B. 分块是为了检索分段,而不是加密。

C. 这是预置吞吐量,而不是分块。

D. 那是转录,与分块无关。

AWS Docs · AWS Docs

Question 49Domain 1 · Task 1.5 · Select 2

Which TWO chunking strategies are offered by Amazon Bedrock Knowledge Bases?

Amazon Bedrock 知识库提供哪两种分块策略?

多选 · 需选 2 项
解析 / Explanation

正确答案 / Answer: A,B

A. Fixed-size splits text into uniform token/character windows.

B. Semantic chunking splits on meaning boundaries to keep coherent ideas together.

C. There is no "temperature chunking"; temperature is an inference parameter.

D. Sharding is a vector-index concept, not a KB chunking strategy.

E. OCUs are OpenSearch capacity units, not a chunking strategy.

A. 固定大小将文本分割成统一的标记/字符窗口。

B. 语义分块在意义边界上进行划分,以保持连贯的想法。

C. 不存在“温度分块”;温度是一个推断参数。

D. 分片是一个向量索引概念,而不是 KB 分块策略。

E. OCU 是 OpenSearch 容量单位,而不是分块策略。

AWS Docs · AWS Docs

Question 50Domain 1 · Task 1.5

A document set has nested structure (sections, subsections) and the team wants retrieval to use both fine-grained child chunks and broader parent context. Which chunking strategy fits best?

文档集具有嵌套结构(节、小节),团队希望检索时同时使用细粒度的子块和更广泛的父上下文。哪种分块策略最适合?

解析 / Explanation

正确答案 / Answer: A

A. Hierarchical chunking creates parent/child chunks so retrieval can use granular matches with broader context.

B. A single large chunk loses granularity and hurts retrieval precision.

C. Fixed-size ignores the document's nested structure.

D. Random splits do not capture hierarchy and degrade relevance.

A. 分层分块创建父/子块,因此检索可以使用具有更广泛上下文的粒度匹配。

B. 单个大块会失去粒度并损害检索精度。

C. 固定大小会忽略文档的嵌套结构。

D. 随机拆分不会捕获层次结构并降低相关性。

AWS Docs · AWS Docs

答题卡 已答 0 · 正确 0 · 错误 0