Share E-Book
Scan to open this page

Scan with your phone to open this page

Author: Abedjan, Ziawasch, Esmailoghli, Mahdi, Galhotra, Sainyam

Rating 3.5 AI initial rating

No description

AI Reading Assistant

Whole-book reading guide from stratified index samples; jump to passages in the text

AI guide
【One-Line Pitch】 A compact, research-grounded map of how to find usable tables inside messy data lakes—covering discovery operations, indexing, systems, and benchmarks—for data management students, researchers, and engineers who need to navigate or build dataset-search systems. 【Book Arc】 - **Opening (~0%–12%)**: Frames the core problem: data lakes retain heterogeneous data cheaply but are underspecified and hard to exploit. Positions data discovery between information retrieval and data integration, and introduces the "vanilla" architecture of queries, table artifacts, operations, indexes, and matching. - **Early (~12%–28%)**: Defines discovery settings (including privacy/governance, multimodal, and temporal dimensions) and catalogs core operations—keyword search, union search, join search, operation sequences/DSLs, natural-language queries, and distributed/federated discovery. - **Early–Middle (~28%–36%)**: Turns to indexing strategies: value-based inverted indexes, column-based indexes (MinHash/LSH and beyond), correlation indexes, row/table/graph-based indexes, and knowledge-graph-enhanced navigation. - **Middle (~36%–48%)**: Surveys systems—academic prototypes such as AURUM, AUCTUS, and BLEND, plus industrial enterprise discovery platforms and marketplace infrastructures—and connects them back to the architectural view. - **Late (~48%–56%)**: Addresses evaluation: effectiveness metrics adapted from IR, benchmarks for closed-world vs. open-world assumptions, and efficiency concerns such as deployment cost, query runtime, and ease of use. - **Ending (~56%–100%)**: Concludes with a summary and future directions: holistic discovery systems, time/lake versioning, decentralized discovery, multimodal discovery, and agentic discovery. 【Key Takeaways】 - **Data discovery is search plus usability, not just retrieval** (Opening): Unlike classic IR, a discovery system must return datasets that can actually be joined, unioned, or used for augmentation—relevance alone is insufficient. - **The vanilla architecture is the organizing spine** (Opening–Middle): User tasks, table artifacts, discovery operations, dedicated indexes, and matching techniques form a reusable mental model for comparing systems. - **Operations form a layered toolkit** (Early): Keyword search, union search, and join search are basic primitives; sequences, domain-specific languages, and natural-language queries compose them into richer tasks. - **Indexes are where performance is won or lost** (Early–Middle): Value-based, column-based, correlation, row/table, and graph indexes each trade off recall, precision, and cost; MinHash/LSH is central for column similarity. - **Systems differ mainly in scope and assumptions** (Middle): Academic prototypes (AURUM, AUCTUS, BLEND) and industrial platforms illustrate how the same operations are packaged for different users and constraints. - **Evaluation is genuinely hard** (Late): Closed-world synthetic benchmarks have limits; open-world, utility-driven evaluation and efficiency metrics (deployment cost, runtime, ease of use) matter for real adoption. - **Distributed and constrained settings change the game** (Early–Middle): Privacy regulations, pricing, and federated environments force discovery to rely on metadata and summaries rather than full data access. - **The field is moving toward holistic, temporal, multimodal, and agentic discovery** (Ending): The authors flag these as promising directions rather than settled solutions. 【Reading Tips】 - **Read Chapter 1 carefully**—it supplies the vocabulary and architecture that every later chapter assumes; skimming it will make the rest feel like a list. - **Use the table of contents as a map**: operations (Ch. 3) → indexes (Ch. 4) → systems (Ch. 5) → benchmarks (Ch. 6) is a deliberate progression; follow it rather than jumping to systems first. - **Deep-read indexing and benchmarking if you build or evaluate systems**; skim the system survey if you mainly need concepts and terminology. - **Treat references as a reading list**: the book is explicitly a structured survey, so follow citations for algorithmic detail the main text only summarizes. - **Note the stated prerequisites**—database management background plus familiarity with ML/AI trends—and brush up if either is weak. 【Coverage Limits】 The available excerpts are dominated by front matter, table of contents, and Chapter 1, so this guide reflects the book's structure and stated aims more than the detailed content of later chapters; specific algorithms, experimental results, and chapter-level arguments beyond the outline are not covered here.
Page 4
believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, expressed or implied, with...
View in text
Page 6
. . 67 4.2 Value-Based Indexes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 4.2.1 Inverted Index ....
View in text
Page 8
. . . . . . . . . . . . . . . . . . . . . . . . . . 139 6.2.5 Evaluating UNION Discovery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14...
View in text
Page 11
ent through additional features, rows, or individual values. The formulation of the discovery task can either be the upload of a dataset whose relevant attri...
View in text
Page 13
.snowflake.com/ 3 https://www.databricks.com/ 4 https://aws.amazon.com/ 5 https://cloud.google.com/solutions/data-lakehouse 6 https://azure.microsoft.com/en-...
View in text
Page 16
round semantic interoperability and cross-portal harvesting. A key technical feature is that portal metadata is stored as Resource Description Framework (RDF...
View in text
Page 18
https://aws.amazon.com/clean-rooms/ 13 https://cloud.google.com/analytics-hub 14 https://www.infosum.com/ 15 https://scale.com/ 16 https://surgehq.ai/ 10 1 D...
View in text
Page 20
r the 20 https://huggingface.co/ 21 https://www.icpsr.umich.edu/sites/icpsr/home
View in text
Tags
AI categories
DataDatabaseBig Data
ISBN: 3032308224
Publisher: Springer
Publish Year: 2027
Language: English
Pages: 168
File Format: PDF
File Size: 6.2 MB
Text Preview (First 20 pages)
Registered users can read the full content for free

Register as a Gaohf Library member to read the complete e-book online for free and enjoy a better reading experience.

Generating text preview…