liveBook · Manning

Data annotation

As explained in Hidden Influences

The process of assigning descriptive or evaluative labels to content so that it can be identified, classified, or analyzed.

A recommendation system can sort books by labels such as “travel literature” or “fiction.” Data annotation is the process of assigning such labels so content can be identified, classified, or analyzed. Recommendations are only as specific and accurate as the labels allow. section 2.3.1

Assigning high-quality labels takes time and money, and different people may classify the same item differently. Fine-grained categories raise further questions: should a book about Italy be labeled by region, genre, country, or all three? These problems become harder when content grows quickly or arrives with few labels. section 2.3.1

Labels also shape attempts to measure whether a social media feed amplifies one side of a public debate. Finding posts that support or oppose a policy is difficult at scale, while labeling users by political stance raises privacy and data-governance concerns. Labels can simplify nuanced content or sensitive identities, so measurement depends on choices that may themselves be difficult or risky. section 4.7 section 4.7.2

How Hidden Influences covers it

How deep it goes
Explains labeling through examples and discusses cost, consistency, scale, nuance, and privacy concerns.
Examples
Uses book genres and regions, plus social posts supporting or opposing a policy and users’ political stances.
What you can do after
Evaluate how labeling choices and constraints affect recommendation quality and amplification measurement.
How it connects
Connects data annotation to recommender systems, content-based recommendation, and algorithmic amplification.

Also explained in

Compare how 14 books explain data annotation →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Hidden Influences. The section links point to the parts of the book it draws on.