Data Accuracy Update: Better Data, Better Words

The Story Behind This Update

When we launched AdjectivesToDescribeAPerson.com, we made a promise: to help you find the perfect word, for the perfect description. But a promise is only as good as the data behind it. That’s why we’ve spent the last few weeks reimagining how we source, curate, and present adjective definitions and examples across our platform.

Today, I’m excited to share what we’ve changed—and why it matters for you.

What Changed: A Complete Data Overhaul

The Old Way vs. The New Way

Previously, our adjective definitions came from aggregated publicly available data sources on the internet, and while convenient, it didn’t give us the transparency or accountability we wanted.

Now, we’ve built a multi-layered data architecture that combines authoritative linguistic sources with human oversight and contextual relevance:

1. Definitions: Powered by WordNet and Wiktionary

We now source our primary definitions from two of the most respected linguistic databases in the world:

• WordNet (Princeton University) provides the primary definitions. WordNet is an open-source lexical database developed and maintained by Princeton’s Cognitive Science Laboratory. It’s widely used in academic and commercial NLP applications because of its rigorous semantic structure.

• Wiktionary serves as a backup for words where WordNet’s coverage is incomplete. Wiktionary is a collaborative dictionary powered by volunteers, offering cultural and contextual nuances that traditional databases sometimes miss.

Why this matters: Every definition you see now has a traceable source. No more black-box origins. Every definition you see on our platform displays its source clearly—whether it comes from WordNet or Wiktionary. This gives you full transparency about where we got it.

2. Examples: A Hybrid Approach Combining Three Sources

We now pull examples from three complementary sources:

Tatoeba (Real-World Usage): Tatoeba is a massive database of sentence translations with over 10 million sentences in 400+ languages. We scan Tatoeba’s English corpus using advanced NLP techniques to find real examples where adjectives are used to describe people.

The process is rigorous: our algorithm uses multiple filters to ensure the example sentence:

• Actually contains the adjective we’re studying
• Uses the word as an adjective (not as part of another phrase)
• Appears in a context about describing a person (e.g., mentions “person,” “people,” “teacher,” “friend,” etc.)

This guarantees you’re seeing how real people use these words in actual conversations and writing.

Wiktionary (Linguistic Context): Just as we use Wiktionary for definitions, it also provides curated example sentences that show formal usage patterns and idiomatic expressions. These examples often show adjectives in contexts that are less common but important to understand.

AI-Generated Examples (With Human Review): Here’s where our AI platform comes in. Since Adjectives to Describe a Person is fundamentally an AI-native tool, we also generate contextual examples using advanced language models. But—and this is crucial—every AI-generated example is manually reviewed by our team to ensure accuracy, naturalness, and relevance.

We generate examples by asking: “Show me this adjective used in a realistic scenario when describing a person.” The AI creates multiple variations, and then we hand-curate the best ones. This hybrid approach gives us examples that are both abundant and reliable.

A Concrete Example: Let’s Look at “Picky”

Here’s exactly how our new system works in practice:

Definition: “Exacting especially about details” (from WordNet) Source tag: WordNet

Examples:

See how each example has a source attribution? This transparency is new, and we think it’s important.

The Technical Foundation

For those curious about how we built this:

• WordNet Integration: We use the Natural NPM package, which wraps Princeton’s WordNet database. For each adjective, we query WordNet specifically for adjective definitions (parts-of-speech tags “a” or “s”).

• Wiktionary Scraping: We built a web scraper using Cheerio that fetches live data from Wiktionary’s English entries, parsing the HTML to extract adjective definitions, examples, synonyms, and more. When WordNet falls short, Wiktionary fills the gap. Our scraping follows Wikimedia Foundation’s Terms of Use by maintaining a clear User-Agent, respecting robots.txt, and ensuring reasonable request rates to avoid server burden.

• Tatoeba Processing: We downloaded Tatoeba’s CSV database (millions of sentences) and built a processor using the Compromise NLP library. For each adjective, it:

  1. Scans sentences for the word using regex
  2. Confirms it’s being used as an adjective (not as part of another phrase)
  3. Verifies the sentence is actually about describing a person
  4. Extracts the top 5 most relevant matches

• Pronunciation: We use the native Web Speech Synthesis API (window.speechSynthesis) to generate IPA pronunciations in British English (UK) and American English (US) in real-time.

Why We Did This: Data Compliance and User Trust

The core reason for this rebuild: We wanted to ensure that every piece of data on our platform is sourced from databases with clear, commercial-friendly licensing:

We’re also explicit about attribution: every definition and example now displays its source so you know exactly where information came from.

A Smaller But Important Detail: Five Words Got Upgraded

Five words in our collection previously used WordNet definitions that were incomplete or unclear. We’ve upgraded these to use Wiktionary’s more comprehensive definitions:

1. Absent-minded
2. Big-headed
3. Daring
4. Easy-going
5. Finicky

These words now have richer definitions and more contextual examples.

What This Means for You

Accuracy: Every definition is sourced from recognized linguistic authorities. No more guessing about whether we got it right.

Context: Examples come from real usage (Tatoeba), academic sources (Wiktionary), or thoughtfully generated scenarios (AI with human review). You’ll see how words are actually used.

Transparency: You’ll always know where information comes from. Each definition and example clearly shows its source.

Reliability: Our data sources are all commercially licensed and clearly attributed, so you can confidently share and use our recommendations without legal concerns.

Thank You

Building a trustworthy language tool requires care, transparency, and collaboration. This update reflects our commitment to both. We hope it makes AdjectivesToDescribeAPerson.com even more useful in your writing, interviewing, and learning.

Questions about our data sources? Feel free to reach out on our contact page.