What is Query Breadth?

Query breadth is a measure of how general or narrow a search query is. Broad queries typically:

• Return a high number of results.
• Often consist of fewer search terms or are used very frequently.

📌 Key Concepts

• 🔹 Related Document Count

  📖 What it means: The number of documents matching a query.

  ✅ High Breadth: Many results
  Example: "shoes" (broad term).

  ✅ Low Breadth: Few results
  Example: "red leather boots size 8" (highly specific).

• 🔹 Adjacent Words

  📖 What it means: Phrases or word pairs that appear together.

  ✅ Example:
  "New York" vs. "York New" (Incorrect order).
  Query: "best Italian restaurant" → Narrower breadth than "best restaurant."

• 🔹 Unknown Entities

  📖 What it means: Terms that aren’t recognized by the system (e.g., slang, misspellings).

  ✅ Example:
  Query: "How to fix a zommbie phone?"
  "Zommbie" is flagged as an unknown entity.

• 🔹 Adding more words = Narrower query.

  ✅ Example:
  Broad: "apple" → Could mean the fruit, company, or movie.
  Narrow: "Apple iPhone 15 release date" → Specific intent.

📌 Applications

• 🔹 Named Entity Recognition (NER)

  📖 What it does: Identifies specific names.

  ✅ Example: "Elon Musk" → Recognized as a named entity.

• 🔹 Triple Creation

  📖 What it does: Builds structured data in a subject-predicate-object format.

  ✅ Example:
  [Subject: Paris] → [Predicate: capital of] → [Object: France]

📌 Process Flow

• 🔹 1️⃣ Retrieve Query/Clicked Result Pair (402)

  📖 What happens? Tracks which results users click for a query.

  ✅ Example:
  Query: "healthy recipes"
  User clicks: A blog about salads.

• 🔹 2️⃣ Determine Popularity Measure (404)

  📖 What happens? Calculates how frequently a result is clicked for that query.

• 🔹 3️⃣ Determine Query Breadth (406)

  📖 What happens? Checks if the query is broad or specific.

  ✅ Example:
  Broad: "coffee shops"
  Specific: "Starbucks near me"

• 🔹 4️⃣ De-weight Popularity Measure (408)

  📖 What happens? Reduces the bias of overly popular results for broad queries.

  ✅ Example:
  Query: "shoes" has millions of results.
  Fix: Prevent one generic page from dominating.

• 🔹 5️⃣ Store Adjusted Popularity (410)

  📖 What happens? Saves the updated weights to improve future search accuracy.