Extracting query intent without relying on AI or Machine Learning requires a rule-based or linguistic approach. This involves analyzing the syntax, keywords, and structural markers within the text to categorize what the user wants.

Here is the general framework for building a non-AI intent extraction engine:

1. Keyword & Seed Word Mapping

The most straightforward method is to create a “Lexicon of Intent.” You categorize specific verbs or nouns that are highly correlated with certain goals.

Intent Category	Primary Keywords (Seeds)
Informational	what, why, how, history, meaning, definition
Transactional	buy, price, cheap, discount, order, download
Navigational	login, official site, homepage, contact page
Instructional	how to, steps, guide, tutorial, recipe, fix
Commercial	best, top, review, vs, comparison

2. Syntactic Pattern Matching

Instead of looking at words in isolation, you look at the structure of the sentence. This is often done using “Regular Expressions” (Regex).

Question Patterns: If a string starts with “Who,” “Where,” or “When” and ends with a “?”, it is flagged as Informational.
Action Patterns: If a string starts with an imperative verb (e.g., “Build,” “Run,” “Create”), it is flagged as Instructional.
Comparison Patterns: Searching for the ” {Noun} vs {Noun} ” structure usually indicates Commercial Investigation.

3. Part-of-Speech (POS) Tagging

By using a standard dictionary-based POS tagger (which identifies nouns, verbs, and adjectives based on a fixed database), you can weigh the intent:

High Verb Density: Usually indicates a task or Instructional intent.
High Proper Noun Density: Usually indicates a Navigational intent (searching for a specific brand or person).
Adjective + Noun: (e.g., “fastest car”, “cheapest flight”) indicates Commercial intent.

4. Logical Boolean Filters

You can apply the logic shown in your prompt snippet to filter results based on the metadata of the source or the structure of the query:

The “IF-THEN” Logic: * IF query contains “How to” AND length > 3 words THEN Set Intent = “Instructional”.
- IF query contains a URL or Domain name THEN Set Intent = “Navigational”.
Negative Constraints: * IF query contains “free” or “wiki”, DEMOTE Transactional weight.

5. Decision Tree Architecture

To combine these, you build a hard-coded decision tree. The system passes the text through a series of “gates”:

Gate 1 (URL Check): Does it look like a web address? (Yes → Navigational).
Gate 2 (Verb Check): Does it start with a command? (Yes → Instructional).
Gate 3 (Entity Check): Does it match a known product list? (Yes → Transactional).
Gate 4 (Default): If no markers are found, categorize as General/Informational.

Summary Table: Non-AI Extraction Methods

Method	Tool Used	Best For
Lookup Tables	Hash maps / Dictionaries	Detecting specific topics or brands.
Regular Expressions	Regex patterns	Identifying question formats or price symbols.
Boolean Logic	If/Else statements	Filtering high-stakes (YMYL) vs. lifestyle content.
Heuristics	Rule-of-thumb logic	Identifying intent based on query length or punctuation.

Are you looking to implement this logic in a specific programming language, or are you designing the logical flow for a database?