EasySearch Rules Syntax: Quick Reference Guide

Rule Structure

Each rule consists of an expression and an optional description. When importing via the _import API, only the expression and description need to be supplied; the system automatically assigns an offset.

{
  "expression": "北京 or 京城",
  "description": "地域_北京"
}

Logical Operators

and

(alias &) – both conditions must be satisfied. or (alias |) – either condition may be satisfied. not (binary, aliases - and &!) – exclude matches.

Operator precedence (from highest to lowest): not (unary) > near > and > or. Parentheses () can be used to override precedence.

(source(新华网) or source(人民网)) and title(北京 or 上海)

Field Constraints

field(expr)

– substring match anywhere in the field (e.g., title(北京)). field[expr] – exact match of the whole field (e.g., sentiment[正面]).

Built‑in fields such as title and content need no declaration. Custom fields must be declared as numeric types at compile time.

POST /_match_rules/repo/_compile
{
  "fields": ["score", "heat"]
}

Regular Expressions

Simple regex: {{regex}} (e.g., {{1[3-9]\d{9}}} matches a mobile number).

Field‑specific regex: field("regex") (e.g., author("张.*")).

Composite regex logic: [expr and/not ...] (e.g., [author("张.*") and not author("张三丰")]).

⚠️ Regex inside a field must be quoted; author(/张.*/) is invalid.

Distance Matching (near)

Syntax:

A near/n B      # distance ≤ n, order irrelevant
A near/+n B     # A appears before B, distance ≤ n
A near/-n B     # A appears after B, distance ≤ n

Distance unit: one Chinese character, one English word, or one numeric string equals 1.

Difference from and:

# and – both words appear anywhere, possibly unrelated
北京 and 发布会

# near – words must be within a limited span, ensuring context
北京 near/10 发布会

Chained example:

(北京 or 京城) near/5 (经济 or GDP)

Numeric Range Matching

score(80, 100)   # positive rating range
score(0, 40)     # negative rating range
heat(10000, 99999)

Numeric fields must be declared as numeric types during compilation.

Three Common Tagging Scenarios

Region Tags

{
  "expression": "北京 or 京城 or 首都 or 中关村",
  "description": "地域_北京"
}
{
  "expression": "深圳 or 鹏城 or 南山区 or 前海",
  "description": "地域_深圳"
}

Sentiment Tags

{
  "expression": "非常好 or 极佳 or 强烈推荐",
  "description": "正面_强烈好评"
}
{
  "expression": "一般 or 还行 or 中规中矩",
  "description": "中性_平淡评价"
}
{
  "expression": "严重不满 or 投诉 or 退款 or 维权",
  "description": "负面_强烈差评"
}

Person Name Tags

{
  "expression": "author(\"张.*\") and source(清华大学 or 北京大学)",
  "description": "人名_高校学者姓张"
}
{
  "expression": "title(\"王.*\") and content(企业家 or CEO or 创始人)",
  "description": "人名_企业家姓王"
}

Common Errors Quick Check

Minus sign is parsed as NOT. Use quotes: "2024-2025年报" instead of 2024-2025年报.

Regex without quotes : write author("张.*") not author(/张.*/).

Numeric field not declared : include the field list in the compile request, e.g., { "fields": ["score"] }.

Separator should be a Tab between expression and description, not a space.

Special Character Escaping Quick Check

Characters that appear as literal values must be quoted: -, &, |, ( ), [ ], { } and spaces.

Rule of thumb: any character that may cause ambiguity should be wrapped in quotes.

Rule Writing Recommendations

Start simple, then expand : begin with pure keywords or OR, then gradually add field constraints and near.

Prioritize title : title(关键词) yields far higher precision than full‑text matching.

Prefer near over and : in the same semantic scenario, near significantly reduces false positives.

Use NOT to filter noise : for sentiment tags, exclude speculative contexts, e.g., and not (预测 or 如果 or 假设).

Regex for fuzzy matches : apply patterns like author("张.*") for names, models, etc.