More Like This

Find documents and pages with similar text content using Elasticsearch's More Like This algorithm.

Indexing with term vectors

According to Elasticsearch documentation, indexing the content field with term vectors enabled (which is the default setting) helps to get better results. But if index storage size is a serious concern and term vectors are disabled (OPENALEPH_SEARCH_CONTENT_TERM_VECTORS=0) "more like this" still works.

How it works

More Like This analyzes text fields to find documents with similar content:

Extracts important terms from source document
Generates query using selected terms
Ranks results by similarity
Returns similar documents and pages

Target schemas

Searches only document entities.

Other entity types (Person, Company, etc.) are excluded.

Fields analyzed

content - Primary text content
text - Secondary text content
name - Entity names (to spot similar file names)

Parameters

`mlt_min_doc_freq`

Minimum document frequency corpus-wide.

Type: int
Default: 0

--args "mlt_min_doc_freq=2"

`mlt_min_term_freq`

Minimum term frequency within source document.

Type: int
Default: 1

--args "mlt_min_term_freq=2"

`mlt_max_query_terms`

Maximum terms to use in query.

Type: int
Default: 25

--args "mlt_max_query_terms=50"

`mlt_minimum_should_match`

Percentage of query terms that must match.

Type: str
Default: 10%

--args "mlt_minimum_should_match=25%"

Parameter effects

`min_term_freq`

Higher values focus on more important terms:

Higher = considers only frequently mentioned terms
Lower = includes more terms from source

`min_doc_freq`

Controls term commonality:

Higher = focuses on common terms
Lower = includes rare/specific terms

`minimum_should_match`

Controls similarity strictness:

Higher percentages = stricter similarity
Lower percentages = broader similarity

`max_query_terms`

Affects query comprehensiveness:

Higher = more comprehensive matching
Lower = focus on most important terms

Query structure

{
  "more_like_this": {
    "fields": ["content", "text", "name"],
    "like": [{"_id": "doc-123"}],
    "min_term_freq": 1,
    "max_query_terms": 25,
    "min_doc_freq": 0,
    "minimum_should_match": "10%"
  }
}

Source entity is automatically excluded from results.