> ## Documentation Index
> Fetch the complete documentation index at: https://labs.laer.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Email threading

Email threading groups related emails into coherent chains based on their metadata and content. This section explains the criteria and methods used by the platform to determine email threading:

### **Email field extraction**

The following metadata fields are extracted for the primary message (topmost email header) and the embedded messages (inline headers).

| Field Name | Description         |
| ---------- | ------------------- |
| FROM       | Sender of the email |
| TO         | Email To field      |
| CC         | Email Cc field      |
| BCC        | Email Bcc field     |
| DATE\_SENT | Email sent date     |
| SUBJECT    | Email subject       |
| BODY       | Email body          |
| SIGNATURE  | Email signature     |
| FOOTER     | Email footer        |

### **Exact Matching Criteria**

Epiq AI performs textual matching of the following email fields. Exact textual match ignores whitespaces and punctuations. 

#### Email sender/recipient matching

* Exact match of the `FROM`, `TO`, and `CC` fields extracted from messages

#### Email sent date matching

* Epiq AI normalizes all time stamps in UTC and compares the time differences. If the time zone is available, maximum allowance is 1 hour. If time zone is unknown, maximum allowance is 24 hours.

#### Email subject matching

* A normalized subject must satisfy exact textual match. Email subjects are normalized by removing prefixes like “RE:”, “FW:”, and variations of these, to identify consistent subjects across threads.

#### **Email Body Matching**

* Textual matching of the Body content
* Inline quoted text, when applicable
* In cases where exact matches are not satisfied, a shingling method is used, which breaks down the comparing texts into overlapping sequences of words (unigrams and bigrams). If the comparing texts exceed the threshold proportion of shared shingles (default value is `0.99`) then they are considered a match.

#### **Features Ignored**

* `BCC` field is ignored in the email matching.
* Footers, common disclaimers, and email signatures are ignored.