{"id":90,"date":"2026-11-07T01:54:15","date_gmt":"2026-11-07T01:54:15","guid":{"rendered":"https:\/\/www.ecsow.com\/blog\/?p=90"},"modified":"2026-11-07T02:34:36","modified_gmt":"2026-11-07T02:34:36","slug":"customize-rules-teaching-ai-what-to-detect","status":"publish","type":"post","link":"https:\/\/www.ecsow.com\/blog\/customize-rules-teaching-ai-what-to-detect\/","title":{"rendered":"Customizing rules.tsv: Teaching Ecsow’s AI What to Detect"},"content":{"rendered":"
\n
The short version<\/div>\n
The AI hears words. A file called rules.tsv decides what those words mean. It sits in the Ecsow install folder, one rule per line, three columns: label, weight, phrase. Separated by Tab, saved as UTF-8, and loaded only at startup \u2014 so a restart is part of every edit. A score of 4 is the bar. Get that far and the call is classified.<\/div>\n<\/div>\n

Ecsow’s AI detection does two separate jobs. First it turns speech into text. Then it decides what the text means: a person, a voicemail, a screening assistant, a transfer message, or a phone menu.<\/p>\n

The second job is a plain text file. You can edit it with Notepad. That is what this page is about.<\/p>\n

\n
Reference document:<\/strong> the full rules.tsv guide, with every label, weight and worked example. Download the PDF<\/a> and keep it next to the file while you edit.<\/div>\n<\/div>\n

Where the file lives<\/h2>\n

rules.tsv<\/code> sits in the folder where you installed Ecsow. Open that folder and you will find it next to the program.<\/p>\n

There is a second place it can be set: the Rules file<\/strong> field in Tools > Options > Voice Activity Detection > AI Settings<\/strong>. If that field points somewhere else, your edits to the install-folder copy do nothing. Check it once and you will not lose an afternoon to it.<\/p>\n

Three columns, five labels<\/h2>\n

Each line is one rule. Three columns, separated by a single Tab character:<\/p>\n

S\t4\twho is calling\nH\t4\tthis is he\nVM\t5\tleave a message after the tone\nT\t5\tplease hold while i connect you\nIVR\t5\tpress 1<\/pre>\n
Read the first line as: label S<\/strong>, weight 4, phrase “who is calling”.<\/p>\n
\n
Label<\/span>What it means<\/span>What happens next<\/span><\/div>\n
H<\/span>A live person<\/span>Your agent is connected<\/span><\/div>\n
VM<\/span>Voicemail or answering machine<\/span>It leaves your message<\/span><\/div>\n
S<\/span>Screening assistant, like Google’s<\/span>It plays your short reply<\/span><\/div>\n
T<\/span>“Please hold while I transfer you”<\/span>It waits for the transfer<\/span><\/div>\n
IVR<\/span>A keypad menu: “press 1 for sales”<\/span>It hangs up<\/span><\/div>\n<\/div>\n
Those five letters are the whole vocabulary. Everything else in the file describes how strongly a phrase points at one of them.<\/p>\n
Weights, and the number 4<\/h2>\n
The weight says how much you trust a phrase. Each match adds its weight to that label’s score. The call is only classified once a score reaches 4.<\/p>\n
So one weight-5 phrase is enough on its own. Two weight-2 phrases also get there. A single weight-2 phrase does not, and that is on purpose \u2014 it stops one loose word from deciding a call.<\/p>\n
Keep weights between 1 and 5:<\/p>\n