◆ AI Learning · Situational

How can a system learn to ignore bad phrases or words?

A system needs a way to remove unwanted information. A user can mark bad phrases or words. This stops the system from processing them in the future.

By Pawan Nayar · from the LLOS AI Learning record

The rule

A system needs a way to remove unwanted information.

A user can mark bad phrases or words. This stops the system from processing them in the future.

Sometimes a computer program finds words that are not useful. These can be parts of a website like "submit" or "close". A special tool helps a user remove these words. The user can click on a word or phrase. A small window appears. The user can choose to ban the whole phrase. The user can also ban only certain words within the phrase. When a word or phrase is banned, it disappears from the screen right away. The system remembers these banned items. It sends this list to the main computer program. This program then ignores these words in the future.

An AI model is reading text from a webpage. It picks up phrases like "click here to accept cookies". A developer sees this and knows "cookies" is not important for the AI's task. The developer uses the tool to ban the word "cookies". Now, the AI model will ignore "cookies" in all future texts.

Words you'll need

Resolved from this lesson's own record — real definitions, not invented for this page.

Unsupervised Learning

Learning patterns without labeled data

Supervised Learning

Learning from labeled data to make predictions

Reinforcement Learning

Learning from feedback signals over trial-and-error interactions

Feature Engineering

Creating inputs for AI models from raw data

Questions people ask

More from the same record, pulled by shared topic — not written new for this page.

Why does an AI model still create near-duplicate text even after case-insensitive matching is applied?

Case-insensitive matching fixes only one type of duplicate. It finds 'Coffee' and 'coffee' as the same.

How can an AI model avoid using copyrighted brand slogans?

A developer must check all slogans in the AI model's training data. This check happens in two steps.

How does the phrase highlighter format work for an AI model?

The phrase highlighter format is an advanced interaction type. It lets a student select a group of words.

How can an AI model improve the sound quality of its generated text?

An AI model can use a three-level sound architecture. Level 1 offers simple controls to the user.