First-hand experience in creating Custom AI Topic

Digesting the nuggets of wisdom from our Machine Learning Expert Team, here is my process of creating Custom AI Topics. If you have yet read our last casual conversation with them, you may wish to read it here.  

With AI assistance, searching and monitoring for news don’t seem like a hassle anymore. A key process to make a good Custom AI topic is to let machines understand your search intent like the content and the context. It will determine the quality of the result.  

I would demonstrate the creation process of creating custom AI topic to share my first-hand experience. I will be creating a Custom AI topic that curates news articles relating to Natural Disasters for this case study.  

1. Defining what I would search   

How do we define Natural disasters? Are we looking at a specific country or a region? To what extent do we need to know about natural disasters? For this case study, I have laid out the necessary parameters relating to Natural Disasters. 

Creation of Custom AI Topic. 

2. Training the machine  

Once I have defined my parameters, the machine will crawl a set of articles relating to the keyword. To train the machine, I will be identifying both relevant and irrelevant articles. A tick indicates relevant articles, and a cross indicates content unrelated to my search intent.  

Example of training the Custom AI Topic 

Example of relevant training article 

Example of irrelevant training article 

The above images are some examples of relevant and irrelevant training articles. Although the irrelevant training article titled “2 weather systems continue to bring rains over parts of PH” did mention keywords relating to natural disasters, the article’s context focuses on the weather forecast. This context may not be fully relevant to my search intent and parameters.   

 Once I have met the minimum requirement of 10 ticks and 10 crosses, I will receive a notification indicating completing Custom AI training. Of course, the more time invested in the machine, the higher the quality of the news search results. 

Completion of training

That’s all I need to do from the user perspective. The Machine Learning Expert Team will then take over to do some initial research and provide feedback about their recommendations. With that, the team will continue fine-tuning your Custom AI with additional training. This process will take up to five working days before materialising them.   

 From the user perspective, there was no complication or required me to learn how to code. I would set the difficulty level to easy. Feel free to explore this feature and share your experience. 

 If you are an existing user, login into ScoutAsia App to explore our Custom AI feature. For enquiries, please contact us at [email protected] or contact your account manager today.