Getting the most out of AI capabilities in your news monitoring
Scrolling through and sieving thousands of search results are a thing of the past. ScoutAsia has injected an AI-powered technology, Scout AI, that speeds up your search and monitoring for news.
With 30+ pre-trained Scout AI topics, ranging from market, macroeconomy, business trends, corporate events to covid-19 related topics, Scout AI reads thousands of news articles, picks up relevant content on respective topics, and keeps you updated with the latest movements.
Apart from the pre-trained Scout AI topics within the subscription package, users can choose to customise their Scout AI according to their requirements.
We have invited our Machine Learning Expert Team to share more about Scout AI and their behind-the-scenes tasks in creating customised Scout AI.
Q1. Tell us what is unique about our Scout AIs?
In a nutshell, Scout AIs is an easy no-code customisation process for users & human-machine learning support is our speciality.
Scout AI topics are customised according to the needs and interests of each user. The team must first understand the search intent to ensure relevant news is delivered through our alerts feature.
To graduate out of the training process, the team monitors two primary quality metrics:
- Predictions are of good quality
- The extent to which the scout is able to predict relevant news
Upon satisfaction, the team will then release the model to the respective client. From time to time, the team monitors the performance of the Scout AI after the release. There may be additional training to ensure that the model performs well and meets the mentioned quality metrics.
Q2. Given that keyword-based search is quite common, what are the benefits of AI assistance?
Keyword-based search has been effective, but there are limitations to it. A keyword-based search pulls out content that mentions the specific keyword. However, there are instances whereby the overall content is not relevant to the particular keyword.
By implementing AI assistance into keyword-based search, Scout AI is able to pick up the main content of the article instead of limiting search results to keywords mentioned, leading to higher precision in search results.
Q3. What kind of topics goes well with Scout AI? Can you give some quick examples?
Generally, subjects that can be reported or described with similar wordings and themes are good candidates for Scout AI. Events such as protests, natural disasters or topics like climate change, autonomous driving are some excellent examples.
Articles relating to “Protests” can be reported without mentioning the keyword itself. Such keyword examples are “strike”, “rally”, “activist”, “conflict”. These will help the machine understand the context.
On the other hand, “natural disasters” covers many different types of events from “wildfire” to “earthquake” and even “tsunami”.
To follow “climate change” related news, you might wish to pick articles about global challenges such as global warming, carbon dioxide, sustainability, greenhouse effect, flooding, etc.
Similarly, on the topic of “autonomous driving”, we would expect to see articles with common words such as self-driving, technologies used in vehicles, smart-driving, driverless.
For quality training purposes, the selected topic should be covered across a few news sources, not just mere mention of keywords but primarily focusing and talking about the said topic. Quality articles will allow the model to comprehensively understand the specific topic, resulting in higher quality and accurate trigger results.
Although providing more explicit content and context is essential, too specific or uncommon topics may not suit Scout AI. The reason is that there may not be sufficient articles to train our machine model, and also, the Scout AI may not be sufficiently precise.
Q4. With ScoutAsia web app subscription, users can create their custom topics. Tell us about the flow of creating custom AI topics?
To get things started, we would get some information from the customer as below:
- Name of topic
- Purpose of creating the Scout AI
- Examples of relevant and irrelevant articles (shown below)
Examples of relevant and irrelevant articles
The information provided will allow the team to understand the requirements better and contribute productively to a smoother research and training process.
We then take a few days to do some initial research about the topic and provide feedback to the user about our recommendations to materialise the topic. Upon confirmation, the team requires up to five working days to deliver the agreed customisation request.
If an overlapping topic is found in our existing standard Scout AI, the team will accordingly design the alert package.
That’s all we have for this round of conversation! Thank you for taking time off to share your daily behind-the-scenes tasks.
Content and context play a part in delivering quality and accurate search results. The clearer your goals and purpose for our Scout AI, the more beneficial it is for your business success.