Welcome to Custom Machine Learning
In KPI6 it is always time for news: a disruptive feature is coming, which will improve the performance of our platform and those of our clients. Custom Machine Learning, an Auto-ML system to offer the customer a new method to filter and enrich the posts of research and monitor an entire industry, a single brand, or enhance use case related to customer understanding.
Would you like to know how a Digital Consumer Intelligence approach can help your Business? Book a Demo to get a free hands-on demonstration of our platform.
Despite the passing of the years and the increase of technologies, the “old-school thinkers” will always remain stable on their ideas: the increase in AI functionality and Machine Learning use will lead to replacing humans with robotic or computer-generated alternatives.
The truth is that one of the primary uses of Artificial Intelligence is to help man and his work, eliminate the “garbage” on the web and help people and computers to identify the best content and the best answers to their questions.
The application of strategies that involve the use of Machine Learning techniques within your company is important: we have seen how the use of a text classifier is necessary for your analysis and monitoring, as well as the use of AI techniques in Social Media Intelligence.
At KPI6, we like to continuously improve our platform, our performance, offering customers more and more precise and customizable services to their needs.
And it is from this frenzy (and from the hard work of our machine learning engineers) that a new feature is born: Custom Machine Learning.
Let’s imagine having large research (volumes of conversations) on a particular topic and having to analyze and classify according to certain themes: the CML gives the possibility of being able to automatically categorize the posts according to labels (categories) chosen by the user.
A brilliant feature for the platform, as well as the uniqueness of having automated Google Bert.
Google Bert Technologies in Digital Consumer Intelligence
BERT, an acronym that stands for Bidirectional Encoder Representations from Transformers, is the Google logarithm based on the use of neural networks, useful for better processing of natural human language and aims to better understand the complete context of the queries and texts provided from the search engine (or from other sources) with even more defined precise and careful understanding thresholds.
This update uses a technique called Natural Language Processing, and Artificial Intelligence field that gives machines the ability to read, understand and draw meaning from human languages. It uses artificial intelligence to understand what people are looking for and how / how much the content actually matches their needs. Google BERT will help the search engine understand what words mean in a sentence, but allowing it to understand all the nuances of this in the semantic context.
Small queries modifiers in regular sentences, such as “not”, “a” or “no”, may have been previously “misunderstood” by Google’s algorithms and by text classifiers. Queries (and individual keywords) will be evaluated in much more detail, taking into account nuances that may have previously been ignored. A technology that Google has released open, which we have customized and inserted into our servers.
CML is arriving in KPI6
Andrea Salvoni, Chief Research Officer in KPI6, explain to us the potential of this technology:
“Custom Machine Learning is an AutoML system. It allows you to train, optimize, and bring predictive models into production in an easy and guided way, providing the system with only tagged examples.”
A feature that will bring various advantages:
“CML was created to offer the customer a new method to filter and enrich the posts of research. By providing examples accompanied by a tag (or class), we can create custom filters that are more powerful than simple text filters, and relatively simpler to create. Instead of creating complex Boolean queries, simply provide the system with examples of posts accompanied by a tag that identifies an attribute of that post. The system will then create a filter that is capable of recognizing those attributes that have been seen in the examples, even in new posts that it has never seen. The system is based on neural networks, so it will be able to understand the meaning based on the syntax and the semantics of the post, a thing that a text filter cannot do easily.”
Important changes, which will also partly redefine the application’s operating interface:
“It will be added a new section to the platform, where it will be possible to tag the examples and train the models. Once trained, it will be possible to use them to enrich and filter the research posts. We will, therefore, have new visualization tools to add inside dashboards.”
Through the use of Custom Machine Learning, it is possible to monitor an entire industry, a single brand, or enhance use cases such as:
– Crisis Management: detecting the posts that could damage your business;
– Brand Reputation: by categorizing customer feedback relating to a product, it is possible to analyze it in a more structured way;
– Customer Care: Understanding customer requests in such a way as to act in a specific way, classifying the type of request from the text.
These are not the only ones, the use cases are manifold, which they will implement only thanks to the continuous use and training of the platform. An important new feature for KPI6, the result of many months of hard work by our machine learning engineers, which will be released shortly on our platform with all its potentialities.
Would you like to know how a Digital Consumer Intelligence approach can help your Business?
Book a Demo to get a free hands-on demonstration of our platform.