In simple terms, text classification applies machine learning algorithms to understand the meaning of a text. Once the meaning is identified, the algorithm puts a label on the text to classify it. For example, websites like Amazon use text classification to classify reviews, which can be positive or negative. The same technique can be used to perform sentiment analysis on social media posts, and determine people’s attitudes towards a particular topic or product.
At Visma Connect, we have developed our own system for support ticket classification using this methodology. We receive 100 support tickets every day on average. As you can imagine, it takes quite a long time to review and route all of these requests. Text classification allows us to sort through incoming requests faster, and assign them to the right person.
Text classification works differently based on the technique. A simple use case would be training the algorithm to look for the frequency of a word on a text. For instance, if the word “meeting” shows up on an email several times, the system could recognize that the message is about scheduling a meeting and therefore suggest a reply (something you’ve probably seen on Gmail).
The more text the algorithm is fed, the more it learns - just like a child learning a new language. This means the algorithm gets better at classifying text over time. Now, this sounds like a pretty cool thing, but what does it do beyond emulating human behaviour?
Let’s go back to the Amazon example. Every day, hundreds (if not thousands) of products get reviews on Amazon. Imagine having to look through them manually to find out if they are good or bad! It would be repetitive and extremely time-consuming. With text classification, this process can be automated, saving Amazon considerable time and personnel costs. That’s not all. The algorithm gets more accurate than humans eventually and makes less mistakes as time passes, so it can be applied and scaled to address a variety of problems. For instance, hospitals are using this machine learning technique to look through patient records and suggest treatments in an effort to prevent death or cure patients more quickly.
Text classification and machine learning/AI are very fast-moving fields. Developments are happening by the day and machines are getting just as good as humans at classifying text. You might be surprised to find out how many of the software products you are using already leverage this technology.
Everyone with an online presence has probably used an app that leverages text classification in one way or another. Text classification is a subset of Natural Language Processing and is typically used in AI applications - whether it’s classifying reviews on e-commerce websites or performing sentiment analysis on social media. Gmail users, for instance, now get reply suggestions based on email content. This feature relies on text classification to dissect and understand the content of a message, and therefore to provide accurate recommendations. During the webcast, we will cover:
Tuesday, November 24 at 16: 00 hrs.
This webinar is a part of the Discover Visma Week. A week full of virtual events where you will get to opportunity to attend presentations and webinars of all the Visma Netherlands companies and our unique cloud solutions. Be inspired, meet the various Visma companies and discover how we are your partner in progress.