Four technologies to change e-commerce
NLP (neuro-linguistic programming), computer vision, moderation and security, personalization—these technologies can certainly become an advantage for online projects in the nearest future.
Bot-sellers, robotized online chats, automated logistics—this is a reality now because of the development of modern technologies. Amazon created the first store without cash desks and registered a patent for delivering goods by drones with parachutes. ClarifAl and Google can define and describe pictures. Ukrainian Ekol Logistics uses neural networks for transporting cargos and Lalafo marketplace for automatically describing goods. Neural networks can define and describe what is shown on a picture, generate neuro-linguistic programming, translate texts, and understand speech. These technologies will soon become a great help for online projects.
There are four most important parts of e-commerce transformation: NLP (neuro-linguistic programming), computer vision, moderation and security, personalization. Let us take a close look at each of them.
NLP (neuro-linguistic programming)
NLP’s features were demonstrated by a New York startup back in 2008. The project team designed an extension for Firefox which analyzed customers’ feedback and automatically generated a short description of two sentences. This technology was used by Amazon and let people skip a tiring process of reading lots of comments. The project was shut down in 2009 because of financial crisis but its technology will play a significant role in e-commerce development.
In the future NLP will search for items with a voice command and ask customer questions to gain clarity: is the clearance of a car important for you, which objects do you take pictures of most often, and others. It is quite possible that artificial intelligence will be able to differentiate between customers’ feelings: excitement, uncertainty, etc. and offer them goods depending on these factors.
Google, Facebook, Pinterest use neural networks when working with images. That is how, because of “face detection” function, Facebook offers you to tag yourself and your friends on a picture and Google offers “search for same pictures” function.
The problem is that most modern services define ordinary and common things: clouds, houses, plants. In e-commerce, it is needed to look at a picture from a sales manager’s point of view and describe goods with characteristics that are important for a customer. Nowadays there are many batch solutions that let online stores sort goods by pictures but most of them are still developing.
Some projects teach neural networks themselves. Lalafo, a Ukrainian buying and selling second-hand goods service, taught neural networks to describe a product on the picture, add category instead of the seller, and suggest the price, based on information about same products.
Imagine how computer vision will change online-trading: it will be enough for a customer just to upload a picture from his smartphone to find a similar product or even line of products.
Personalization and simplicity of usage are the most important things nowadays. NLP, deep learning, defining forms are directed at using knowledge about customers more wisely. If earlier online-stores could create a simple algorithm for product recommendation, today, with the growth of mobile traffic, this task became harder. In mobile apps, users rarely search for products by categories, they just scroll the feed. Usually this takes 5–10 minutes. Within this time it is essential to guess user’s preferences and show him products to grab his attention. Moreover in mobile apps half of the sales are happening through recommended deals, not through categories or filters. Technologies of neural networks can make recommendations that are as close to user’s interests as possible. Network can analyze user’s browsing history, define topics that he finds interesting, and suggest products depending on these factors.
Imagine a user, opening an app and starting to scroll feed on his smartphone. He looks at a thermos and wants more information about it. Together with a thermos, the neural network will show him a teapot and if, according to previous browsing history, neural network decided that the user is a girl it will also show a pair of shoes because 90% of girls will definitely react to shoes. After some time the neural network will differentiate between types of girls: active, sporty, businesswoman, and so on and will suggest goods based on this information.
Moderation and security
Online trading platforms have to deal with a great amount of information and usually hundreds of moderators are involved in this process. Neural network helps to automate the process of checking the content and reduce the amount of people involved, to one. Moreover, if you use neural networks you can influence the quality of moderation. People tend to make mistakes, machines do it less often. When analyzing users’ profiles and finding out different patterns, neural networks can even find violators. Disqus, Pinterest, Google Photos, and others are already using classification and filtration of users’ content with the help of neural networks.
Many of these solutions are still only developing but very soon each of them can become an undeniable competitive plus for e-commerce businesses.