Dialog agent for the shopping portal
Yosh.AI implements automatic conversational systems that simulate conversation with people by interacting in real-time via a proprietary application. Our artificial intelligence tool can learn how to respond to queries and conduct conversations using deep learning algorithms. The primary purpose of chatbots is to provide fast and effective customer service, whose expectations and requirements are growing along with the development of technology. As a result, chatbots enable faster and more effective communication with customers and improve internal processes in companies. Yosh.AI implements chatbots in various business sectors, from the financial industry to e-commerce. Below are presented examples of the functionality of the implemented application for e-commerce.
Dialogue agent welcome window
The welcome window displays a message to the user (including the data processing policy) and query hints in the form of buttons. The user can get to know the functionality of the agent by asking. In response, he will receive a carousel with possible functionalities, such as searching for products in the customer dataset, information about products and promotions, searching for similar products by image, searching for the nearest stationary stores based on the user’s location, displaying hints of the most frequently asked questions in the form of buttons.
Product search through dialogue
After selecting the “Looking for a product” functionality, users will receive buttons with search suggestions. They can use these prompts or send their natural language text message. For example, the user selected the “Women’s sandals” button in the case below. In response, he received a carousel with suggested shoes matching this query. In the next step, the user chose the hint “Black,” which narrowed the search results to this color. Then he decided on the prompts “Material” and “Suede,” followed by “Price” and “About PLN 150.00”. As a result, the search results were narrowed down to these three combined criteria. The user can obtain a similar search result without using hints, by writing a message containing his criteria.
Product search with image
The functionality allows you to search for products similar to the product in the photo sent by the user. The method uses an image search service. Please follow our article about Visual Search. In the example use case, the user uploads a photo of the product using the icon in the lower right corner of the chat, next to the user’s message box. The dialog agent sends an image query to the image search service, which response with a list of visually similar products from the customer’s database. In addition, the agent presents a list of products in the carousel.
Recommendation engine in the dialogue agent
The recommendation engine is another component that considers the user’s preferences based on the dialogue with the virtual assistant. In the product feedback carousel, recommended products that are closely related to the user’s previous query are presented.
Promotion search
The user can receive information about current store promotions by selecting the “What’s on sale?” option in the carousel. The agent will display a message with the address of the subpage with sales and the location of the store’s promotion regulations.
Searching for the nearest stationary store
The functionality allows the user to obtain information about the location of stationery stores closest to the user’s location. After the user selects the “Nearest store” option, the agent will request the area via the browser. Suppose the user allows the browser to obtain the location of the device. In that case, the agent will display a carousel listing the areas of the physical stores, ordered by the distance of the stores from the user’s device. The agent will display an appropriate message if the user does not allow the browser to obtain the device’s location.
Frequently Asked Questions
The agent can answer the most frequently asked questions by store users. Selecting the “Frequently asked questions” button returns a message with a list of suggested topics. For each case, the user receives hints to narrow down the query. Examples of user questions, agent responses, and button prompts are shown below.
