Create chatbots as a widget on the Russian Railways educational portal website.
The widget is supposed to process the main array of user requests, recognizing natural speech and selecting a pre-programmed response from the database.
We prepared a system prototype based on the chatbot in Telegram and Viber, which will allow to test the solution on a sample of 100 users. While creating the system, a machine learning algorithm to work with new data was developed.
In addition, data preparation methodology (preprocessing, cleaning) for new array operational processing was created.
The project is at the testing stage.
A pilot version of the product was developed, allowing even on a limited sample to automate 3 main types of standard user questions.
To conduct a semantic analysis and tone value analysis of all product reviews in order to compile an aggregated review with the key benefit identification.
To rank the product characteristics according to the importance criterion for the final user (buyer).
An algorithm to analyze each individual review of a product in terms of tonality in relation to the characteristic was developed , as well as a system of “weights” for the following parameters:
To demonstrate the system capabilities, a demo-bot in Telegram was developed, which at the entrance receives an article or product name, and at the output gives a ranked list of characteristics and a model key advantage based on customer preferences.
The MVP system was developed in 24 hours.
The algorithm was taken by the company for implementation.
1. Create a system that checks if the consultant / operator’s speech matches the scripts.
2. To analyze the emotional component of communication with the client.
A system that allows analyzing the correspondence of speech to scripts was developed. The system consists of two machine learning models of:
An extensive study of the available tagged data on the voice and its emotional component was conducted. An algorithm that determines the basic voice emotions was created.
The project continues; at the moment an intelligent IVR system is being developed, which allows to automate the questions routing and select the most appropriate answers from the database to offload the first support line.
Analyze the data of more than 1000 receipts with the purchase of one of the goods and determine the key patterns in the commodity neighborhood, customer profile and purchase place.
A complex data analysis was performed using the following technology stack:
An analytical statement has been prepared for the company on the following factors:
The resulting insights enriched the customer’s understanding of their brand and helped to improve positioning and planning of marketing activities.
Develop algorithms for specific information searching in various parts of the Internet (databases, trading platforms, websites).
1. All the online trading platforms in the Russian Federation (more than 20 pieces) were analyzed for medical product resale – references to the goods for specific requests were looked for. Integration with each trading platform was configured, algorithms for necessary information search and saving links with references in a separate database were made.
2. A system for automatic collection of the main text from a large number of links was developed. It was adapted to high loads – up to 30,000 pieces in one request.
The client is engaged in marketing research, for which reason he needed a system that automatically highlighted the main text on sites, cutting off the unnecessary things like advertising, table of contents, additional modules.
3. A search system for online tools connected to the site was developed (Jivosite, Yandex.Metrica, etc.). The system reads the site code and searches it for references to various connected tools and then puts all the data into a table.
4. A system for collecting feedback on customer products, their segmentation, markup and classification was developed.
An automatic parsing system was developed that can quickly process large volumes of specific information. The implementation of our development saves more than 20 hours of staff time per week and automates a number of processes in the customer’s company.
Create a parameter accounting system to optimize the decision-making process in the pharmaceutical industry for all levels – from ordering product components and transporting them to distributing the load on production lines, reconfiguring them and overtime work shifts planning.
Customer Development was made with 10 major players in the Russian pharmaceutical market to clarify all the parameters that affect the system efficiency.
An integrated solution architecture was developed, which is built around four main components:
1. Virtual Production Model Reflecting Optimized Processes
2. Module with expert knowledge of the specifics and structure of production
3. The unique optimizing algorithm
4. Convenient user interf
A unique algorithm was developed based on AI technologies, reinforcement learning and advanced mathematical methods. The algorithm is a universal tool for the production of any format that can take into account an unlimited number of parameters. MVP was developed based on market data collected by Cleverbots.
Create an application / chatbot combining the basic useful functions for quality daily monitoring of the disease.
A detailed analysis of the user’s path to the target audience was conducted – plunged into how people with diabetes are monitoring the disease at the moment and what difficulties they have.
The following problems were identified:
A project based on Telegram was implemented.
The main feature of the “dia friend” was the digital intelligence that recognizes the type of food from the photo. In the process of work, datasets were collected (100,000 images per 100 products) and they trained a neural network on image classification.
A comprehensive solution to support people with diabetes was implemented, combining all the important functions of disease control.
Received the highest rating of this decision at the International Digital Pharma Conference.
Finding the target audience by analyzing images on social media.
A system of parsing and image processing from open sources for further analysis of the user’s compliance with the parameters of the target audience was built.
Two pilot launches with FMCG and Pharma are scheduled for early 2019.
Detect illegal fishing in the Gulf of Finland
A system for processing and analyzing aerial photographs of the territory was developed to detect changes in the position of fishing vessels and to compare them with the coast guard data.
The MVP solution was developed in 48 hours and won the international European developer competition.
Automate the process of checking receipts for the presence of purchased goods in it, which is suitable for the promotion.
Develop a widget for the site, as well as chatbots for all popular messengers.
The solution structure was analyzed the best way to solve this problem was found – by checking receipts through the FTS service. Made a connection with the API FTS and learned how automatically and in large quantities to check the receipts contents.
Developed a widget on the customer’s site and chat bots in popular messengers.
Since its launch, more than 5,000 people have used this system.
Work on the system capabilities is proceeded. .
5 different actions were carried out together with the brand SPLAT.
An example of an active campaign: https://avia.splat.ru/
Communication infrastructure construction for 1000 forum participants for the technological leaders of “OSTROV 10-21” with the functionality for networking and team building;
Creating a data graphical display in the web interface for the situational center.
A communication interface (Telegram and VKontakte) with networking and team building functions based on a recommender system was developed.
Minimum time was allocated for the project implementation. The team was divided; part of the team went to The Russian island in Vladivostok to deploy the situation center on the spot. The project Manager from the development side was in Boston, USA, and the main part of the team was in Moscow and St. Petersburg. To eliminate the time difference, the team moved to a 24-hour development cycle.
The system was developed and deployed at the event, the situational center is equipped with dashboards, a lot of positive feedback was received on the recommendation system in terms of networking and team building recommendations.
Develop alternatives to the mobile version of the online store.
An extended version of user interaction scripts with a chat bot was built, taking into account the display of a large number of products from the catalog. A chat bot based on Telegram was developed. A user-friendly interface was created.
A chat bot in Telegram was created, which has the full functionality of an online store: from reading the catalog to placing an order and dispatch. The system has been integrated into General CRM.
Quickly and conveniently collect data from customers for a single action “Rosa Khutor” – organize a mobile laundry at a ski resort in Sochi.
A chat bot in Telegram and VKontakte was developed, which promptly collected contact information from customers and provided users with promotional codes.
Smart Tips is Cleverbots’ own startup, which involves creating a service for paying tips in restaurants using a bank card online.
The service is a web application with which the guest can leave a tip for the service using a credit card.
The service is an online platform and does not imply complex IT integration with ERP and CRM institutions. After the decision to use the service in the institution, the connection takes place within two weeks.
The key component of the service is the payment mechanics worked out from the technical and legal point of view.
Smart Tips is the first Cleverbots startup. In plans-development to the full-fledged application with expanded functionality for users.
The development of the service started in April 2018 and will be ready for launch in the first quarter of 2019. The service will be piloted in three restaurant chains in January – February.