The bot for searching in the Schneider Electric directory

Schneider Electric Russia is a manufacturer of equipment and machinery for mechanical engineering and energy, as well as household goods.

Search for information in the Schneider Electric technical directory.

Customer

Task

Solution

The chatbot that is well-versed in the directory and gives the most relevant answers.

The service is addressed to design-engineers who help customers select equipment and other goods.

01/

The engineer writes the name or characteristic of the product to the bot

The bot searches through the directory and gives a list of relevant answers

The customer applies to the company

02/

01

Problem

Lack of technical texts for bot training — initially the accuracy of the answers was low.

Solution

Schneider Electric helped to train the service by adding alternative product names and formulations to the text database.

02

Problem

An inconvenient document format is – PDF with text descriptions in Russian.

Solution

We translated the document into the table format, in which we manually prescribed the properties of each product.

03/

Debugging: testing and training the bot together with the customer

Development and selection of the most appropriate method of search through the bot's knowledge base

Step 03

Step 02

Analytics: filling the bot's knowledge base

Step 01

04/

1 month

March-April 2021

05/

EORA platform for the development of dialog systems: chatbots, voice assistants, automated contact centers.

06/

Team leader

Aidar Valeev

Back-end developer

Alexey Kazantsev

Anna Mikhnik

Analyst

Ivan Starkov

Project Manager

08/

Ivan Starkov

Project Manager

«Ordinary directory bots choose one of several options for responding to the user's request. In the case of Schneider Electric, the search is based on an index of bigrams and trigrams of keywords, so that the bot can not only search through the query text, but also process possible typos in it»

«The special complexity of the project was in the content: we did not have knowledge in this highly specialized area. The customer helped us a lot — thanks to the participation of one of the managers, the chatbot became an order of magnitude "smarter", and the accuracy of the search increased. Thank him so much!"

Aidar Valeev

EORA Team leader

09/

The bot is running in pilot mode (it searches for 16 pages out of 200).

Other projects of EORA

Universal bot for Purina websites

Chatbot for wallpaper selection on the Leroy Merlin website

Chatbot for generating commercials in the style of ESSA advertising

11/