Azure NLP and Power BI solution for chemical manufacturing

Natural Language Processing (NLP) to generate customer insights in color consultation platform for MEIA region

Company operates in four segments: Decorative Paints, Marine Coatings, Protective Coatings, and Powder Coatings. Each segment offers its own solutions, but they all share our vision of using paints and coatings to protect property.
Company use WhatsApp Bot for Business for Middle East India and Africa in different languages like English, Arabic, Hindi etc.
Basic reporting is being currently used which shows basic KPI information related of marketing and demographics like total conversations, key points etc.

The existing reporting is unable to provide detailed analysis of the customer conversations like interests of customer, which paint colour they are more interested in, which part of house is commonly enquired and about the customer sentiments during the conversation.

To meet the key business requirement, an AI based NLP (Natural Language Processing) solution is required.
This AI based solution will read the conversation and provide a text summary of understand the meaning of the conversation.
From the text summary, key phrases can be extracted to know main pointers of the conversations
From the conversation, sentiment can be analyzed to get a comparative idea of customer satisfaction.
Also, as conversations are happening in multiple languages hence a language detection and conversion is also required for better analysis.
For showing the end-result a reporting layer is used

Azure cloud-based solution.
Incremental data load to load only newly arrived data to increase and optimize performance.
Calculating response time of the conversations and identifying conversations which did not initiate or in which either the customer or consultant did not get back.
Mapping numeric responses in conversation like 1,2 to their appropriate categories of language English/Arabic or paint area type Interior/Exterior etc.

Removing bot conversations and compressing the conversation in a list based on Conversation ID.
Perform NLP by detecting the language, translating Arabic text, and summarizing it. Perform sentiment analysis, extract key phrases, and translate the summary, sentiment, and key phrases into Arabic.

Exception handling and workflow notification on Outlook were created to keep a note of the data push.
Triggers are created at each part of processing for smooth data transfers.