Train sentiment analysis model

Train ThinkAutomation to grade and score your company messages based on sentiment. Already primed with built-in training data, you can also feed your own training data into the sentiment analyser for pinpoint accuracy.

Train sentiment analysis model – overview

• As one of its baked-in actions, ThinkAutomation can automatically grade messages based on constructs such as positive/negative, or sales/support query, for example
• You can use data from incoming messages – emails, web form submissions, survey attachments etc – to train the sentiment analyser
• You can also use a dedicated sentiment analysis control panel to add training data, run tests, and view your sentiment database
• You should train roughly equal numbers of the constructs you’re scoring (i.e. 50% positive, 50% negative), and seek to use actual messages rather than keywords alone
• Scored messages will be returned as a number between 1 (maximum negative) and 100 (maximum positive)
• You could then use the scores to perform specific actions as part of a broader workflow/process

Train sentiment analysis model – use cases

• Prioritise and escalate emails based on sentiment
• Send an SMS alert if an incoming email contains a high negative sentiment score
• Score emails for sales-readiness and email prospects to sales
• Segment customers based on satisfaction
• Automatically push positive feedback to a Teams channel
• Shortlist CVs based on keywords
• Plus countless more, depending on your desired use case