Customer complaints as well as compliments can be automatically categorised by companies using DataSift's new machine learning tool.
The Big Data analysis firm will classify people's rants and raves into those two aptly-named categories, while other online utterings could be pigeonholed into "purchase intent" or "churn".
"Companies are now hungry to go beyond sentiment analysis into more advanced insights," said Tim Barker, chief product officer at DataSift.
"Social data has evolved. Everyone from financial institutions through to the United Nations refers to it and we recognised that people need actionable, nuanced insights from social data to better understand their audience's mood and intent."
DataSift is trying to provide this with its latest product, VEDO Intent, which aims to replace manual analysis completed by data scientists with an automated approach to save time and cut costs.
Bypassing analysts means VEDO Intent allows end users in marketing, customer relations and other departments to build and fill categories for unstructured data by themselves.
At first, the tool learns how the data is classified as the task is performed manually, before building its own machine-learning model to automate the classification of millions of social media snippets.
Privacy is retained by having this automation complete the process, rather than data scientists, according to DataSift.
It means firms can ask questions of their data to find out which of their products prove popular, how satisfied customers are and the strength of their marketing campaigns.
DataSift launched VEDO Intent today, as well as announcing an initiative called Forum that aims to accelerate innovation in machine learning's applications to social media data.
It hopes Forum will help its partner ecosystem of developers and data scientsists develop more algorithms for human-created data.
To kick things off, DataSift has released its keyworld relationship model to its partners, helping them analyse hashtags, keywords and topics that are all related to one another.
