3 ms·
My company builds software to analyze customer feedback. We use "real" ML for sentiment classification, as well as some of our natural language processing and
by got2surf 10y ago
My company builds software to analyze customer feedback.
We use "real" ML for sentiment classification, as well as some of our natural language processing and opinion mining tools. However, most of the value comes from simple statistical analysis/probabilities/ratios, as other commenters mentioned. The ML is really important for determining that a certain customer was angry in a feedback comment, but less important in highlighting trending topics over time, for example.
- activatedgeek 10y agoWhat do you mean by "real"?
- got2surf 10y agoSorry, using "real" in quotes wasn't too descriptive. A few machine learning-based classifiers (we've used Bayesian and SVM approaches). Word embeddings and topic modeling (similar to word2vec) which are based on shallow neural networks. Those are a few of what I would consider the "real" machine learning tools we use. Most of the application, though, is statistics/pattern recognition/visualizations on top of the data calculated by the ML approaches. The interesting thing is (in my opinion/experience) that a 10% improvement in some of the ML performance (a 10% increase in accuracy, for example) will translate to a 1-3% improvement in end user experience (they see slightly better insights and patterns, but it is a marginal improvement). On the other hand, layering a new visualization or statistical heuristic on top of the data can lead to a significant boost in user experience. Again, this is just for our specific application/domain, but we focus on making the ML results more accessible to users instead of focusing on the marginal accuracy of the ML results themselves.