Pienso is different.

Other companies offer topic modeling. What sets Pienso apart?

Machine learning is incredibly powerful, but there’s no replacement for human expertise. That’s why Pienso uses a process known as semi-supervised, interactive machine learning to put an expert in the algorithmic loop — because algorithms and humans are both at their best when they work together.

In  conventional  “supervised”  machine  learning,  “human-in-the-loop”  typically  refers  to  the  process  of  manually  tagging  individual documents in a training data set. This time-consuming work is usually farmed out to an external army of low-wage workers, or “mechanical  turks,”  who  lack  the  expertise  to  classify  nuanced, specialized subjects. In this approach, the human supervision only happens at the beginning of the process (data preparation) and at the end (validation of deployed model’s results).

“Interactive”  machine  learning  means  that  you’re  able  to  directly manipulate your model’s topic definitions. But for most services, this just means that they include the ability to directly edit the code of your algorithm. This means that your model-trainer needs to be proficient in programming languages like Python.

Pienso is different. It doesn’t need tagged training data to initi-ate a model or coding skills to refine it. Instead, Pienso reads every raw document in your training data and clusters them into possible topics. Then a human expert (that’s you!) reviews, refines, and labels these detected topics in the model-training workspace. Adjust topic  definitions  without  ever  seeing  a  line  of  code,  and  validate your  changes  in  real-time  by  watching  documents  redistribute  as you work. We call it “expert-in-the-loop” machine learning.

At the end of this process, your output is a trained model, used to  analyze  new  or  existing data right in the Pienso platform, or wherever your data is stored (or streamed from) — or exported as an API and implemented in your own analytical tools.