Blog archive

What is the Dasha command line interface? It's a NodeJS command-line script that helps you with the configuration and maintenance of your Dasha integration.

AI in banking can detect risk, assist employees, answer customers, and coordinate workflows. The safe design keeps identity, policy, credit, fraud controls, and money movement inside bank-controlled systems rather than handing authority to a model.

A few years ago mentioning conversational UX raised eyebrows. Today, it is par for the course. We can talk for ages about using conversational AI design to create amazing user experiences (and will soon). Instead, today, I want to tell you how you can use a key feature of the Dasha AI Platform -…

The customer is always right. As small business owners, we know that this isn’t always the case. Yet, we will and should do everything in our power to ensure that every experience our customer has with our business is a delightful one.  A delighted customer is more likely to spread the good word…

The machine learning random forest algorithm is one of the most amazing ML algorithms invented by Leo Breiman and Adele Cutler back in the last century. It has come down to us in its "original form" (no heuristics have been able to improve it significantly) and is one of the few universal…

Let’s talk about self-supervised machine learning - a way to teach a model a lot without manual markup, as well as an opportunity to avoid deep learning when setting a model up to solve a problem. This material requires an intermediate level of preparation; there are many references to original…

Let’s talk about something that cannot but excite your imagination - let’s talk about key breakthroughs that have occurred in speech recognition thanks to transformers.

Surprisingly, it is extremely difficult to find a good overview of all the methods of model calibration - a process as a result of which the “black boxes” not only qualitatively solve the classification problem, but also correctly assess their confidence in the answer given.

The logistic loss or cross-entropy loss (or simply cross entropy) is often used in classification problems. Let's figure out why it is used and what meaning it has. To fully understand this post, you need a good ML and math background, yet I  would still recommend ML beginners to read it (even…

Whether you need to drown out extraneous noise in recorded speech, get rid of echoes, or simply separate the voice from the music, this guide can be very helpful to you. It's okay if you've never worked with Speech Enhancement models and Denoising tools - I’m here to describe all the steps and…

Let’s say you have a working and a developer-friendly .NET ecosystem. There are a lot of services and your team doesn’t cherish the idea of having a service built without .NET. Additionally, there is a pending request to develop software to serve some machine learning models.

In the process of Dasha's dialogue with a person, depending on the phrase said by the person and the algorithm implemented in DashaScript, Dasha chooses a response phrase. After that, to convey that phrase to the person, the TTS service comes into play.

In this post, we will be addressing the quality of machine learning algorithms for solving classification problems. We will consider the "recall-precision" curves as well as Gain, Lift, K-S (machine learning curves), and a table for analyzing profitability. Most importantly, we will define all…

My name is Ilya, and as a machine learning researcher, it’s kind of my duty to think about AI and the impact it has on our reality. In my previous article, I tackled questions that you might have tried to wrap your mind around: will AI take our jobs – and if so, which ones would be the first? And…

We’ll talk about two things - a deep learning approach that solves the problem of reducing the sample size and an even more ambitious task - creating synthetic data that stores all the useful information about the sample. In this post you'll be learning to distill, so to speak.

This blog post is for you, Python lovers. Last year there was an experiment conducted.

Let’s see how you can easily create a voice & text assistant, and teach it to control your robot vacuum cleaner like Xiaomi or iRobot.

If you’re considering voice AI as a way to boost your call center performance, sooner or later you’ll face the big question: should you use synthesized or pre-recorded speech for AI voice output?

Let’s discuss the highlights of the recent speech separation and recognition challenge as well as some tricks used by the winners.

Do you really think you should outsource lead generation offshore? Let me change your mind.

The call center outsourcing market has lately been booming and estimated at USD 75 billion. But recent pandemic disruptions have taken their toll on the industry, bringing old problems to light and creating new ones (enter WFH).

GPT-2 and other tools now can generate texts. But so far, the results are not that... impressive. Let's talk about Neural Text Degeneration.

What if I told you that Alan Turing’s work started impacting today’s study of AI before computers were invented?

Average Handle Time. Out of all call center metrics and industry standards, AHT sticks out like a sore thumb. Managers chase it, call center operators hate it. Let’s dig in and try to figure out the value of this metric, as well as its impact on your call center’s day to day productivity - both…

Everybody hates long wait times. Here is how you make them history.