Azure Text Analytics

The multi-language support and ability to create custom language models has been beneficial to our organization, which operates across multiple regions globally and has many different language-speaking communities. The custom language model capabilities have allowed us to test the platform with niche use cases. The overall performance of the platform has been strong, especially with regard to speed and accuracy. We've been able to use the platform with large sets of language content and have not noticed any degradation in platform performance or diminished accuracy. The ability to integrate with other Azure services (especially Azure Cognitive Services) has been another strong feature, as our organization has extensive use of Azure cloud services and a need to integrate sentiment analysis into various application stacks.

My preferred part is the flexibility and the scalability of it. It's able to handle different workloads easily and supports over 10 languages, even if we don't use all of them it's great for our global business environment. The specialized modules and the real-time opinion are super effective and consistent so it delivers really good insights.

Our Use-Case made it, so that we need a tool to analyze a large amount of texts without branching out into a completely new technical world but rather stay in a known Azure environment, with it being a reliable integration to our current setup of cloud tools.

We've had some issues with context sensitivity around using Azure Text Analytics, as the sentiment models have largely been trained on product and service reviews and don't support some of our domain-specific needs (such as employee feedback surveys) as well as we had hoped. Due to the training set of data, the Text Analytics platform isn't as context aware for text that isn't related to product or service reviews. The pricing structure is not as intuitive as we would like and the costs started to mount quickly as we were using the service with larger sets of text data and with more frequency. The documentation for the service is not as comprehensive and well-structured as other products - we've found gaps in certain key areas and the organization of documentation could be improved. Additionally, the support for implementation could be stronger.

The only area I would like to see improved is the pricing flexibility, I don't want to pay for things I won't use and I would like more options tailored to different usage levels, that would make it easier to scale and optimize cost

There are certain features missing like a cache to save your work after you initially analyzed a piece of text, where you need to go through it all and cant cache it. Also i'm missing a double review button when you analyze a text and delete something you just did its directly gone, but other steps require more confirmation to go ahead with deletion. Last part is id love to have a more direct API-approach, where i can send a marked text to and get its contents back by the API. Maybe it is possible but i havent found it so far.