With new NLP hyperparameters and the power of automation, you can run new AI experiments much faster with DataRobot AI Cloud, which takes text data prediction explanations to the next level. Understand the impact of text, such as patterns of positive and negative feedback in customer reviews or product feedback. Access more granular insights to help determine new use cases and increase the trust, usability, and explainability of text features.
Find scale and speed with improvements in scoring code for time series models that involve large datasets. Now you can download the scoring code for time series models and run predictions on your largest datasets outside DataRobot AI Cloud in the environment of your choice.
On top of these new features, we are excited to announce that Algorithmia is now integrated with the DataRobot AI Cloud platform. Reduce operational costs for AI projects by deploying custom models to the Algorithmia environment. Take advantage of autoscaling deployment while benefiting from the DataRobot MLOps built-in capabilities.
Learn more about other features that are only found in the DataRobot AI Cloud platform.
Text Prediction Explanations (Public Preview)
Text data conveys a lot of information, and you use it more as you build models. Therefore, it is crucial to explain how a machine learning model interprets text and how accurate the prediction is. With advanced text prediction explanations exclusive to DataRobot users, you can access more granular insights and understand the text and its influence on the model at a word or phrase level, both with negative and positive impacts. You can access granular insights via GUI or API and identify the limitations of the explanations to make better decisions and build trust in your models. Advanced text prediction explanations allow you to unlock new use cases and explore new patterns while running experiments for AI projects.
An example based on movie reviews is shown below where a model is used in an attempt to predict the sentiment of movie review text. Text Prediction Explanations indicate that the word “movie” (and others highlighted in red) is indicative of a strong positive impact on the prediction, whereas the word “right” (and others highlighted in blue) is indicative of a strong negative impact on the prediction.
Enhanced Language Detection for Autopilot (GA)
Your business is global, and you have datasets in several different languages. Not a problem anymore! There is no need to manually identify your texts’ languages and spend countless hours adjusting your models to find the best set of hyperparameter values. We have implemented a best-in-class deep learning model for language detection that will automatically optimize hyperparameters for you to hit the ground running. Your dataset can include several languages, and the AI Cloud platform will do the rest of the work. This feature is available to all users and exclusive to DataRobot.
New Text Preprocessing Composable ML Tasks (GA)
Are you interested in using Composable ML to take the next step in experimenting with building models? Do you have text data that you want to leverage? We’ve introduced new Text AI tasks to help you featurize texts in various ways and build an even greater variety of blueprints. New tasks that cover lemmatization, PoS tagging, and stemming are now available through Composable ML.
Time Series Scoring Code (GA)
Export and consume time series models with a few clicks. Scoring code for time series models now runs scalably on your largest datasets. You can directly export time series models in a Java-based Scoring Code package and run predictions on the environment of your choice.
Save time and resources to manage your models. You can easily download Scoring Code from the Leaderboard or the deployment with a few clicks. With time series scoring code, you can generate predictions outside the DataRobot AI Cloud platform in the environment of your choice, private cloud, or hybrid cloud. With MLOps agents, you can also monitor them as well.
Now Algorithmia users can leverage model monitoring capabilities within DataRobot MLOps. The integration allows you to use API calls to connect external models and use DataRobot’s built-in monitoring features to ensure that models perform in an expected way. With that, you can access monitoring statistics, such as service health, drift, and accuracy, to evaluate the performance of your production AI. At the same time, AI Cloud platform users can deploy custom models into Algorithmia infrastructure, take advantage of deployment autoscaling, and reduce operational costs for AI projects.