Downloading information about text analytics models
Download information about your text models, such as the model binary file, taxonomy and training data, performance reports, and feedback.
Access model information in one of the following ways:
Choices Actions From a model
In the navigation pane of Prediction Studio, click Models.
Click the model for which you want to download information.
In the Text categorization - Topic model window, click Download, and then select a file to download.
From a text prediction
In the navigation pane of Prediction Studio, click Predictions.
Open a text prediction.
Click Models, and then click Model data.
For a model that you are interested in, click Download, and then select a file to download.
- Taxonomy (csv)
- A taxonomy file contains the topic structure and any keywords (Should, Must, And, Not words) provided for each topic. You can download the taxonomy that you created for a keyword-based topic detection model, and then use the taxonomy as part of a machine learning topic model. Or you can use the taxonomy file in text analyzers to perform keyword-based topic detection.
- Training data (zip)
- A compressed CSV file that contains the training and test data that
was used to create the model. The training data file is downloaded
from the repository configured in Prediction Studio. You can download the training data from one model, and then use
it to train another model. This option is only enabled for models that have been trained. It is disabled for pure keyword-based models or if the report is not available due to repository changes or due to model migration.
- Report (zip)
- A compressed file that contains the performance report for the
model. The report is based on the algorithm that was used to
generate the model during the last model update. The file contains
two CSV files per algorithm. The first CSV file contains the list of
test records of the model and the predicted outcome of the model
against each test record. The second CSV file contains the score
sheet that provides the precision, recall, and F-score measures for
individual topics or entities, and for the model as a
whole.This option is only enabled for models that have been trained. It is disabled for pure keyword-based models or if the report is not available due to repository changes or due to model migration.
- Model file (jar)
- A model binary file that is created at the end of each model building process. This binary file is run against incoming data to generate the model output.
- Feedback data (zip)
- A compressed file that contains feedback data given to the model from Pega Intelligent Virtual Assistant (IVA) channels (email or chatbots) or by using the pxCaptureTaFeedback activity. The file includes both reviewed and unreviewed feedback.