ACRE Technologies



Click a link below to learn more about the technologies that make ACRE different.

Or, refer to our help files to learn more about how ACRE works.

>        Architecture

By defining four types of user jobs whose performance can be independently tuned and scaled, ACRE ensures that both large and small text processing tasks can be completed efficiently.

>        Label Models

Any number of label models can be executed.  Each model execution adds a new set of metadata to the Model Explorer and Data View.  The possibilities are endless.

Models can utilize binning, pattern match, rules, machine learning, or a combination of these to select a label for data.

>       Category Trees

Category Trees store a hierarchical representation of all possible label values for a model.   Many existing taxonomies (such as org charts, folder trees, directory trees, etc.) can be imported into ACRE.  The tree structure also defines the data aggregation points for reports, summaries and word clouds.

>        Natural Language Processing (NLP)

ACRE provides a variety of NLP functions that further classify text and define the vocabulary to be used in analysis.  

>        Machine Learning

ACRE's Nearest Word Cloud algorithm is designed specifically for text analytics, providing some advantages over algorithms typically used for numerical analytics.  The algorithm is intuitive, transparent, tunable and scalable.

Users can rapidly prototype ML models with default settings.  Simply Train and Execute!  But for expert analysts who want to tune performance, ACRE provides a variety of modeling parameters.

>        Combined Models

ACRE's simple 1-click model combinations extend analysis in useful ways without requiring technical knowledge.

o   The Machine Learning Extension feature extends Rules-based model matches (which must be exact) to add more labels using fuzzy matching (based on similarity).

o   Models can be chained so that multiple models are attempted until one succeeds.

o   Results from one model can be used to train another model.



Click a help topic to learn more.