Text Classifier Model Pruner
Palladian for KNIME version 2.3.0.202009251618 by palladian.ws; Philipp Katz, Klemens Muthmann, David Urbansky
The “Text Classifier Model Pruner” allows to reduce the size of a text classification model by applying different pruning methods. On the one hand, low frequency terms which were encountered during training can be removed. Our experience shows, that setting this value e.g. to “2” roughly reduces the number of terms in the model to half the amount without significantly harming classification quality (your mileage may vary).
On the other hand, an information-gain-based pruning strategy is available, which scores the terms and their associated category probabilities. A good explanation of the information gain method can be found in “A Comparative Study on Feature Selection in Text Categorization”, Yiming Yang and Jan O. Pedersen, 1997.
Options
- Minimum term count
- The minimum count (i.e. the number of training documents in which it has to appear) to keep a term, set to one to keep all terms.
- Minimum information gain
- The minimum information gain to keep a term. Set to zero to keep all terms.
Input Ports
Output Ports
Best Friends (Incoming)
- Text Classifier Learner (100 %) Streamable
Best Friends (Outgoing)
- Text Classifier Predictor (62 %)
- Text Classifier Model to Table (27 %)
- Text Classifier Learner (5 %) Streamable
- Text Classifier Model Writer (5 %)
Installation
To use this node in KNIME, install Palladian for KNIME from the following update site:
A zipped version of the software site can be downloaded here.
You don't know what to do with this link? Read our NodePit Product and Node Installation Guide that explains you in detail how to install nodes to your KNIME Analytics Platform.
Developers
You want to see the source code for this node? Click the following button and we’ll use our super-powers to find it for you.