Adding an analyzer

THIS PAGE IS A WORK IN PROGRESS

This page explains the higher-level steps needed to add an Analyzer to TMA and will use the LDAAnalyzer as a running example: ldaanalyzer.py

1. Add the analyzer file to TMA/src/backend, e.g. ldaanalyzer.py

2. Make the analyzer class a subclass of TMAnalyzer and define an initialization that takes a list of parameters as input: from tmanalyzer import TMAnalyzer class LDAAnalyzer(TMAnalyzer): def __init__(self,params): """ implementation goes here"""
   the input parameters should correspond (mostly) to the parameters that will be passed to the topic model software.

3. Implement the methods do_analysis and create_relations,
   do_analysis should create the topic model from the data
   create_relations should populate the tma database tma.sqlite, which is used to create the output displays -- see ldaanalyzer.py for a template on how to populate the database.

4. Implement the perplexity determination if applicable (TODO expand this section)

5. Add the frontend for the analyzer in src/backend/forms.py

• Add the algorithm to std_algo
• create the appropriate fields prefixed with the algorithm, e.g. lda_topic_init, lda_alpha, etc
• create a retriever to retrieve the given fields (see the other retrievers for an example)

6. Add the algorithm to the algorithm form template under src/templates/algorithm-form.html (follow the example of the other algorithms)

7. Update the wiki with information about the new topic model. see Topic Models