posthoc - Functions for ML/NLP experiment post-processing and analysis
David Andrzejewski (david.andrzej@gmail.com)
DESCRIPTION
posthoc is a Clojure package containing functions for the detailed
analysis of ML/NLP experiments (especially latent topic modeling). In
order to understand algorithm/model performance, it is often helpful
(or necessary) to answer detailed (but arbitrary) questions which were
not considered before running the experiment. For example, after
running Latent Dirichlet Allocation, we may wish to inspect word-topic
associations for a certain term broken out by the co-occurrence of
some keyword within a 10-token window. The goal of this package is
then two-fold:
1) Provide a kit of 'building block' functions which can be assembled
in order to construct novel analyses.
2) Save especially common/useful specific analyses.
LICENSE
This software is open-source, released under the terms of the GNU
General Public License version 3, or any later version of the GPL (see
COPYING).
SLOPPY/INFORMAL/INCOMPLETE API
fileio.misc
fileio.parse-labels
fileio.parse-topics
fileio.wordcount
annotate-table
-if words are positively labeled, print with \textbf{}
annotate-topics
-are words positively annotated?
best-topics
-find 'best' standard LDA topics wrt word enrichment
check-excl
-eval satisfaction of sentence exclusion
check-incl
-eval satisfaction of sentence inclusion
dependency
-parse/manipulate/analyze Stanford-style dependency parses
eval-relevance
eval-labels
-eval Top N topic words wrt relevance to target concept
mean-avg-prec
-mean average precision of a given ranking
pool-topics
-take union of Top N concept-topic words from different runs
topicutil
-data structs and fcns related to topic models
util
-very general Clojure convenience/utility functions