# ////////////////// #
# /// maxmodelr /// #
# //////////////// #

# SET UP

# ... by cloning the repository
$ git clone https://github.com/dh-thesis/maxmodelr.git
$ cd maxmodelr


# REQUIREMENTS

--> check if you fullfill the requirements listed
--> at https://github.com/herreio/maxplanckr


# PUBLICATION SEARCH

# start R console
$ R

# if it's the first time, packrat will install itself
# after that simply run the following command to set up R environment
> packrat::restore()
# this will take same time, grab yourself a coffee...

# load package
> devtools::load_all()

# load topic coupling data (k = number of topics)
> couple <- load_couple(k=25)
> format(object.size(couple), units="MB")
# [1] "541 Mb" <-- be aware of this memory usage, you need some free RAM!

# set search query
> query <- "information retrieval"

# choose topic model
> models <- c("lda_all","lda_mpi","lda_pers","btm_all")
> model <- sample(models, 1)

# get n most probable items with same topic as query
> n = 20
> search_topic_items(couple, query=query, model=model, n=n)

# take c most probable items with same topic as query
# and compute js-divergence between theta of items and query
# return the first n items
> c = 100
> search_dist_items(couple, query=query, model=model, n=n, c=c)

# take all items with same topic as query
# and compute js-divergence between theta of items and query
# return the first n items
# !!
# !! depending on the number of topics in the model, this means comparing lots of vectors
# !! the lower the number of topics, the more publications are in the same topic
# !! and need to be compared. it is suggested not to use this function with the
# !! example coupling data provided in this repository with k = 25.
# !! instead create coupling data with k >= 100 (see below)
# !!
> search_topic_items_dist(couple, query=query, model=model, n=n)


# DATA PREPARATION

# create models
$ exec/runbtm > btm.log 2>&1
$ exec/runlda > lda.log 2>&1

# infer titles
$ exec/infer > infer.log 2>&1

# couple titles and topics
$ exec/couple