ropensci_demo.Rmd

## rOpenSci Demo

[rOpenSci](http://ropensci.org/)

### Getting data from the literature - PLOS

This example demonstrates how you can easily get literature data from Public Library of Science from R. 

#### Install rplos

```{r install, eval=FALSE}
install.packages(c("rplos","tm","ggplot2"), dependencies=TRUE)
```

#### Load rplos

```{r load}
library(rplos)
```

#### Search for mentions of Fisher in the author field, returning title and author fields, searching in full papers (not including figure captions, etc.), returning only 25 results.

```{r searchplos, message=FALSE, warning=FALSE}
key <- "key"
out <- searchplos(terms='author:"Fisher"', fields='title,author', toquery='doc_type:full', limit=25, key=key)
head(out) # first six rows
```

#### Quickly visualize variation in frequency of word usage in PLOS journals

A built in function in `rplos` called `plosword` makes a bar plot of frequency of terms across PLOS articles. 

```{r plosword, message=FALSE, warning=FALSE}
library(ggplot2)
plosword(list('monkey','Helianthus','sunflower','protein','whale'), key=key, vis = TRUE)$plot + 
  theme_grey(base_size=18)
```

Or, we can do the same thing manually. First, install a few packages if you don't have them already.

```{r eval=FALSE}
install.packages(c("RCurl","RJSONIO"))
```

Then get counts of terms across papers, then plot using `ggplot2`

```{r visualize, message=FALSE, warning=FALSE}
library(RCurl)
library(RJSONIO)
search <- function(x) {
  args <- list(q = x, fl = 'id', wt = 'json')
  tt <- getForm("http://api.plos.org/search", .params = args)
  fromJSON(I(tt))$response$numFound
}
terms <- c('monkey','Helianthus','sunflower','protein','whale')
temp <- sapply(terms, search, USE.NAMES=FALSE)
df <- data.frame(Term = terms, No_Articles = temp)
ggplot(df, aes(x=Term, y=No_Articles)) + 
  theme_grey(base_size=18) +
  geom_bar(stat="identity")
```


#### Get abstracts of 500 papers, and use the tm package for text mining. 

Get 500 abstracts from PLOS One only. The `*:*` is special syntax to denote *give back everything*

```{r getabstracts, message=FALSE, warning=FALSE}
out <- searchplos(terms='*:*', fields='abstract', toquery=list('cross_published_journal_key:PLoSONE', 'doc_type:full'), limit=500, key=key)
out$abstract[1:3] # take a peek
```

Load the tm package, and create a document library

```{r corpus, message=FALSE, warning=FALSE}
library(tm)
(corpus <- Corpus(VectorSource(out$abstract)))
```

Create a term-document matrix from the corpus, and inspect it. 

```{r tdm, message=FALSE, warning=FALSE}
tdm <- DocumentTermMatrix(corpus,
                          control = list(removePunctuation = TRUE,
                                         stopwords = TRUE, 
                                         removeNumbers = TRUE))
inspect(tdm[1:5,1:5])
```

Various operations on document term matrices

```{r freqterm, message=FALSE, warning=FALSE}
# find terms that occur at least five times across documents
findFreqTerms(tdm, 250)

# find associations (terms which correlate) with at least 0.3 correlation for the term "result"
findAssocs(tdm, "cells", 0.3)
```