Permalink
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
105 lines (69 sloc) 3.55 KB

Apps for counting words or categories and viewing them in context

This is experimental code with no guarantee of working at all. All subject to change.

Installing Java

You'll need an up to date Java install for all these things. Specifically, the JDK. You can get the currently latest one here after accepting the terms and conditions radio button and running the installer.

You can check on Unix / Mac by opening a terminal window and typing

java -version

This should give you three lines of output. The top one should be

java version "1.8.0_60"

All good.

Tools

The command line tools now run off a common root, this is either the jarfile jca-<version>.jar or a bash script called jca (that just wraps a java -jar jca-<version>.jar call).

Here are the tools:

  • cat: Apply a Yoshikoder dictionary to a bunch of files and record the result in a CSV file.

  • word: Count words and drop the results in LDAC, Matrix Market, or CSV format into a folder. If you have more than a handful of documents you probably don't want to choose the CSV option.

  • conc: Construct concordances (a.k.a. 'keyword in context') for a word, a phrase, or a category from a content analysis dictionary, for a bunch of documents.

  • line: Turn folders of documents into Mallet's preferred one-per-line format. This has three columns. The first two are the file's old name and the final one is the text in one long line.

  • desc: List some summary statistics for your files.

In case you're wondering, this code deals with multi-word pattern matches in dictionaries and outside them. That code just hasn't made it into the Yoshikoder yet.

Usage: Command line

The release file jca-<version>.zip contains a jar file and tiny bash script jca to drive it.

Put the jar file anywhere you like and note down its path. Then type

java -Xmx512M -jar <path>jca-<version>.jar <tool> [options] <files>

where <path> is the folder where you left the jar file (~/bin/ is a pretty good place for Unix / Mac folk), <version> is the latest version (currently 0.2.4.1), <tool> is one of the tools listed above, options vary by tool, and <files> are names of text files and folders full of text files.

You can give any mixture of files and folder names. If a folder name is provided then the documents within that folder will be analyzed, but the tools don't look into any sub-folders that might be down there too.

Unix / Mac folk might find it more convenient to add

alias jca='java -Xmx512M -jar /path/to/jca-<version>.jar'

to their ~/.bash_profile and then use

jca <tool> [options] <files>

instead.

In case you're curious -Xmx512M gives the application half a gigabyte of memory (512M) for the tools to work within. This should be plenty. If it isn't, then either you're building CSV files from really large numbers of documents, your documents are absurdly large, or my code is taking a memory leak. Let me know if you think it's the latter.

Usage: R

If you don't really love the command line, you might prefer to interact with these tools via an R package. That R package is here.

Compilation

If you want to compile the code, pull the source and type ant dist. You know, ant - that thing they built back when XML seemed like a good idea.

Does she go? Build Status

Will Lowe, October 2015