SemanticSpaceExplorer

Manually inspecting S-Spaces

Introduction

The SemanticSpaceExplorer tool is a quick and easy way to check the contents of one or more sspaces. It is useful for gaining access to sspaces, especially those in [binary format] (/fozziethebeat/S-Space/wiki/FileFormats_ . Another useful feature is its ability to act as a scripting engine.

Using SemanticSpaceExplorer

SemanticSpaceExplorer understands the following commands:

• load file1.sspace [file2.sspace...]
• unload file1.sspace [file2.sspace...]
• get-neighbors word [number (default 10)] [similarity measure]
• get-similarity word1 word2 [similarity measure (default cosine)]
• compare-sspace-vectors word sspace1 sspace2 [similarity measure (default: COSINE)]
• help
• set-current-sspace filename.sspace
• get-current-sspace
• alias filename.sspace name
• write-command-results output-file command...
• print-vector word
• print-words [string-prefix]

Note 1: most of these commands can be used in abbreviated form, using the first letter of each word (e.g. get-neighbors can be invoked using gn). Note 2: similarity measure is the string equivalent of the [Similarity.SimType] (http://fozziethebeat.github.com/S-Space/apidocs/edu/ucla/sspace/common/Similarity.SimType.html) enumeration. These can be abbreviated with any unique prefix of the string.

Run commands from prompt

Running:

java -cp classes/ edu.ucla.sspace.tools.SemanticSpaceExplorer

will open up a prompt for direct keyboard input. Use ctrl-c to exit.

Run commands from script

Running:

java -cp classes/ edu.ucla.sspace.tools.SemanticSpaceExplorer --executeFile awesome.script

will run the commands in awesome.script, allowing you to easily reproduce your work.

Examples

The first step is to load the sspace:

load processed.sspace

To test that a word retained its meaning one could check:

get-neighbors violin 5 COSINE

to ensure that all 5 words are valid neighbors (semantically speaking).

Or to test that two words are related:

get-similarity violin guitar COSINE