Líonra Séimeantach na Gaeilge:
This is the home page for Líonra Séimeantach na Gaeilge (the "LSG", or, in English, the Irish Language Semantic Network), a database consisting of Irish words and the semantic relationships among them. A semantic network of this kind is sometimes called a wordnet, after Princeton's English-language WordNet, which was the first-ever full-scale semantic network (dating back to the mid-1980's). Semantic networks are much richer than traditional thesauri, which generally only record (near) synonyms and sometimes antonyms. The LSG, like most other wordnets, encodes a richer set of relationships, including hypernyms and hyponyms (broader and narrower terms), meronyms and holonyms (part vs. whole), etc.
Semantic networks have many applications in Natural Language Processing. They are used in systems for word sense disambiguation, document summarization and indexing, and information retrieval. When a semantic network in one language contains mappings to a second language (ours is linked to the English WordNet), it can be used in various ways to improve machine translation. In general terms, from an artificial intelligence perspective, a semantic network encodes some of the "real-world knowledge" that is required for computers to understand and process texts in a non-trivial way.
The image to the left is a depiction of the full LSG (click it for a full-size version). In fact, this is a simplification of the true picture — each node in the image represents a whole set of synonymous words which could be added as additional branches. Something like this network, but probably thousands of times more complex, is contained in the brain of every Irish speaker — semantic connections like these are made instantly and intuitively. The 3D graph browser (see below) allows you to "fly" through this network and manipulate it in various ways.
- As an enriched "thesaurus" in PDF format. Note that each word in the body of the text is a hyperlink cross-reference. You can download the (nearly 50MB) PDF directly, or else the LaTeX source if you want to build the PDF yourself (e.g. with different fonts). Please save the file locally on your computer to conserve bandwidth.
- Integrated into the free OpenOffice.org office suite. You can install this directly using the "Install new dictionaries..." wizard ("Suiteáil foclóirí nua..." in the Irish version). Alternatively, you can download the ZIP file containing the thesaurus and install it manually. Here's a screenshot of the thesaurus in action.
- As a three-dimensional network browser, hosted at aimsigh.com.
- In RDF-XML format, for use by other computer applications. Contact me if you are interested in this.
And if you are interested in the source code that I used to create the network, visit our development site at Google Code.
- Comprehensive database. There are 32742 synsets, 36262 headwords and 77596 individual word senses, including a great deal of modern terminology, as well as literary and dialect forms, slang, etc.
- Free license. Like the Princeton WordNet (but, unfortunately, unlike nearly all other wordnets in existence), the LSG is free software. Specifically, all data, including the PDF thesaurus, are released under the terms of the GNU Free Documentation License. This means, in short, that you have the freedom to copy and redistribute the data, with or without modification, as long as you do so under the same license.
- English mappings. Entries in the LSG are linked to synsets in the Princeton WordNet. This is a key element in my ongoing work on English-Irish machine translation.
- Frequent updates. I am planning on providing regular updates, incorporating corrections and refinements, but also reflecting Irish as a living language via new terminology, shifting usages, etc.
- Common lexicon. The database used to generate the thesaurus is the same one I use to generate the GaelSpell family of spellcheckers and the Gramadóir grammar checker. Improvements to one project will be reflected in the others automatically.
This work is licensed under a Creative Commons Attribution-ShareAlike 3.0 Unported License.