Etytree lightning talk at Wikimania 2017: transcript

Key points

• The tool is developed with a Wikimedia Foundation grant and combines a database of lexical and etymological relationships with an interactive visualization.
• Typing a word such as pistachio or wig shows a graph of its etymological relationships and the dictionary pages the data came from.
• The workflow takes the latest English Wiktionary dump, parses it into an RDF database, sets up a query endpoint, and builds the visualization in JavaScript with D3.
• The English Wiktionary was chosen because it is multilingual and its etymology sections follow fairly standard rules that are relatively easy to parse.
• The parser uses regular expressions and a grammar to turn etymology sentences into triples such as one lemma being etymologically derived from another.
• Goals include exporting the data to Wikidata, serving linguists, data scientists and language learners, debugging, and finding new collaborators.

Chapters

1. 0:00Introducing etytree and the pistachio example
2. 0:59Source pages and the etymology of wig
3. 2:17Four-step workflow and the Virtuoso database
4. 2:54Why the English Wiktionary was chosen
5. 3:24How the etymology parser works
6. 3:59Goals, uses and next steps

Full transcript

0:00My name is Esther, and I'll present a tool that I'm developing with a grant of the Wikimedia Foundation, which is called ET3, from Etymology Plus 3, which consists of a database of lexical information and etymological relationships, and an interactive tool, and it's based on the initial design. I'll show you the interactive tool, which looks like a test version of the interactive tool which looks like a dictionary, so you can type a word, for example pistachio, and you press enter and you visualize the dictionary, the graphical, the graph of themological relationships So for example here, the English word Pistaccio, here, it's derived from Italian Pistacchio, which in turn derived from Pistaccio in Latin, from Pistaculum in Latin.

0:59And in the same visualization you can see which dictionary pages contributed to this information. So for example here, these four pages are extracted etymological relationships in this graph for this word from these four pages. I'll show you another example that I'm interested in, which is the word wig. So when you type wig, you get a list of words in wiktionary that are spelled the same, which is the one you're interested in. In this case, I'm interested in the English wig, and you get this big etymological tree, not so big actually, but...

1:45So, the word wig derives from periwig in English, which in turn derives from perruc, which I cannot spell, it's French, from belo, from pius Latin. And again, you can mouse over the word to have the lexicographical information of the word. So, going back to the presentation. The workflow consists of four steps.

2:17I take the dump of the English dictionary, the most recent one. I parse it and I create a database, an RDF database. Then I set up a stacker endpoint in virtuoso, at this link from the Wikimedia Foundation labs. And then I create the visualization and the repository is available on GitHub with D3 in JavaScript, which is the tool that I just showed you. I can show you the database, how it looks like. So this is the faceted browser of Virtuoso, and this is the page for English pistachio.

2:54You can click here to get lexicographical information about the word, part of speech, language, science, and so on. Or you can click here and get the etymological information, so the word derives from. Why did I choose the English dictionary? Because it's multilingual, like many old dictionaries, and because etymology sections in the English dictionary are relatively easy to parse.

3:24They follow a clear set of standard rules, although there are many exceptions. For example, this is how the parser works in the simplest case. The etymology section looks like a sentence. It's something like from Italian Pistacchio, from Latin Pistacchio. This is the page for English Pistacchio. And this is how the wiki text looks like. So I use regular expression and the grammar to encode this wording into elements like from, language and lemma, or comma and dot.

3:59And from this I obtained triples, lemma etymologically derived from lemma 1, so pistacchio etymologically derived from pistacchio, lemma 1 etymologically derived from lemma 2, and after substitution I obtained the RDF database, so English pistacchio etymologically derived form Italian pistacchio. So, to wrap up, the idea is to create a database of etymological relationships that eventually can be exported into Wikidata. That could be a resource for linguists, for data scientists, or more people, if you have ideas, and to create a tool to visualize the tree of any word for the general users who is interested in etymology, for people that are learning a language through meaningful learning, which is like connecting to prior learning in such

5:01a way that your knowledge is more highly retainable and generalizable. Future steps are debugging, eventually merging with your key data and hopefully find new collaborators.

Source and licence

Recording: File:Etytree lightning talk.ogv by Epantaleo (2017-08-11), via Wikimedia Commons. Licence: CC BY-SA 4.0. The transcript, summary and chapters on this page are shared under the same licence (CC BY-SA 4.0).

The transcript is machine speech-to-text by the TidyTools Audio & video transcriber; the summary, key points and chapters were written by Claude from that transcript. Both may contain errors: check the recording before quoting it.