α] warning

Alpha/draft version. READ THE WARNING. DO NOT LINK widely. Typo/bug/feature reports not helpful ATM :)

Many results from radix will be ! INCORRECT !, confusing, misleading, or incomplete.

You should quasi-verify results by clicking around the wiktionary pages and reading etymology sections. Radix is a good tool for discovery of plausible cognates, but unfortunately is not yet very close to being clear and reliable.

Alpha disclaimer

If you're reading this, you're interacting with an alpha version of radix. That means:

Things might break, e.g. the website might go down or some pages might not work.
Things might suck, e.g. be unfinished or have a messed up UI, or the server might be really slow.
Things might change, e.g. the output format/styling or url format might change.
Features might come and go.
The quality of the output might be notably worse than it will be later.
The server might not be secure.

Kinds of error

Here are ways that the radix page for word W (more precisely, for one sense of W) can be wrong.

In many cases, the page will have a:	Which means that the page:
False cognate	shows another word V in a list of cognates of W and/or in a pastward or futureward cognate poset of W, when in reality V is not a cognate of W.
Missing cognate	fails to list a true cognate V.
Misleading sense	correctly shows a word V as a cognate because V has a sense S that's cognate with W, but S is a surprising, uncommon, or obscure sense of V that is homographic with a more common sense Z of V that is not cognate with W, and this is not made clear.
False sense	correctly lists a cognate V, but includes senses of V that are not cognate with W.
Missing sense	correctly lists a cognate V, but fails to include the sense of V that is cognate with W.
Overconfident cognate	shows a word V that is only possibly, plausibly, perhaps cognate with W.
Non-existent word	shows a word V that is not actually a word, e.g. it is misspelled or contains markup or punctuation.
False descent	shows V as descending from another word U, when in reality V does not descend from U.
Missing descent	fails to show V as descending immediately from U, when in reality U is a most recent ancestor of V.

In many cases, the page will have a:

Which means that the page:

False cognate

shows another word V in a list of cognates of W and/or in a pastward or futureward cognate poset of W, when in reality V is not a cognate of W.

Missing cognate

fails to list a true cognate V.

Misleading sense

correctly shows a word V as a cognate because V has a sense S that's cognate with W, but S is a surprising, uncommon, or obscure sense of V that is homographic with a more common sense Z of V that is not cognate with W, and this is not made clear.

False sense

correctly lists a cognate V, but includes senses of V that are not cognate with W.

Missing sense

correctly lists a cognate V, but fails to include the sense of V that is cognate with W.

Overconfident cognate

shows a word V that is only possibly, plausibly, perhaps cognate with W.

Non-existent word

shows a word V that is not actually a word, e.g. it is misspelled or contains markup or punctuation.

False descent

shows V as descending from another word U, when in reality V does not descend from U.

Missing descent

fails to show V as descending immediately from U, when in reality U is a most recent ancestor of V.

Sources of error

Radix

This is a partial list of ways that radix introduces errors. (See radix.ink/_future for more discussion.)


In general, missing features of how wiktionary works	For a long time ~~radix~~ I did not know about redirections; there may be other such features.
In general, making assumptions about how wiktionary works	For example, radix assumes that all articles in the wiktionary xml dumps are contained in <page> tags and have a title in <title> tags, which is probably true, but there are lots of assumptions like this, each of which could be false.
Misparsing articles	E.g. not handling ===, {{, }}, and \| correctly, especially nested like {{\|{{\|}}\|}}.
Omitting or misinterpreting headers	There are a bunch of headers in wiktionary articles. Radix doesn't know about all of them, and might misinterpret them.
Mispartitioning headers into senses	There's no strict layout in wiktionary articles, and in particular it's not always determined how to partition subsections of one language's article into sections that discuss senses of the word which all share a single etymology.
Omitting or misinterpreting templates	There are many wiktionary templates--markup sections that, among other things, crucially specify etymological links between words. Many templates are discarded by radix, and might be misparsed.
Failing to parse out word texts	Wiktionary has some way of formatting the texts of words in templates using [[]] and stuff, which I don't understand and which therefore sometimes produces garbage in radix results
Applying the wrong text transformation to word texts	The texts of words in wiktionary templates are written to be displayed, which is often not identical to the text of the word in the wiktionary article for that word (e.g. a word displays with diacritics, but some of the diacritics are absent in the article title). The makeEntryName function and other code does this translation in wiktionary (see en.wiktionary.org/wiki/Module:languages), but I don't know how to plug the actual code into radix. So radix uses a janky old partial version of that code, which is okayish but probably incorrect.
Incorrectly inferring etymological links	Some links between words are straightforward, e.g. the inherited template. But radix goes further and uses abstract regexs to infer more links, e.g. saying that W comes from V if some other article says in its etymology section "... W, from V, ...". This works somewhat well, but sometimes falsely infers links.
Incorrectly in- or ex-cluding senses	If a word has multiple senses, radix uses a few heuristics to guess which senses are and are not really cognate with the present word. The heuristics are far from perfect.

In general,

missing features of how wiktionary works

For a long time ~~radix~~ I did not know about redirections; there may be other such features.

In general,

making assumptions about how wiktionary works

For example, radix assumes that all articles in the wiktionary xml dumps are contained in <page> tags and have a title in <title> tags, which is probably true, but there are lots of assumptions like this, each of which could be false.

Misparsing articles

E.g. not handling ===, {{, }}, and | correctly, especially nested like {{|{{|}}|}}.

Omitting or misinterpreting headers

There are a bunch of headers in wiktionary articles. Radix doesn't know about all of them, and might misinterpret them.

Mispartitioning headers into senses

There's no strict layout in wiktionary articles, and in particular it's not always determined how to partition subsections of one language's article into sections that discuss senses of the word which all share a single etymology.

Omitting or misinterpreting templates

There are many wiktionary templates--markup sections that, among other things, crucially specify etymological links between words. Many templates are discarded by radix, and might be misparsed.

Failing to parse out word texts

Wiktionary has some way of formatting the texts of words in templates using [[]] and stuff, which I don't understand and which therefore sometimes produces garbage in radix results

Applying the wrong text transformation to word texts

The texts of words in wiktionary templates are written to be displayed, which is often not identical to the text of the word in the wiktionary article for that word (e.g. a word displays with diacritics, but some of the diacritics are absent in the article title). The makeEntryName function and other code does this translation in wiktionary (see en.wiktionary.org/wiki/Module:languages), but I don't know how to plug the actual code into radix. So radix uses a janky old partial version of that code, which is okayish but probably incorrect.

Incorrectly inferring etymological links

Some links between words are straightforward, e.g. the inherited template. But radix goes further and uses abstract regexs to infer more links, e.g. saying that W comes from V if some other article says in its etymology section "... W, from V, ...". This works somewhat well, but sometimes falsely infers links.

Incorrectly in- or ex-cluding senses

If a word has multiple senses, radix uses a few heuristics to guess which senses are and are not really cognate with the present word. The heuristics are far from perfect.

There's a lot of code in radix that isn't touched on in the above table, and of course any of that code could have bugs that introduce more errors. E.g. the order handling (transitive closure, etc.), handling language code aliases, pruning posets for display, traversing posets for display, etc.

Other sources

the etymologies themselves are in their heart of hearts confusing. eg "influenced by". eg heathen

we don't know the etymologies; they're lost to time, or we havent' studied them yet theres disagreement and uncertainty the etymologoise haven't been copied into wiktionary wiktionary has typos radix fails to use the info in wiktionary correctly gap between