So I’ve started charting how *ts gets rendered in my idolect(s), and in the process, I am learning new things about Colloquial Finnish I had never noticed.

The usual deal is that *ts in native vocabulary gets rendered as /ts/ only in Standard Finnish. The spoken dialects instead tend to have /tt/ or /ht/. Modern colloquial Finnish of most of southern and western Finland sides with /tt/ — while the southeastern Ostrobothnian savolaiskiila dialect, that my mother and her parents speak, also allows /ht/. So I’ve native familiarity with all three main variants. /ts/ however gets left as-is in all sorts of newer vocabulary that never went through the traditional dialects: learned vocabulary (hertsi, otsoni, satsuma), recent loans (koutsi ‘coach’, matsi ‘sports match’), Helsinki slang (funtsia ‘to think’, kartsa ‘street’) and general neologisms (itsari ‘suicide’, rantsu ‘beach’).

Factor in consonant gradation and additional interesting things happen. Normally, /tt/ and /ht/ that come from original *tt and *ht gradate to /t/ and /hd/ ~ /hr/ ~ /h/; but depending on the dialect, /tt/ and /ht/ that come from *ts may be exempt from this. There’s also a fifth pattern where *ts gives /ht/ : /t/.

(Here’s a map of the situation, including also rarer variants that I’m not sure are in use anymore today.)

I was expecting to find some variation in which words allow /ht/, versus which allow /tt/. No such thing seems to come up in substantial numbers however. Almost any word that allows /tt/ will also allow /ht/, and vice versa (metsä ‘forest, woods’ → mettä ~ mehtä; seitsemän ‘7’ → seittemän ~ seihtemän).

Interestingly enough though, there is some variation depending on the position of *ts in a word. Between a first and second syllable, I only find three nonstandard gradation patterns grammatical: /tt/ : /tt/, /tt/ : /t/ and /ht/ : /ht/ (mettän, metän, mehtän). But between a second and third syllable, things differ: in my heritage dialect a pattern /ht/ : /h/ now turns up (häirihten ~ häirihen ‘I bother’, for standard Fi. häiritsen), while in colloquial Finnish, the pattern /tt/ : /t/ disappears entirely (häiritten, but not ˣhäiriten). This is also followed by the contracted form tarvitten ‘I need’ (standard tarvitsen) → tartten, but not ˣtarten.

Word derivation also has pretty big effects. For example katsoa ‘to look’ is a typical basic-register word that allows the variants kattoa, kahtoa (katon, kahton, etc.) just fine. However, its derivative katse ‘sight, gaze’ is firmly in the literary register. I would deem ˣkatte, ˣkate, ˣkahte clearly ungrammatical (and I might have trouble even parsing the latter two).

This has also led me to notice an interesting unrelated pattern in verb inflection. Finnish verbs ending in -tse- are currently somewhat recessive: especially colloquially they tend to be replaced with corresponding basic vowel-stem or contraction verbs (with which they are already homophonous in the infinitive and a couple of other forms). So, for example tupakoida : tupakoitsen : tupakoitsee ‘to smoke tobacco’ (infinitive, 1PS, 3PS) tends towards the inflection tupakoin : tupakoi instead. Similarly tarvita : tarvitsen : tarvitsee ‘to need’ may also yield a third (!) inflection type in colloquial Finnish: tarviin, tarvii.

But this seems to fail in some cases. E.g. tulkita : tulkitsen ‘to interpret’ does not produce an expected colloquial form ˣtulkkiin; similarly merkitä : merkitsen ‘to mark down, note’, but not ˣmerkkiin. The key element seems to be consonant gradation: -tse-stem verbs have the key feature that the word root is always in the weak grade, while contraction verbs allow alternation (as in merkata : merkkaan ‘to mark’). The corresponding merkitä : ˣmerkkiin (with consonant gradation before -itA : -ii-) has, however, no precedents, and seems to therefore fail to materialise. Alternately though, gradation may be simply “forgotten”. While tarvita comes from earlier *tarbittak (compare the noun tarve : tarpeen ‘need’, or the adjective tarpeellinen ‘useful, necessary’), /rv/ is common enough also as an underlying basic consonant cluster, and hence the colloquial contraction-verb inflection comes out as tarviin and not ˣtarpiin.

How well do Google and Microsoft and recognize speech across dialect, gender and race?

If you’ve been following my blog for a while, you may remember that last year I found that YouTube’s automatic captions didn’t work as well for some dialects, or for women. The effects I found were pretty robust, but I wanted to replicate them for a couple of reasons:

I only looked at one system, YouTube’s automatic captions, and even that was over a period of several years instead of at just one…