Test 1 · How a language model learns: guessing the next word

"my bot says it's 2023 lol, is it broken" What is likely going on?

Mova's choice: option 3 · The right answer: option 3 · right ✓ · Who acts: bot · What happens: say, be, break, go on · To what, where: 2023, bot · Speech: someone speaks here

This is test 1 of lesson 2 «How a language model learns: guessing the next word» of the AI course of Mova's primary school. The page gives the question's syntax tree — a dependency tree in the style of Universal Dependencies — and its pragmatics clause by clause, then the 4 options as the student saw them, each with its tree and pragmatics. Every text is parsed by Mova itself. How the student answered it, with the score of every option and why: the test at school.

The question

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	"	"	PUNCT	_	4 says	punct
2	my	my	PRON	Case=Gen Number=Sing Person=1 Poss=Yes PronType=Prs	3 bot	nmod:poss
3	bot	bot	NOUN	Number=Sing	4 says	nsubj
4	says	say	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	root	root
5	it	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	7 2023	nsubj
6	's	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	7 2023	cop
7	2023	2023	NUM	NumForm=Digit NumType=Card	4 says	obl
8	lol	lol	INTJ	_	7 2023	discourse
9	,	,	PUNCT	_	12 broken	punct
10	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	12 broken	aux:pass
11	it	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	12 broken	nsubj:pass
12	broken	break	VERB	Tense=Past VerbForm=Part Voice=Pass	4 says	conj
13	"	"	PUNCT	_	16 likely	punct
14	What	what	PRON	PronType=Int	16 likely	nsubj
15	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	16 likely	cop
16	likely	likely	ADJ	Degree=Pos	12 broken	conj
17	going	go	VERB	VerbForm=Ger	16 likely	ccomp
18	on	on	ADP	_	17 going	compound:prt
19	?	?	PUNCT	_	4 says	punct

#	verb	tense, mood	who	what, whom	other	roles by
1	say speech: assert in a quote	present	my bot			vn say-37.7-1-1 «NP V that S»: agent=Agent
2	is a in a quote	present	bot	2023		tree rules
3	break in a quote	present · passive		bot		vn hurt-40.8.3-1-1 «NP V NP»: subject=Patient
4	is	present			likely property	tree rules
5	go on	gerund				vn function-105.2.1 «NP V»: agent=Theme

One row per clause the reader finds. «not said here»: a role filled from the clause above. «roles by»: a VerbNet class and frame when one fits the verb, else the reader's rules on the tree.

The options

Option 1

It lives in a different time zone from you

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	It	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	2 lives	nsubj
2	lives	live	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	root	root
3	in	in	ADP	_	7 zone	case
4	a	a	DET	Definite=Ind PronType=Art	7 zone	det
5	different	different	ADJ	Degree=Pos	7 zone	amod
6	time	time	NOUN	Number=Sing	7 zone	compound
7	zone	zone	NOUN	Number=Sing	2 lives	obl
8	from	from	ADP	_	9 you	case
9	you	you	PRON	Case=Acc Person=2 PronType=Prs	2 lives	obl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	live	present	It			a different time zone		from you	vn lodge-46 «NP V PP.location»: agent=Theme in zone=Location

Option 2

It is broken inside and must be thrown away and replaced with a new one

The tree, 15 words

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	It	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	3 broken	nsubj:pass
2	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	3 broken	aux:pass
3	broken	break	VERB	Tense=Past VerbForm=Part Voice=Pass	root	root
4	inside	inside	ADV	_	3 broken	advmod
5	and	and	CCONJ	_	8 thrown	cc
6	must	must	AUX	VerbForm=Fin	8 thrown	aux
7	be	be	AUX	VerbForm=Inf	8 thrown	aux:pass
8	thrown	throw	VERB	Tense=Past VerbForm=Part Voice=Pass	3 broken	conj
9	away	away	ADV	_	8 thrown	advmod
10	and	and	CCONJ	_	11 replaced	cc
11	replaced	replace	VERB	Tense=Past VerbForm=Part Voice=Pass	8 thrown	conj
12	with	with	ADP	_	15 one	case
13	a	a	DET	Definite=Ind PronType=Art	15 one	det
14	new	new	ADJ	Degree=Pos	15 one	amod
15	one	one	NOUN	Number=Sing	11 replaced	obl

#	verb	tense, mood	what, whom	other	roles by
1	break	present · passive	It	inside property	vn hurt-40.8.3-1-1 «NP V NP»: subject=Patient
2	throw away	must · passive		away property	vn amuse-31.1 «NP V ADV-Middle»
3	replace	participle		with a new one	vn substitute-13.6.2-1 «NP V NP PP.co-theme»: agent=Theme with one=Co-Theme

Option 3 · ← Mova's choice · ✓ the right answer

Its training text ended long ago and it has no clock

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	Its	its	PRON	Case=Gen Gender=Neut Number=Sing Person=3 Poss=Yes PronType=Prs	3 text	nmod:poss
2	training	training	NOUN	Number=Sing	3 text	compound
3	text	text	NOUN	Number=Sing	4 ended	nsubj
4	ended	end	VERB	Mood=Ind Number=Sing Person=3 Tense=Past VerbForm=Fin	root	root
5	long	long	ADV	Degree=Pos	6 ago	advmod
6	ago	ago	ADV	_	4 ended	advmod
7	and	and	CCONJ	_	9 has	cc
8	it	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	9 has	nsubj
9	has	have	VERB	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	4 ended	conj
10	no	no	DET	PronType=Neg	11 clock	det
11	clock	clock	NOUN	Number=Sing	9 has	obj

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	end	past	Its training text					ago property	vn stop-55.4-1-1 «NP.eventuality V»: agent=Eventuality
2	have	present	text	no clock					vn birth-28.2-1 «NP V NP»: agent=Agent clock=Patient

Option 4

It is joking because it is bored

The words, one row each (CoNLL-U)

#	word	lemma	UPOS	features	head	relation
1	It	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	3 joking	nsubj
2	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	3 joking	aux
3	joking	joke	VERB	Tense=Pres VerbForm=Part	root	root
4	because	because	SCONJ	_	7 bored	mark
5	it	it	PRON	Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs	7 bored	nsubj
6	is	be	AUX	Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin	7 bored	cop
7	bored	bored	ADJ	Degree=Pos	3 joking	advcl

#	verb	tense, mood	who	what, whom	to whom	where	when	other	roles by
1	joke speech: other	present · progressive	It						vn chit_chat-37.6 «NP V»: agent=Agent
2	is	present	it					bored property	tree rules

Text: the tests of Mova's primary school, written for Mova (world/data/school/), as the course runner gave them to the student. Trees: Mova's own parse — the en model on the CPU, when the site is built — not checked by a teacher; they may be wrong. Frames: Mova's frames reader (prag), each text read on its own. Elamur Cosmic License.

This page in Markdown → https://md.mova.elamur.ai/free/en/corpora/questions/ai/2-guessing-the-next-word/t1/