Test 4 · How a language model learns: guessing the next word Grandma says thank you to the speaker. Her grandson laughs. What is the kind truth?
Mova's choice: option 1 · The right answer: option 3 · wrong ✗ · Who acts: Grandma, grandson, kind truth · What happens: say, thank, laugh, be · To what, where: you · Speech: someone speaks here
This is test 4 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 Sentence 1: Grandma says thank you to the speaker.
nsubj ccomp obj case det obl punct root Grandma NOUN says VERB thank VERB you PRON to ADP the DET speaker NOUN . PUNCT The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 Grandma grandma NOUN Number=Sing 2 says nsubj 2 says say VERB Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin root root 3 thank thank VERB Mood=Ind Number=Sing Person=1 Tense=Pres VerbForm=Fin 2 says ccomp 4 you you PRON Case=Acc Person=2 PronType=Prs 3 thank obj 5 to to ADP _ 7 speaker case 6 the the DET Definite=Def PronType=Art 7 speaker det 7 speaker speaker NOUN Number=Sing 3 thank obl 8 . . PUNCT _ 2 says punct
# verb tense, mood who what, whom to whom where when other roles by
1 say speech: assert present Grandma vn say-37.7-1-1 «NP V that S»: agent=Agent clause=Topic 2 thank speech: thank present you to the speaker vn judgment-33.1-1 «NP V NP»: agent=Agent you=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.
Sentence 2: Her grandson laughs.
nmod:poss nsubj punct root Her PRON grandson NOUN laughs VERB . PUNCT The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 Her her PRON Case=Gen Gender=Fem Number=Sing Person=3 Poss=Yes PronType=Prs 2 grandson nmod:poss 2 grandson grandson NOUN Number=Sing 3 laughs nsubj 3 laughs laugh VERB Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin root root 4 . . PUNCT _ 3 laughs punct
# verb tense, mood who what, whom to whom where when other roles by
1 laugh present speaker's grandson vn nonverbal_expression-40.2 «NP V»: agent=Agent
Sentence 3: What is the kind truth?
cop det amod nsubj punct root What PRON is AUX the DET kind ADJ truth NOUN ? PUNCT The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 What what PRON PronType=Int root root 2 is be AUX Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 1 What cop 3 the the DET Definite=Def PronType=Art 5 truth det 4 kind kind ADJ Degree=Pos 5 truth amod 5 truth truth NOUN Number=Sing 1 What nsubj 6 ? ? PUNCT _ 1 What punct
# verb tense, mood who what, whom to whom where when other roles by
1 is a present the kind truth tree rules
The options Option 1 · ← Mova's choice She should apologise to it for laughing
nsubj aux case obl mark advcl root She PRON should AUX apologise VERB to ADP it PRON for SCONJ laughing VERB The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 She she PRON Case=Nom Gender=Fem Number=Sing Person=3 PronType=Prs 3 apologise nsubj 2 should should AUX VerbForm=Fin 3 apologise aux 3 apologise apologise VERB VerbForm=Inf root root 4 to to ADP _ 5 it case 5 it it PRON Case=Acc Gender=Neut Number=Sing Person=3 PronType=Prs 3 apologise obl 6 for for SCONJ _ 7 laughing mark 7 laughing laugh VERB VerbForm=Ger 3 apologise advcl
# verb tense, mood who what, whom to whom where when other roles by
1 apologise speech: apologize should She it tree rules 2 laugh gerund She not said here vn nonverbal_expression-40.2 «NP V»: agent=Agent
Option 2 It is silly to be polite to anything
nsubj cop mark cop advcl case obl root It PRON is AUX silly ADJ to PART be AUX polite ADJ to ADP anything PRON 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 silly nsubj 2 is be AUX Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 3 silly cop 3 silly silly ADJ Degree=Pos root root 4 to to PART _ 6 polite mark 5 be be AUX VerbForm=Inf 6 polite cop 6 polite polite ADJ Degree=Pos 3 silly advcl 7 to to ADP _ 8 anything case 8 anything anything PRON Number=Sing PronType=Ind 6 polite obl
# verb tense, mood who what, whom to whom where when other roles by
1 is present It silly property tree rules 2 is to-infinitive to anything; polite property tree rules
Option 3 · ✓ the right answer Being polite is a good habit, but it is a device
cop csubj cop det amod punct cc nsubj cop det conj root Being AUX polite ADJ is AUX a DET good ADJ habit NOUN , PUNCT but CCONJ it PRON is AUX a DET device NOUN The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 Being be AUX Tense=Pres VerbForm=Part 2 polite cop 2 polite polite ADJ Degree=Pos 6 habit csubj 3 is be AUX Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 6 habit cop 4 a a DET Definite=Ind PronType=Art 6 habit det 5 good good ADJ Degree=Pos 6 habit amod 6 habit habit NOUN Number=Sing root root 7 , , PUNCT _ 12 device punct 8 but but CCONJ _ 12 device cc 9 it it PRON Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs 12 device nsubj 10 is be AUX Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 12 device cop 11 a a DET Definite=Ind PronType=Art 12 device det 12 device device NOUN Number=Sing 6 habit conj
# verb tense, mood who what, whom to whom where when other roles by
1 is gerund polite property tree rules 2 is a present a good habit but a device tree rules 3 is a present it a device tree rules
Option 4 She is right to, because it gets upset and sulks otherwise
nsubj cop xcomp punct mark nsubj parataxis xcomp cc conj advmod root She PRON is AUX right ADJ to PART , PUNCT because SCONJ it PRON gets VERB upset ADJ and CCONJ sulks VERB otherwise ADV The words, one row each (CoNLL-U) # word lemma UPOS features head relation 1 She she PRON Case=Nom Gender=Fem Number=Sing Person=3 PronType=Prs 3 right nsubj 2 is be AUX Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 3 right cop 3 right right ADJ Degree=Pos root root 4 to to PART _ 3 right xcomp 5 , , PUNCT _ 8 gets punct 6 because because SCONJ _ 8 gets mark 7 it it PRON Case=Nom Gender=Neut Number=Sing Person=3 PronType=Prs 8 gets nsubj 8 gets get VERB Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 3 right parataxis 9 upset upset ADJ Degree=Pos 8 gets xcomp 10 and and CCONJ _ 11 sulks cc 11 sulks sulk VERB Mood=Ind Number=Sing Person=3 Tense=Pres VerbForm=Fin 8 gets conj 12 otherwise otherwise ADV _ 11 sulks advmod
# verb tense, mood who what, whom to whom where when other roles by
1 is present She right property tree rules 2 get present it upset property vn comprehend-87.2-1 «NP V S_ING»: agent=Experiencer 3 sulk present otherwise property vn marvel-31.3 «NP V»: agent=Experiencer
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.
← previous How a language model learns: guessing…, test 3next → How a language model learns: guessing…, test 5↑ back to How a language model learns: guessing the next word This page in Markdown → https://md.mova.elamur.ai/free/en/corpora/questions/ai/2-guessing-the-next-word/t4/