Analysis of the restricted corpus

[vcvpaiva]

According to stats (https://github.com/own-pt/rte-sick/blob/master/expanded/conllu/len.3-5.conllu.stats.xml) 390 sentences (385 nsubj+5nsubjpass), 1 conjunction only, 4 dependencies 'dep', 13 copulas, 140 direct objects, 26 expletives, 19 negations and 20 noun-nouns.

1 prep and 9 particles.

The rest is: "acomp"=2 "advmod"=34 "amod"= 20 "partmod"= 10 "pobj"= 1

also 30 ADJ, 45 ADV "quantmod"= 3

[vcvpaiva]

only conjunction is:

Paper and scissors both cut

scissors has lemmatization problem, no concept found. paper is wrong concept, newspaper. cut is wrong concept, process only.

want to distribute? enhanced CoreNLP does.

[vcvpaiva]

13 copulas: (one can check https://github.com/own-pt/rte-sick/blob/master/expanded/conllu/len.3-5.conllu)

1. Some people are silent
2. The child is silent
3. A classroom is empty
4. A man is motionless
5. A man is scared
6. The fish are immobile
7. A man is silent
8. A biker is naked
9. The man is training ****
10. Some kittens are hungry
11. The man is rock climbing ****
12. A woman is grating carrots ****
13. The woman is dicing garlic ****

4 copulas are wrong as far as dependencies are concerned. others are wrong as far as the mapping is concerned, later.

[vcvpaiva]

19 negations marked:

1. A fish is not swimming
2. Two men are not fighting
3. The men are not dancing
4. A person is not frying
5. People are not playing cricket
6. The men are not talking
7. The band isn't singing (noticed n't working fine!)
8. The person is not drawing
9. A man is not dancing
10. The man is not dancing
11. The man is not exercising
12. The woman is not waterskiing
13. The woman is not dancing
14. Someone is not playing piano
15. Some women are not talking
16. A horse is not racing
17. Two women are not dancing
18. A jet is not flying
19. The man is not drawing

all dependencies seem fine.

However there are several semantic negations that are not marked.

1. No person is hiking
2. There are no men fighting
3. There is no man speaking
4. There is no man drawing
5. There is no dog barking
6. There is no person writing
7. There is no man spitting
8. There is no band playing
9. There is no parrot speaking
10. There is no man exercising
11. There is no baby talking
12. There is no man mixing
13. There is no man dancing
14. There is no panda climbing
15. There is no lion walking (amb?)
16. There is no hamster singing
17. There is no lemur eating (amb)
18. There is no woman exercising
19. There is no clown singing
20. There is no one typing
21. There is no woman dancing
22. There is no puppy rolling
23. There is no man praying
24. There is no man screaming
25. There are no mimes performing
26. There are no men sawing
27. There is no man thinking

and also:

1. Nobody is riding a bike
2. Nobody is playing ping pong
3. Nobody is holding a hedgehog
4. Nobody is feeding an animal
5. Nobody is slicing a tomato
6. Nobody is brushing a cat
7. Nobody is beating an egg
8. Nobody is playing the guitar
9. Nobody is riding a bike
10. Nobody is playing ping pong

[vcvpaiva]

Looking for the 20 nns is difficult, as the reps use NN for noun. However, the first 7 noun-nouns below seem correct. and to give them concepts, they need to be considered compounds.

1. ping pong
2. golden retriever
3. sumo wrestlers
4. tiger cub
5. sumo ringers
6. baby pandas 2x
7. cartoon airplane

Wrong ones:

1. men sawing (nn)
2. man screaming (dep)
3. puppy rolling
4. woman dancing
5. clown singing
6. hamster singing
7. lion walking
8. panda climbing
9. man dancing
10. rock climbing
11. baby talking
12. band playing
13. dog barking
14. blonds girl (typo)

[vcvpaiva]

there are 4 dependencies "dep"

1. There is no man speaking (speaking)
2. Paper and scissors both cut (both)
3. A kitten is getting bored (getting)
4. There is no man screaming (screaming)

[vcvpaiva]

"acomp"=2 ===> falling asleep. "quantmod"===> 3 "a few" "amod"= 20 ===> correct ones: old woman, young female, young kids, cold cyclist, happy baby, young girl, white horse, sad man, hungry woman, small animal

wrong ones: a few kittens, few men, several children, lemur eating, one typing, grating carrots, dicing garlic, golden retriever, animated airplane, man thinking,

[vcvpaiva]

particle verbs:

1. take off 3
2. laying down
3. sitting down
4. waking up 2
5. standing up 2 Note that this particles do mess up the concepts assignment, e.g.

# text = A toddler is standing up
1   A   a   DET DT  _   2   det _   DT|?|?
2   toddler toddler NOUN    NN  _   4   nsubj   _   NN|10714465-n|NonFullyFormed+
3   is  be  VERB    VBZ _   4   aux _   VBZ|02604760-v|Entity+
4   standing    stand   VERB    VBG _   0   ROOT    _   VBG|01546111-v|PhysicalAttribute+
5   up  up  PRT RP  _   4   prt _   RP|00097011-r|Increasing+

only one preposition: Someone is on a horse

[vcvpaiva]

expletives

1. There are no men fighting (wrong root)
2. There is no man speaking
3. There is no man drawing
4. There is no dog barking
5. There is no person writing
6. There is no man spitting
7. There is no band playing

[vcvpaiva]

5 passive voices only, all correct.

1. A boat is anchored
2. A pencil is being sharpened
3. An onion is being chopped
4. An onion is being sliced
5. Some paper is being cut

[vcvpaiva]

advmodifiers: (34 in total)

1. quickly 2
2. happily 2
3. frantically
4. gracefully 2
5. slowly
6. wildly
7. noisily
8. passionately
9. cheerfully
10. fervently
11. loudly
12. fearlessly

Others:

1. together
2. still
3. outside
4. inside
5. indoors
6. around
7. alone
8. downhill
9. uphill

[vcvpaiva]

actually in the whole corpus we have 112 adverbs ending in "ly" -- this is for future use in the nomlex for adjective/adverbs. they're attached. sadverbs.txt

[vcvpaiva]

I do not understand why only the verb "sit" gives a different meaning to the auxiliary "to be". Sentences: A person is sitting, A man is sitting, A baby is sitting, A tiger is sitting, A toddler is sitting, the man is sitting indoors-- all give "is" as PhysicalAttribute instead of Entity+. Yes, "sit" is an state, not an action, but so is "standing".

[arademaker]

@vcvpaiva once more, so answer this puzzle we need an interactive implementation of UKB with a traces! We would need to see how the algorithm evolves during the computation to choose a sense A instead of a sense B for a given word W.

[vcvpaiva]

1 occ of man mapping to Man, 111 occs mapping to Hominid? 140 occs of woman mapping to Woman 9 Guitars, 11 Canines 32 playing ==> DramaticActing, 32 dancing ==> Dancing+ 24 Separating, from slice, dismantle and tear

[vcvpaiva]

17 correct adjectives: empty, hungry, immobile naked, motionless, silent scared, animated, bored happy, sad golden, white young, old small asleep

some of the adjectives are past participles: scared, animated, bored which causes problems.

[vcvpaiva]

only 50 nsubjs, kind of:

1. man
2. male
3. boy
4. woman
5. lady
6. girl
7. female
8. child
9. kid
10. toddler

11. baby

12. clown
13. cheerleader
14. speaker
15. biker
16. ringer (sumo)
17. wrestler
18. mime
19. cyclist
20. band
21. team
22. someone
23. nobody/ no one
24. people

25. person

26. fish
27. kitten
28. cat
29. dog
30. golden retriever
31. lemur
32. cow
33. panda
34. tiger
35. lion
36. horse
37. parrot
38. cub
39. hamster
40. squirrel

41. animal

42. classroom (place, not the people)
43. plane
44. airplane
45. jet
46. boat
47. pencil
48. keyboard
49. barbells
50. paper
51. onion

[vcvpaiva]

dobjs:

1. play cricket/soccer/rugby
2. receive volleyball
3. play piano/keyboard/guitar/flute
4. strum guitar
5. cut wood
6. cut shrimps/tomatoes/garlic/potatoes/butter
7. slice garlic/butter/potatoes/herbs/tomato/bread/tofu
8. chop onion/garlic
9. dice onion
10. mix eggs
11. beat eggs
12. dry noodles/eggs
13. saw logs
14. likes parrots
15. cook prawns/eggs
16. boil shrimps/eggs/noodles
17. peel banana/food
18. store broccoli
19. break eggs/
20. crack eggs
21. tear sheets
22. hold a hedgehog
23. cleanse animal
24. brush a cat
25. dirty animal
26. feed animal
27. spread dough
28. lump dough
29. eat cereal/grass
30. amalgamate eggs
31. lift barbells
32. mow grass
33. read email