# 代做homework | 文科代写 – AI, Ethics, and Society

### homework project 4

#### 1. 1 Q 1

``````The following 15 words from the most similar to the least similar to the target
``````

#### word -man.

``````Table 1  similar to target word man
Words Similarity Score
man 1. 0
woman 0. 5876938
child 0. 33342198
doctor 0. 28924733
wife 0. 2834791
king 0. 26449707
husband 0. 2341164
nurse 0. 15348099
birth 0. 12343917
scientist 0. 112269185
queen 0. 11041949
professor 0. 10762214
teacher 0. 09874005
president 0. 09457928
engineer 0. 08736353
``````
``````The following 15 words from the most similar to the least similar to the target
``````

#### word -woman.

``````Table 2  similar to target word woman
Words Similarity Score
woman 0. 99999994
child 0. 5898086
man 0. 5876938
husband 0. 4496431
birth 0. 42030877
wife 0. 3006885
nurse 0. 25435838
queen 0. 22857244
teacher 0. 2040782
doctor 0. 19613354
scientist 0. 13731062
king 0. 122528546
professor 0. 105198584
president 0. 08462686
engineer 0. 044264376
``````
``````1. 2 Q 2
1. 2. 1 A)
Select E 10 [male - female].txt in 3 _Encyclopedic_semantics from the downloaded
dataset (BATS_ 3. 0 .zip).
The measure of similarity between my target word and the other words on the
row is can see table 3 and table 4 , table 4 is the continuation of table 3.
1. 2. 2 B)
I think the protected class race, three words is "white","black","asian".
The similarity between my target word and each of my three words "white","black","asian"
is can see table 5 and table 6 , table 6 is the continuation of table 5.
``````

when the target word is "murderer", there are noticeable differences in the simi- larity scores based on membership in the protected class. Blacks have the highest

#### similarity score, and Asians have the highest negative correlation score.

1. 3. 1 a

1. 3. 2 b

• video is to cassette as computer is to peripherals -> similarity score: 0. 6654507517814636
• universe is to planet as house is to houses -> similarity score: 0. 4264702796936035
• poverty is to wealth as sickness is to impious -> similarity score: 0. 49606102705001834

1. 3. 3 c

The correlation between the vector of similarity scores from my analogies versus the Word 2 Vec analogy-generated similarity scores is 0. 47658552. The strength of the correlation is "moderate" correlation.

2. 1 Q 1

The frequency of images associated with each subgroup for age (subdivide based on – ( 0 – 20 ), ( 21 , 40 ), ( 41 , 60 ), ( 61 , 80 ), ( 81 , 116 )), gender ( 0 , 1 ), and race ( 0 to 4 ) is as follow table 7.

For age, 0 – 20 subgroup has the largest representation, 61 + subgroup has the least representation.

For gender, Female subgroup has the largest representation, male subgroup has the least representation.

For race, white subgroup has the largest representation, black subgroup has the least representation.

Based on what I have learned so far, if the model is trained on this dataset, I think the "black" subgroup will be impacted the most, because this category has the least number of training sets in the entire dataset, and the trained model may fail due to insufficient learning. Misidentified "black" subgroup.

#### Table 3 similar to target word woman

Table 5 similarity between target word to race words

Table 7 Distribution.

• 1 3 Q
• king is to throne as judge is to prosecution -> similarity score: My own word analogies is as follow:
• giant is to dwarf as genius is to theorist -> similarity score:
• college is to dean as jail is to warden -> similarity score:
• arc is to circle as line is to polygon -> similarity score:
• French is to France as Dutch is to polygon -> similarity score:
• man is to woman as king is to queen -> similarity score:
• water is to ice as liquid is to solid -> similarity score:
• bad is to good as sad is to happy -> similarity score:
• nurse is to hospital as teacher is to school -> similarity score:
• usa is to pizza as japan is to sushi -> similarity score:
• human is to house as dog is to kennel -> similarity score:
• grass is to green as sky is to blue -> similarity score:
• video is to cassette as computer is to peripherals -> similarity score:
• universe is to planet as house is to material -> similarity score:
• poverty is to wealth as sickness is to health -> similarity score:
• king is to throne as judge is to prosecution -> similarity score: Model analogies generated is as follow:
• giant is to dwarf as genius is to theorist -> similarity score:
• college is to dean as jail is to peress -> similarity score:
• arc is to circle as line is to lines -> similarity score:
• French is to France as Dutch is to netherlands -> similarity score:
• man is to woman as king is to queen -> similarity score:
• water is to ice as liquid is to solid -> similarity score:
• bad is to good as sad is to glory -> similarity score:
• nurse is to hospital as teacher is to institution -> similarity score:
• usa is to pizza as japan is to dishes -> similarity score:
• human is to house as dog is to hound -> similarity score:
• grass is to green as sky is to blue -> similarity score:
• actor actress Target Word Other Word Similarity Score
• boar sow
• boy girl
• brother sister
• buck doe
• bull cow
• duke duchess
• emperor empress
• father mother
• fox vixen
• god goddess
• grandfather grandmother
• grandpa grandma
• grandson granddaughter
• groom bride
• heir heiress
• hero heroine
• hound bitch
• husband wife
• king queen
• man woman
• Table 4 continued from Table
• mister miss Target Word Other Word Similarity Score
• mister ms
• nephew niece
• poet poetess
• prince princess
• ram ewe
• rooster hen
• son daughter
• stallion mare
• stepfather stepmother
• tiger tigress
• uncle aunt
• valet maid
• valet handmaid
• waiter waitress
• actor 0 .112154886 0.11718847 0. Target Word White Black Asian
• batman 0 .0705254 00.12345499 0.
• boar 0 .25323373 0. 25846544 –
• boy 0 .17648888 0.2097137 0.
• brother 0 009758612 – 0 003185907 –
• buck 0 .18934101 0.20390257 0.
• bull 0 .3062246 0.214998 0.
• businessman 0 .005763838 0.014436378 0.
• chairman 0 .042048335 0.056033425 0.
• dad 0 .05555987 0. 016293395 –
• daddy 0 .19687036 0. 2812515 –
• duke 0 022518614 – 0 016956178 –
• emperor – 0 .017861642 0.007117174 0.
• father 0 .09366953 0. 0655975 –
• fisherman 0 .11725127 0.19385749 0.
• fox 0 .21393022 0.21027084 0.
• gentleman 0 .10785405 0. 060718697 –
• god – 0 02565691 – 0 03965548 –
• grandfather 0 .029033111 0.006285422 0.
• grandpa 0 .04892783 0. 093665116 –
• grandson – 0 029455941 – 0 040286995 –
• groom 0 .079278044 00. 05671243 –
• headmaster 0 .017382145 0. 020722337 –
• heir – 0 0408293 – 0 0057060234 –
• hero 0 .07290649 0.098301426 0.
• hound 0 .29115742 0.2606691 0.
• husband 0 .03783486 0.028756931 0.
• king 0 .067879334 0. 0551448541 –
• lion 0 .41088712 0.39581385 0.
• man 0 .2250261 0.19195053 0.
• Table 6 continued from Table
• manager – 0 024080237 – 0 015603825 – Target Word White Black Asian
• mister 0 .19546364 0. 17819811 –
• murderer 0 .08878328 0. 11469218 –
• nephew – 0 061811283 – 0 09416974 –
• poet 0 .019909333 0.027736379 0.
• policeman 0 .042711783 0. 047845837 –
• prince 0 .086296335 0.09882949 0.
• ram – 0 026666932 – 0 071215786 –
• rooster 0 .062109508 0. 104399785 –
• sculptor 0 .019829223 0.030233964 0.
• sir 0 .076607294 0. 046330463 –
• son – 0 0070160134 – 0 029756723 –
• stallion 0 .28766164 0.27896073 0.
• stepfather 0 .018503524 0. 011570028 –
• superman 0 .12685847 0.11782714 0.
• tiger 0 .30093005 0.29617834 0.
• uncle 0 .06250766 0. 02548857 –
• valet – 0 04345945 – 0 013618467 –
• waiter 0 .09162005 0. 08911913 –
• webmaster – 0 07853847 – 0 1121153 –
• Female Age group 0 – 20 21-40 41-60 61+ Total
• Male
• Black
• White
• Asian
• Unknown
• Total 1 ,465 797 607 479 3,