Dataset Viewer
Auto-converted to Parquet
id
stringlengths
4
7
context_1
stringlengths
10
129
context_2
stringlengths
10
187
target_word
stringlengths
2
13
pos
stringclasses
2 values
target_word_location_1
dict
target_word_location_2
dict
language
stringclasses
1 value
label
int64
0
1
EN_10
It is important to use flux when soldering or oxides on the metal will prevent a good bond.
His opinions are in flux.
flux
N
{ "char_end": 27, "char_start": 23 }
{ "char_end": 24, "char_start": 20 }
EN
0
EN_21
The solution took three hours.
They were trying to find a peaceful solution.
solution
N
{ "char_end": 12, "char_start": 4 }
{ "char_end": 44, "char_start": 36 }
EN
0
EN_22
They stopped at an open space in the jungle.
The astronauts walked in outer space without a tether.
space
N
{ "char_end": 29, "char_start": 24 }
{ "char_end": 36, "char_start": 31 }
EN
0
EN_42
Settling a score.
The player with the highest score is the winner.
score
N
{ "char_end": 16, "char_start": 11 }
{ "char_end": 33, "char_start": 28 }
EN
0
EN_51
The astronauts walked in outer space without a tether.
The space between his teeth.
space
N
{ "char_end": 36, "char_start": 31 }
{ "char_end": 9, "char_start": 4 }
EN
0
EN_63
A unit of wheat is a bushel.
He was a member of a special police unit.
unit
N
{ "char_end": 6, "char_start": 2 }
{ "char_end": 40, "char_start": 36 }
EN
0
EN_70
He predicted that although it is of small account now it will rapidly increase in importance.
The account of his speech that was given on the evening news made the governor furious.
account
N
{ "char_end": 49, "char_start": 42 }
{ "char_end": 11, "char_start": 4 }
EN
0
EN_76
The course of the river.
He took a course in basket weaving.
course
N
{ "char_end": 10, "char_start": 4 }
{ "char_end": 16, "char_start": 10 }
EN
0
EN_78
He wanted to live his own life without interference from others.
She's my love, my life.
life
N
{ "char_end": 30, "char_start": 26 }
{ "char_end": 22, "char_start": 18 }
EN
0
EN_80
It takes time to learn the setup around here.
The laboratory included an elaborate setup for measuring the energy.
setup
N
{ "char_end": 32, "char_start": 27 }
{ "char_end": 42, "char_start": 37 }
EN
0
EN_87
The pursuit of love.
Unremitting pursuit of wealth doesn't bring happiness, particularly if successful.
pursuit
N
{ "char_end": 11, "char_start": 4 }
{ "char_end": 19, "char_start": 12 }
EN
0
EN_115
He wanted to feel a part of something bigger than himself.
She played the part of Desdemona.
part
N
{ "char_end": 24, "char_start": 20 }
{ "char_end": 19, "char_start": 15 }
EN
0
EN_125
What do we have here?
I have two years left.
have
V
{ "char_end": 15, "char_start": 11 }
{ "char_end": 6, "char_start": 2 }
EN
0
EN_133
The teacher asked how many people made up his home.
Deliver the package to my home.
home
N
{ "char_end": 50, "char_start": 46 }
{ "char_end": 30, "char_start": 26 }
EN
0
EN_140
They stopped at an open space in the jungle.
They tested his ability to locate objects in space.
space
N
{ "char_end": 29, "char_start": 24 }
{ "char_end": 50, "char_start": 45 }
EN
0
EN_151
England scored a goal at the death to even the score at one all.
It was the death of all his plans.
death
N
{ "char_end": 34, "char_start": 29 }
{ "char_end": 16, "char_start": 11 }
EN
0
EN_160
The life of this milk carton may be thousands of years in this landfill.
There is no life on the moon.
life
N
{ "char_end": 8, "char_start": 4 }
{ "char_end": 16, "char_start": 12 }
EN
0
EN_170
He was unable to make his motion because he couldn't get recognition by the chairman.
Increasing recognition that diabetes frequently coexists with other chronic diseases.
recognition
N
{ "char_end": 68, "char_start": 57 }
{ "char_end": 22, "char_start": 11 }
EN
0
EN_173
The formation of sentences.
A defensive formation.
formation
N
{ "char_end": 13, "char_start": 4 }
{ "char_end": 21, "char_start": 12 }
EN
0
EN_194
He sailed around the world.
All the world loves a lover.
world
N
{ "char_end": 26, "char_start": 21 }
{ "char_end": 13, "char_start": 8 }
EN
0
EN_195
A life of luxury and ease.
He passed all the exams with ease.
ease
N
{ "char_end": 25, "char_start": 21 }
{ "char_end": 33, "char_start": 29 }
EN
0
EN_199
They agreed about the difficulty of the climb.
Finished the test only with great difficulty.
difficulty
N
{ "char_end": 32, "char_start": 22 }
{ "char_end": 44, "char_start": 34 }
EN
0
EN_200
Pottery was his life.
He appointed himself emperor for life.
life
N
{ "char_end": 20, "char_start": 16 }
{ "char_end": 37, "char_start": 33 }
EN
0
EN_203
The general term of an algebraic equation of the n-th degree.
He was sentenced to a term of six years in prison.
term
N
{ "char_end": 16, "char_start": 12 }
{ "char_end": 26, "char_start": 22 }
EN
0
EN_211
In a static system oil can not be replaced by water on a surface.
A vast system of production and distribution and consumption keep the country going.
system
N
{ "char_end": 18, "char_start": 12 }
{ "char_end": 13, "char_start": 7 }
EN
0
EN_212
We are in accord with your proposal.
The accord of light and shade in painting.
accord
N
{ "char_end": 16, "char_start": 10 }
{ "char_end": 10, "char_start": 4 }
EN
0
EN_218
He seemed more content in death than he had ever been in life.
Her death came as a terrible shock.
death
N
{ "char_end": 31, "char_start": 26 }
{ "char_end": 9, "char_start": 4 }
EN
0
EN_220
A rifle action.
Gawkers always try to get as close to the action as possible.
action
N
{ "char_end": 14, "char_start": 8 }
{ "char_end": 48, "char_start": 42 }
EN
0
EN_229
The professor didn't like his friends to use his formal title.
He held the title for two years.
title
N
{ "char_end": 61, "char_start": 56 }
{ "char_end": 17, "char_start": 12 }
EN
0
EN_248
Every number has a unique position in the sequence.
He has an unlisted number.
number
N
{ "char_end": 12, "char_start": 6 }
{ "char_end": 25, "char_start": 19 }
EN
0
EN_254
Victorian furniture is an interest of mine.
In the common interest.
interest
N
{ "char_end": 34, "char_start": 26 }
{ "char_end": 22, "char_start": 14 }
EN
0
EN_285
He gave us a general formula for attacking polynomials.
X = \frac -b \pm \sqrtb^2-4ac2a is a formula for finding the roots of the quadratic equation ax2 + bx + c = 0.
formula
N
{ "char_end": 28, "char_start": 21 }
{ "char_end": 44, "char_start": 37 }
EN
0
EN_291
I'll save you the trouble.
This move will save money.
save
V
{ "char_end": 9, "char_start": 5 }
{ "char_end": 19, "char_start": 15 }
EN
0
EN_313
I just don't get him.
Where did she get these news?
get
V
{ "char_end": 16, "char_start": 13 }
{ "char_end": 17, "char_start": 14 }
EN
0
EN_321
The connection between overeating and obesity.
There was a connection between eating that pickle and having that nightmare.
connection
N
{ "char_end": 14, "char_start": 4 }
{ "char_end": 22, "char_start": 12 }
EN
0
EN_324
He threw a flare to the fullback who was tackled for a loss.
Infection can cause a lupus flare.
flare
N
{ "char_end": 16, "char_start": 11 }
{ "char_end": 33, "char_start": 28 }
EN
0
EN_326
I went into the salon to get a clip.
He could do ten at a clip.
clip
N
{ "char_end": 35, "char_start": 31 }
{ "char_end": 25, "char_start": 21 }
EN
0
EN_327
The rule of St. Dominic.
He determined the upper bound with Descartes' rule of signs.
rule
N
{ "char_end": 8, "char_start": 4 }
{ "char_end": 50, "char_start": 46 }
EN
0
EN_340
A man of action.
He saw action in Korea.
action
N
{ "char_end": 15, "char_start": 9 }
{ "char_end": 13, "char_start": 7 }
EN
0
EN_358
He was elected to a seat in the Senate.
He sat on the arm of the chair rather than the seat, which always annoyed his mother.
seat
N
{ "char_end": 24, "char_start": 20 }
{ "char_end": 51, "char_start": 47 }
EN
0
EN_364
The conflict between the government and the rebels began three years ago.
He was immobilized by conflict and indecision.
conflict
N
{ "char_end": 12, "char_start": 4 }
{ "char_end": 30, "char_start": 22 }
EN
0
EN_380
He tried to blame the victim but his success on that score was doubtful.
What was your score on your homework?
score
N
{ "char_end": 58, "char_start": 53 }
{ "char_end": 19, "char_start": 14 }
EN
0
EN_388
I paid with a twenty and pocketed the change.
The product is undergoing a change in order to improve it.
change
N
{ "char_end": 44, "char_start": 38 }
{ "char_end": 34, "char_start": 28 }
EN
0
EN_423
We'll get them!
She didn't get the joke.
get
V
{ "char_end": 9, "char_start": 6 }
{ "char_end": 14, "char_start": 11 }
EN
0
EN_424
I expected a brief account.
A person of considerable account.
account
N
{ "char_end": 26, "char_start": 19 }
{ "char_end": 32, "char_start": 25 }
EN
0
EN_440
A military buildup in preparation for the invasion.
The buildup of leaves blocked the drain pipes.
buildup
N
{ "char_end": 18, "char_start": 11 }
{ "char_end": 11, "char_start": 4 }
EN
0
EN_442
In a small number of horses, muscle weakness may progress to paralysis.
The weakness of the span was overlooked until it collapsed.
weakness
N
{ "char_end": 44, "char_start": 36 }
{ "char_end": 12, "char_start": 4 }
EN
0
EN_447
The number of parameters is small.
In English the subject and the verb must agree in number.
number
N
{ "char_end": 10, "char_start": 4 }
{ "char_end": 56, "char_start": 50 }
EN
0
EN_460
I can't get these boots off (or on).
She didn't get the joke.
get
V
{ "char_end": 11, "char_start": 8 }
{ "char_end": 14, "char_start": 11 }
EN
0
EN_461
His novels always have a lot of action.
The United Nations must have the power to propose and organize action without being hobbled by irrelevant issues.
action
N
{ "char_end": 38, "char_start": 32 }
{ "char_end": 69, "char_start": 63 }
EN
0
EN_463
Our sense of sight enables us to locate objects in space.
Can you locate your cousins in the Midwest?
locate
V
{ "char_end": 39, "char_start": 33 }
{ "char_end": 14, "char_start": 8 }
EN
0
EN_471
The life of this milk carton may be thousands of years in this landfill.
Get a life.
life
N
{ "char_end": 8, "char_start": 4 }
{ "char_end": 10, "char_start": 6 }
EN
0
EN_472
The time of year for planting.
This time he succeeded.
time
N
{ "char_end": 8, "char_start": 4 }
{ "char_end": 9, "char_start": 5 }
EN
0
EN_490
The medical resident finished a two-week rotation in pediatrics and began one in orthopaedics.
The earth's rotation about its axis is responsible for its being slightly oblate rather than a sphere.
rotation
N
{ "char_end": 49, "char_start": 41 }
{ "char_end": 20, "char_start": 12 }
EN
0
EN_501
They tested his ability to locate objects in space.
Write your name in the space provided.
space
N
{ "char_end": 50, "char_start": 45 }
{ "char_end": 28, "char_start": 23 }
EN
0
EN_506
During his first year in office.
He rented an office in the new building.
office
N
{ "char_end": 31, "char_start": 25 }
{ "char_end": 19, "char_start": 13 }
EN
0
EN_516
He was at the top of his form.
To apply for the position, complete the application form.
form
N
{ "char_end": 29, "char_start": 25 }
{ "char_end": 56, "char_start": 52 }
EN
0
EN_545
Messages must go through diplomatic channels.
Do you think the sofa will go through the door?
go
V
{ "char_end": 16, "char_start": 14 }
{ "char_end": 29, "char_start": 27 }
EN
0
EN_568
The wee small voice of conscience.
The voice of the law.
voice
N
{ "char_end": 19, "char_start": 14 }
{ "char_end": 9, "char_start": 4 }
EN
0
EN_573
All the salesmen met their quota for the month.
The quota for Japanese imports was negotiated.
quota
N
{ "char_end": 32, "char_start": 27 }
{ "char_end": 9, "char_start": 4 }
EN
0
EN_575
He hoped for a new life in Australia.
Pottery was his life.
life
N
{ "char_end": 23, "char_start": 19 }
{ "char_end": 20, "char_start": 16 }
EN
0
EN_595
That mountain is solid rock.
He threw a rock at me.
rock
N
{ "char_end": 27, "char_start": 23 }
{ "char_end": 15, "char_start": 11 }
EN
0
EN_605
The development and printing of his pictures took only two hours.
After he saw the latest development he changed his mind and became a supporter.
development
N
{ "char_end": 15, "char_start": 4 }
{ "char_end": 35, "char_start": 24 }
EN
0
EN_607
The cinema relies on apparent motion.
He made a motion to adjourn.
motion
N
{ "char_end": 36, "char_start": 30 }
{ "char_end": 16, "char_start": 10 }
EN
0
EN_612
I don't have any money left.
What do we have here?
have
V
{ "char_end": 12, "char_start": 8 }
{ "char_end": 15, "char_start": 11 }
EN
0
EN_613
The alkaline inclination of the local waters.
Her inclination is for classical music.
inclination
N
{ "char_end": 24, "char_start": 13 }
{ "char_end": 15, "char_start": 4 }
EN
0
EN_633
Metal fatigue.
The soldiers were put on fatigue to teach them a lesson.
fatigue
N
{ "char_end": 13, "char_start": 6 }
{ "char_end": 32, "char_start": 25 }
EN
0
EN_648
The ball was already emptying out before the fire alarm sounded.
The ball rolled into the corner pocket.
ball
N
{ "char_end": 8, "char_start": 4 }
{ "char_end": 8, "char_start": 4 }
EN
0
EN_654
A five-pound beef heart will serve six.
He drew a heart and called it a valentine.
heart
N
{ "char_end": 23, "char_start": 18 }
{ "char_end": 15, "char_start": 10 }
EN
0
EN_655
He built himself a cabin in a hollow high up in the Appalachians.
The hollow of the hand or of a tree.
hollow
N
{ "char_end": 36, "char_start": 30 }
{ "char_end": 10, "char_start": 4 }
EN
0
EN_656
While we're in Brighton, let's get a stick of rock!
He threw a rock at me.
rock
N
{ "char_end": 50, "char_start": 46 }
{ "char_end": 15, "char_start": 11 }
EN
0
EN_669
There was a light at every corner.
They played as long as it was light.
light
N
{ "char_end": 17, "char_start": 12 }
{ "char_end": 35, "char_start": 30 }
EN
0
EN_687
Players had to reserve a court in advance.
Pay court to the emperor.
court
N
{ "char_end": 30, "char_start": 25 }
{ "char_end": 9, "char_start": 4 }
EN
0
EN_705
He's in the plumbing game.
The game is 6 all.
game
N
{ "char_end": 25, "char_start": 21 }
{ "char_end": 8, "char_start": 4 }
EN
0
EN_709
Add it to my score and I'll settle later.
A score were sent out but only one returned.
score
N
{ "char_end": 18, "char_start": 13 }
{ "char_end": 7, "char_start": 2 }
EN
0
EN_740
The devotion of his time and wealth to science.
Their devotion to each other was beautiful.
devotion
N
{ "char_end": 12, "char_start": 4 }
{ "char_end": 14, "char_start": 6 }
EN
0
EN_746
You take a chance when you let her drive.
We had the chance to meet the president last week.
chance
N
{ "char_end": 17, "char_start": 11 }
{ "char_end": 17, "char_start": 11 }
EN
0
EN_751
He held the title for two years.
He had no documents confirming his title to his father's estate.
title
N
{ "char_end": 17, "char_start": 12 }
{ "char_end": 40, "char_start": 35 }
EN
0
EN_757
The winning goal came with less than a minute left to play.
My lifelong goal is to get into a Hollywood movie.
goal
N
{ "char_end": 16, "char_start": 12 }
{ "char_end": 16, "char_start": 12 }
EN
0
EN_763
Printing at higher resolution will cause a reduction in performance.
The resolution of an equation.
resolution
N
{ "char_end": 29, "char_start": 19 }
{ "char_end": 14, "char_start": 4 }
EN
0
EN_766
A swelling with a disposition to rupture.
Salt has a disposition to dissolve in water.
disposition
N
{ "char_end": 29, "char_start": 18 }
{ "char_end": 22, "char_start": 11 }
EN
0
EN_775
One end of the box was marked 'This side up'.
We have given it at the end of the section since it involves the calculus.
end
N
{ "char_end": 7, "char_start": 4 }
{ "char_end": 27, "char_start": 24 }
EN
0
EN_779
He came to a bad end.
One end of the box was marked 'This side up'.
end
N
{ "char_end": 20, "char_start": 17 }
{ "char_end": 7, "char_start": 4 }
EN
0
EN_790
It was debated whether saving grace could be obtained outside the membership of the church.
The victor's grace in treating the vanquished.
grace
N
{ "char_end": 35, "char_start": 30 }
{ "char_end": 18, "char_start": 13 }
EN
0
EN_823
There's no place like home.
The teacher asked how many people made up his home.
home
N
{ "char_end": 26, "char_start": 22 }
{ "char_end": 50, "char_start": 46 }
EN
0
EN_834
They endured a torrent of inquiries.
A torrent of abuse.
torrent
N
{ "char_end": 22, "char_start": 15 }
{ "char_end": 9, "char_start": 2 }
EN
0
EN_836
Birds of prey often seize small mammals.
To seize smuggled goods.
seize
V
{ "char_end": 25, "char_start": 20 }
{ "char_end": 8, "char_start": 3 }
EN
0
EN_849
The arrangement of the furniture.
A flower arrangement.
arrangement
N
{ "char_end": 15, "char_start": 4 }
{ "char_end": 20, "char_start": 9 }
EN
0
EN_858
I couldn't tell what the thing was.
To say the same thing in other terms.
thing
N
{ "char_end": 30, "char_start": 25 }
{ "char_end": 21, "char_start": 16 }
EN
0
EN_864
My Russian vocabulary is very limited.
The vocabulary of any language is influenced by contacts with other cultures.
vocabulary
N
{ "char_end": 21, "char_start": 11 }
{ "char_end": 14, "char_start": 4 }
EN
0
EN_873
Your header is too long; "Local Cannibals" will suffice.
He took a header into the shrubbery.
header
N
{ "char_end": 11, "char_start": 5 }
{ "char_end": 16, "char_start": 10 }
EN
0
EN_876
We had a feast to celebrate the harvest.
The Thanksgiving feast.
feast
N
{ "char_end": 14, "char_start": 9 }
{ "char_end": 22, "char_start": 17 }
EN
0
EN_877
He asked to see the executive who handled his account.
He predicted that although it is of small account now it will rapidly increase in importance.
account
N
{ "char_end": 53, "char_start": 46 }
{ "char_end": 49, "char_start": 42 }
EN
0
EN_888
I have known him all his life.
She's my love, my life.
life
N
{ "char_end": 29, "char_start": 25 }
{ "char_end": 22, "char_start": 18 }
EN
0
EN_901
The computer weather model did not correctly predict the path of the hurricane.
A model of clarity.
model
N
{ "char_end": 26, "char_start": 21 }
{ "char_end": 7, "char_start": 2 }
EN
0
EN_902
The queen and her court traveled to the city to welcome back the soldiers.
The house was built around an inner court.
court
N
{ "char_end": 23, "char_start": 18 }
{ "char_end": 41, "char_start": 36 }
EN
0
EN_921
He took a course in basket weaving.
She prepared a three course meal.
course
N
{ "char_end": 16, "char_start": 10 }
{ "char_end": 27, "char_start": 21 }
EN
0
EN_940
The very picture of a modern general.
The political picture is favorable.
picture
N
{ "char_end": 16, "char_start": 9 }
{ "char_end": 21, "char_start": 14 }
EN
0
EN_943
The conflict between the government and the rebels began three years ago.
I wanted to attend the meeting but there's a conflict in my schedule that day.
conflict
N
{ "char_end": 12, "char_start": 4 }
{ "char_end": 53, "char_start": 45 }
EN
0
EN_951
It may rain in which case the picnic will be canceled.
A window case.
case
N
{ "char_end": 25, "char_start": 21 }
{ "char_end": 13, "char_start": 9 }
EN
0
End of preview. Expand in Data Studio

Dataset Card for PS-Eval Dataset

Dataset Summary

The PS-Eval Dataset is a suite of polysemous and monosemous contexts extracted and filtered from the WiC dataset. It aims to evaluate the ability of Sparse Autoencoders (SAEs) to disentangle polysemantic activations into monosemantic features within large language models (LLMs). The dataset contains 1,112 samples balanced between two classes:

  • Poly-contexts: Target words with different meanings across two contexts (Label: 0).
  • Mono-contexts: Target words with the same meaning across two contexts (Label: 1).

Each sample includes two sentences (contexts) containing the target word, along with a label indicating whether the target word's meaning is the same or different.

This dataset is particularly useful for evaluating methods and models that address polysemy in LLMs, such as feature-based interpretability techniques.

Supported Tasks and Leaderboards

  • Polysemy Detection: Classify whether the target word has the same or different meaning across contexts.
  • Feature Interpretability: Evaluate whether Sparse Autoencoders (SAEs) can map polysemantic activations into monosemantic features.

This dataset can also serve as a benchmark for context-sensitive word representations.

Languages

The dataset is in English.

Dataset Structure

Data Instances

Each instance in the dataset is stored in JSON format with the following structure:

{
      "id": "EN_22",
      "context_1": "They stopped at an open space in the jungle.",
      "context_2": "The astronauts walked in outer space without a tether.",
      "target_word": "space",
      "pos": "N",
      "target_word_location_1": {
          "char_start": 24,
          "char_end": 29
      },
      "target_word_location_2": {
          "char_start": 31,
          "char_end": 36
      },
      "language": "EN",
      "label": 0
}

Data Fields

  • target_word (string): The polysemous or monosemous word shared across the two contexts.
  • context1 (string): The first sentence containing the target word.
  • context2 (string): The second sentence containing the target word.
  • label (integer): Binary label where:
    • 0 = Different meanings (poly-contexts).
    • 1 = Same meaning (mono-contexts).

Data Splits

The dataset is provided as a single split with 1,112 samples:

  • Poly-contexts (Label 0): 556 samples
  • Mono-contexts (Label 1): 556 samples

Dataset Creation

Source Data

The PS-Eval dataset is built on top of the WiC Dataset (Word-in-Context) – a rich resource for polysemous words originally introduced in Pilehvar and Camacho-Collados (2019).

Filtering Process

We carefully selected instances from WiC where the target word is tokenized as a single token in GPT-2-small. This ensures consistency when analyzing activations in Sparse Autoencoders.

Annotations

Labels are derived from the WiC dataset:

  • Different meanings: Target words in poly-contexts (Label 0).
  • Same meaning: Target words in mono-contexts (Label 1).

Dataset Usage

Intended Use

This dataset is designed for evaluating models and methods that:

  • Analyze polysemantic and monosemantic activations in LLMs.
  • Detect context-sensitive meanings of polysemous words.
  • Test Sparse Autoencoders (SAEs) for interpretability.

Example Usage

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("gouki510/wic_eval_data")

# Inspect a sample
print(dataset["train"][0])

Metrics

The dataset supports evaluation metrics such as:

  • Accuracy
  • Precision
  • Recall
  • F1 Score
  • Specificity

These metrics are particularly important for evaluating polysemy detection models and Sparse Autoencoders.

For implementation details of the evaluation metrics, please refer to the GitHub repository: [link_to_your_repo].

Dataset Curators

This dataset was curated by Gouki Minegishi as part of research on polysemantic activation analysis in Sparse Autoencoders and interpretability for large language models.

Citation

If you use the PS-Eval Dataset in your work, please cite:

@inproceedings{minegishi2024ps-eval,
  title={Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Words},
  author={Gouki Minegishi, Hiroki Furuta, Yusuke Iwasawa, Yutaka Matsuo},
  year={2024},
  url={hoge}
}
Downloads last month
47