fix readme

.gitattributes

@@ -79,3 +79,4 @@ data/t_rex.filter.min_entity_12_max_predicate_10.jsonl filter=lfs diff=lfs merge
 data/t_rex.filter.min_entity_16_max_predicate_100.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.min_entity_4_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
 data/t_rex.filter.min_entity_8_max_predicate_25.jsonl filter=lfs diff=lfs merge=lfs -text
+data/t_rex.filter_unified.jsonl filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1 @@
+data_raw

README.md

@@ -21,12 +21,13 @@ We split the raw T-REX dataset into train/validation/test split by the ratio of
 ### Filtering to Remove Noise
 
 We apply filtering to keep triples with alpha-numeric subject and object, as well as triples with at least either of subject or object is a named-entity.
+After the filtering, we manually remove too vague and noisy predicate, and unify same predicates with different names.
 
-| Dataset | Raw | Filter |
-|--------:|----:|-------:|
-| Triples | 941,663 | 662,482 |
-| Predicate| 931 | 818 |
-| Entity | 270,801 | 197,302 |
+| Dataset   | Raw       | Filter    | Unification   |
+|----------:|----------:|----------:|--------------:|
+| Triples   | 941,663   | 583,333   | 432,795       |
+| Predicate | 931       | 659       | 247           |
+| Entity    | 270,801   | 197,163   | 149,172       |
 
 ### Filtering to Purify the Dataset
 We reduce the size of the dataset by applying filtering based on the number of predicates and entities in the triples.
@@ -54,11 +55,11 @@ we choose top-`max predicate` triples based on the frequency of the subject and
 
 - distribution of entities
 
-<img src="https://huggingface.co/datasets/relbert/t_rex/resolve/main/data/stats.entity_distribution.png" alt="" width="500" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
+<img src="https://huggingface.co/datasets/relbert/t_rex/resolve/main/data/stats.entity_distribution.png" alt="" width="600" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
 
 - distribution of predicates
 
-<img src="https://huggingface.co/datasets/relbert/t_rex/resolve/main/data/stats.predicate_distribution.png" alt="" width="500" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
+<img src="https://huggingface.co/datasets/relbert/t_rex/resolve/main/data/stats.predicate_distribution.png" alt="" width="600" style="margin-left:'auto' margin-right:'auto' display:'block'"/>
 
 
 ## Dataset Structure

check_predicate.py

@@ -1,11 +0,0 @@
-import json
-import pandas as pd
-
-with open("data/t_rex.filter.jsonl") as f:
-    data = pd.DataFrame([json.loads(i) for i in f.read().split('\n') if len(i) > 0])
-freq = data.groupby("predicate").count()['title']
-data['freq'] = [freq.loc[i] for i in data['predicate']]
-tmp = data.groupby("predicate").sample(10, replace=True)
-tmp = tmp.drop_duplicates()
-tmp.to_csv("data/t_rex.filter.predicate_check_sample.csv", index=False)
-

data/t_rex.filter.jsonl

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:1b80224241efafb2b053bc630288a3adfbaf69422454b57c620c6f810da20aee
-size 603995857
+oid sha256:ce5fdbf59ad4aaa7de18198863aa1d1b9e451386748f0a854b3b7ab1bff7f79f
+size 502849464

data/t_rex.filter_unified.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0aeaa043780559579cc3326e9eb923e09de1842590add1b1d92e4b19454f586
+size 408520515

filtering_denoise.py

@@ -1,47 +0,0 @@
-import json
-import string
-import re
-
-stopwords = ["he", "she", "they", "it"]
-list_alnum = string.ascii_lowercase + '0123456789 '
-
-
-def filtering(entry):
-
-    def _subfilter(token):
-        if len(re.findall(rf'[^{list_alnum}]+', token)) != 0:
-            return False
-        if token in stopwords:
-            return False
-        if token.startswith("www"):
-            return False
-        if token.startswith("."):
-            return False
-        if token.startswith(","):
-            return False
-        if token.startswith("$"):
-            return False
-        if token.startswith("+"):
-            return False
-        if token.startswith("#"):
-            return False
-        return True
-
-    if not _subfilter(entry["object"].lower()):
-        return False
-    if not _subfilter(entry["subject"].lower()):
-        return False
-
-    if entry['object'].islower() and entry['subject'].islower():
-        return False
-
-    return True
-
-
-with open(f"data/t_rex.raw.jsonl") as f:
-    data = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
-print(f"[before]: {len(data)}")
-data = [i for i in data if filtering(i)]
-print(f"[after] : {len(data)}")
-with open(f"data/t_rex.filter.jsonl", 'w') as f:
-    f.write('\n'.join([json.dumps(i) for i in data]))

predicate_manual_check.csv

@@ -0,0 +1,660 @@
+unique predicates,ok,pretty relation name,pretty relation name is reverse,same as,reverse of,remove (noisy),remove (too vague)
+abolished,,,,ending,,,
+academic discipline,,,,discipline,,,
+studies,x,[Academic Subject] studies [Topic],,,,,
+actor,,,,film starring,,,
+actress,,,,film starring,,,
+signatory,x,[Act] is signed by [Person],,,,,
+adjacent to,,,,next to,,,
+administered by,,,,maintained by,,,
+maiden flight,x,[Aircraft] makes the maiden flight on [Date],,,,,
+advisor,,,,studied at,,,
+hub,x,[Airline] has a hub in [Location],,,,,
+airline alliance,x,[Airline] is in [Airline Alliance],,,,,
+alma mater,,,,studied at,,,
+fleet,x,[Army] has [Fleet],,,,,
+alumni of,,,,studied at,,,
+alumnus of,,,,studied at,,,
+amalgamation of,,,,,,x,
+garrison,x,[Army] is based in [Location],,,,,
+series,x,[Art Work] follows after [Art Work],,,,,
+painting of,x,[Art Work] is a painting of [Person],,,,,
+architect,,,,architectural style,,,
+translation of,x,[Art Work] is a translation of [Art Work],,,,,
+arena,,,,home field,,,
+artery,,,,,,x,
+artist,,,,,,x,
+artists,,,,,,x,
+aspect of,,,,,,x,
+assassin,,,,murdered by,,,
+assassinated by,,,,murdered by,,,
+adapted from,x,[Art Work] is adapted from [Art Work],,,,,
+assembly of,,,,,assembly,,
+director,x,[Art Work] is directed by [Person],,,,,
+painter,x,[Art Work] is painted by [Person],,,,,
+published in,x,[Art Work] is published in [Magazine],,,,,
+authority,,,,council,,,
+published,x,[Art Work] is published on [Date],,,,,
+awarded,,,,award,,,
+awarded by,,,,,award,,
+awards,,,,award,,,
+sculptor,x,[Art Work] is sculpted by [Person],,,,,
+band,,,,music by,,,
+based on,,,,series,,,
+battle,,,,war,,,
+began,,,,starting,,,
+beginning,,,,starting,,,
+belongs to,,,,,divisions,,
+bestowed by,,,,founder,,,
+author,x,[Art Work] is written by [Person],,,,,
+engine,x,[Artifact] has [Engine],,,,,
+birthday,,,,birth place,,,
+birthplace,,,,birth date,,,
+book by,,,,author,,,
+border,,,,next to,,,
+bordered by,,,,next to,,,
+borders,,,,next to,,,
+born at,,,,birth place,,,
+born in,,,,birth place,,,
+born on,,,,birth date,,,
+boyfriend,,,,,husband,,
+branch,,,,sector,,,
+brand,,,,,product,,
+shape,x,[Artifact] has a shape of [Shape],,,,,
+coat of arms,x,[Artifact] is a coat of arms of [Group],x,,,,
+motif,x,[Artifact] is a motif for [Art Work],x,,,,
+built by,,,,builder,,,
+built from,,,,constructed from,,,
+burial place,,,,tomb,,,
+buried in,,,,tomb,,,
+patron saint,x,[Artifact] is a patron saint of [Country],x,,,,
+capital,,,,,capital of,,
+capital city,,,,,capital of,,
+result,x,[Artifact] is a result of [Artifact],,,,,
+type of,x,[Artifact] is a type of [Type],,,,,
+career,,,,job,,,
+edition of,x,[Artifact] is a version of [Artifact],,,,,
+cartridge,,,,,,x,
+cast member,,,,film starring,,,
+established,x,[Artifact] is built on [Date],,,,,
+cause,,,,caused by,,,
+cause of,,,,,caused by,,
+discoverer,x,[Artifact] is discovered by [Person],,,,,
+formation,x,[Artifact] is formation of [Army],,,,,
+CEO,,,,chief executive,,,
+formed from,x,[Artifact] is formed from [Artifact],,,,,
+chairperson,,,,chairman,,,
+contained within,x,[Artifact] is found in [Artifact],,,,,
+colour,x,[Artifact] is in [Color],,,,,
+channel,,,,broadcast on,,,
+influenced by,x,[Artifact] is influenced by [Artifact],,,,,
+listing,x,[Artifact] is listed in [List],,,,,
+chief executive officer,,,,chief executive,,,
+children,,,,,father,,
+maintained by,x,[Artifact] is maintained by [Company],,,,,
+chromosome,,,,,,x,
+manufacturer,x,[Artifact] is manufactured by [Company],,,,,
+citizen of,,,,residence,,,
+citizenship,,,,residence,,,
+named for,x,[Artifact] is name of [Artifact],,,,,
+city council,,,,council,,,
+named after,x,[Artifact] is named after [Person],,,,,
+operating system,x,[Artifact] is the OS of [Software],,,,,
+co-driver,,,,,,x,
+coach,,,,head coach,,,
+platform,x,[Artifact] is the platform of [Game],x,,,,
+used by,x,[Artifact] is used by [Person],,,,,
+namesake,x,[Artifact] is used for its namesake of [Artifact],,,,,
+color,,,,colour,,,
+producer,x,[Artist] is produced by [Person],,,,,
+movement,x,[Artists] leads [Movement],,,,,
+commissioned by,,,,producer,,,
+compiled by,,,,,,x,
+composed by,,,,music by,,,
+athletes,x,[Athlete] joins [Competition],x,,,,
+composer,,,,music by,,,
+presented by,x,[Award] is presented by [Company],,,,,
+conflict,,,,,,x,
+connected with,,,,,,x,
+conservation status,,,,,,x,
+central bank,x,[Bank] is the central bank of [Country],x,,,,
+carries,x,[Bridge] crosses [Artifact],,,,,
+bridge over,x,[Bridge] crosses [River],,,,,
+architectural style,x,[Building] has an architectural style of [Person],,,,,
+contains,,,,is in,,,
+constructed from,x,[Building] is built from [Material],,,,,
+cathedral,x,[Cathedral] is in [City],x,,,,
+costume designer,,,,,work,,
+ore,x,[Chemical Material] is made from [Stone],,,,,
+country,,,,,made from,,
+country of origin,,,,,,x,
+capital town,x,[City] is a capital town of [Country],x,,,,
+county seat of,,,,,county seat,,
+twin city,x,[City] is a twin city of [City],,,,,
+craft,,,,job,,,
+city,x,[City] is in [Country],,,,,
+created by,,,,,work,,
+created from,,,,formed from,,,
+creator,,,,,work,,
+crosses,,,,,carries,,
+capital of,x,[City] is the capital of [Country],,,,,
+HQ,x,[Company] has the headquarter in [Location],,,,,
+divisions,x,[Company] is a subsidiary of [Company],x,,,,
+sector,x,[Company] is in a sector of [Sector],,,,,
+dad,,,,father,,,
+daughter,,,,,father,,
+daughters,,,,,father,,
+dead,,,,died on,,,
+death,,,,died on,,,
+operated by,x,[Company] operates [Vehicle],,,,,
+product,x,[Company] owns [Product],,,,,
+publisher,x,[Company] publishes [Art Work],,,,,
+depicts,,,,painting of,,,
+derived from,,,,,,x,
+designation,,,x,,,,
+designed by,,,,,work,,
+designer,,,,,work,,
+sport,x,[Competition] is a league of [Sport],,,,,
+developer,,,,developed by,,,
+topic,x,[Conference] is about [Topic],,,,,
+council,x,[Council] is the council of [Country],x,,,,
+claimed by,x,[Country] claims [City],x,,,,
+diet,,,,,,x,
+diocese,,,,,,x,
+diplomatic relation,,,,,,x,
+directed by,,,,director,,,
+history,x,[Country] has [History],,,,,
+director of photography,,,,director,,,
+assembly,x,[Country] is [Political Party] assembly,,,,,
+disciple,,,,,student of,,
+disciple of,,,,student of,,,
+disciples,,,,,student of,,
+member of,x,[Country] is a member of [Group],,,,,
+discography,,,,,,x,
+discovered by,,,,discoverer,,,
+enclave within,x,[Country] is enclaved by [Country],,,,,
+disestablished,,,,,,x,
+continent,x,[Country] is in [Continent],,,,,
+distinct from,,,,,,x,
+distributed by,,,,film studio,,,
+distribution,,,,,,x,
+distributor,,,,film studio,,,
+divided into,,,,,city,,
+division,,,,sports league,,,
+athlete,x,[Country] joins [War],x,,,,
+county seat,x,[Country]'s county seat is [Location],,,,,
+flag,x,[Country]'s flag is [Artifact],,,,,
+culture,x,[Culture] is originated in [Country],x,,,,
+edited by,,,,author,,,
+currency,x,[Currency] is used in [Country],x,,,,
+editions,,,,,,x,
+editor,,,,author,,,
+educated at,,,,studied at,,,
+education,,,,studied at,,,
+effect,,,,,,x,
+symptoms,x,[Disease] has [Symptoms],,,,,
+employed by,,,,works at,,,
+employment,,,,job,,,
+caused by,x,[Disease] is caused by [Virus],,,,,
+precedes,x,[Era] comes before [Era],,,,,
+endemic to,,,,,,x,
+ending,x,[Event] ends on [Date],,,,,
+time,x,[Event] happens on [Date],,,,,
+epoch,,,,era,,,
+since,x,[Event] is since [Date],,,,,
+starting,x,[Event] starts on [Date],,,,,
+ethnic group,,,,,,x,
+ethnicity,,,,nationality,,,
+event,,,,took part,,,
+took part,x,[Event] takes place at [Location],x,,,,
+executive body,,,,government,,,
+executive branch,,,,government,,,
+executive producer,,,,film studio,,,
+faith,,,,religion,,,
+family,,,,,,x,
+famous works,,,,work,,,
+mascot,x,[Fictional Character] is a mascot of [Sport Team],x,,,,
+fields,,,,discipline,,,
+film director,,,,director,,,
+film editor,,,,author,,,
+characters,x,[Fictional Character] is from [Art Work] ,x,,,,
+film producer,,,,film studio,,,
+made from,x,[Food] is made from [Ingredient],,,,,
+cuisine,x,[Food] is originated in [Country],,,,,
+fruit of,x,[Fruit] is [Type],,,,,
+video game publisher,x,[Game] is produced by [Company],,,,,
+first language,,,,native language,,,
+deity of,x,[God] is the god of [Religion],,,,,
+government,x,[Government] is the government of [Country],x,,,,
+jurisdiction,x,[Government] is the jurisdiction of [City],,,,,
+followed by,,,,,series,,
+follows,,,,series,,,
+fork of,,,,,,x,
+form of government,,,,government,,,
+section of,x,[Group] has a section of [Group],,,,,
+religion,x,[Group] is [Religion],,,,,
+formed out of,,,,,,x,
+successor,x,[Group] is a predecessor of [Group],,,,,
+founded,,,,foundation,,,
+founded by,,,,founder,,,
+religious order,x,[Group] is a religious order of [Group],,,,,
+created,x,[Group] is created on [Date],,,,,
+founded in,x,[Group] is founded at [Location],,,,,
+founder,x,[Group] is founded by [Person],,,,,
+game engine,,,,video game publisher,,,
+gameplay,,,,,,x,
+foundation,x,[Group] is founded on [Date],,,,,
+gauge,,,,,,x,
+general manager,,,,team manager,,,
+funded by,x,[Group] is funded by [Person],,,,,
+girlfriend,,,,husband,,,
+given by,,,,presented by,,,
+legislative body,x,[Group] is legislative body of [Country],x,,,,
+graduate of,,,,studied at,,,
+graduated from,,,,studied at,,,
+ground,,,,home field,,,
+group,,,,music by,,,
+happens in,,,,,,x,
+legislature,x,[Group] is legislature of [Country],x,,,,
+merged into,x,[Group] is merged into [Group],,,,,
+parliament,x,[Group] is the parliament of [Country],x,,,,
+heir,,,,,,x,
+official language,x,[Group] speaks [Language],,,,,
+highest point,,,,highest peak,,,
+highway system,,,,,,x,
+leader,x,[Group]'s leader is [Person],,,,,
+appointed by,x,[Head of Government] is appointed by [Head of Government],,,,,
+home ground,,,,home field,,,
+home venue,,,,home field,,,
+honors,,,,award,,,
+honours,,,,award,,,
+public holiday,x,[Holiday] is a national holiday of [Country],x,,,,
+island,x,[Island] is [Country],,,,,
+head of state,x,[Job] is the head of state in [Location],x,,,,
+land,x,[Land] is [Country],,,,,
+alphabet,x,[Language] consists of [Alphabet],,,,,
+illustrated by,,,,,work,,
+illustrator,,,,,work,,
+imprisoned in,,,,prison,,,
+inaugurated,,,,established,,,
+include,,,,,is in,,
+includes,,,,,is in,,
+includes part,,,,,,x,
+incorporated,,,,established,,,
+incumbent,,,,,,x,
+dialect,x,[Language] is a dialect of [Language],x,,,,
+indigenous to,,,,culture,,,
+industry,,,,sector,,,
+inflow,,,,,,x,
+enacted by,x,[Law] is enacted by [Group],,,,,
+informed by,,,,,,x,
+ingredient,,,,made from,,,
+inspiration,,,,influenced by,,,
+inspired by,,,,influenced by,,,
+instance of,,,,is a,,,
+instruction set,,,,,,x,
+instrument,,,,plays,,,
+intersex,,,,male,,,
+invented by,,,,,work,,
+inventor,,,,,work,,
+is a,,,,,,,x
+is a type of,,,,is a,,,
+is an,,,,is a,,,
+is in,,,,,,,x
+is in the borough of,,,,town,,,
+is in the city of,,,,city,,,
+is in the county of,,,,town,,,
+is in the district of,,,,town,,,
+is in the municipality of,,,,town,,,
+is in the parish of,,,,,,x,
+is in the province of,,,,state,,,
+is in the region of,,,,is in,,,
+is in the state of,,,,state,,,
+is in the town of,,,,state,,,
+is in the village of,,,,town,,,
+is located in,,,,is in,,,
+is not,,,,,,x,
+is on,,,,,,x,
+is owned by,,,,,product,,
+ISA,,,,,,x,
+approved by,x,[License] is approved by [Organization],,,,,
+ballpark,x,[Location] is a ballpark of [Sport Team],x,,,,
+river mouth,x,[Location] is a river mouth of [Bay],,,,,
+kids,,,,,father,,
+killed by,,,,murdered by,,,
+killer,,,,murdered by,,,
+kind of,,,,is a,,,
+sovereign state,x,[Location] is a sovereign state of [Location],,,,,
+known for,,,,work,,,
+label,,,,record label,,,
+administrative centre,x,[Location] is an administrative center of [Location],,,,,
+language,,,,native language,,,
+language official,,,,native language,,,
+language spoken,,,,native language,,,
+latitude,,,,is in,,,
+launch date,,,,established,,,
+launch vehicle,,,,,,x,
+Indian reservation,x,[Location] is an Indian reservation in [Country],,,,,
+league,,,,sports league,,,
+closed,x,[Location] is closed on [Date],,,,,
+exclave of,x,[Location] is exclave of [Country],,,,,
+librettist,,,,libretto by,,,
+planet,x,[Location] is in [Planet],,,,,
+licence,,,,license,,,
+next to,x,[Location] is next to [Location],,,,,
+life partner,,,,,husband,,
+on the coast of,x,[Location] is on the coast of [Ocean],,,,,
+literary works,,,,work,,,
+lived in,,,,residence,,,
+locality,,,,city,,,
+located in,,,,is in,,,
+location,,,,,,x,
+longitude,,,,,,x,
+lover,,,,husband,,,
+split from,x,[Location] is split from [Location],,,,,
+lyricist,,,,music by,,,
+lyrics by,,,,music by,,,
+made by,,,,,product,,
+highest peak,x,[Location] is the highest peak in [Country],x,,,,
+made of,,,,made from,,,
+lowest point,x,[Location] is the lowest point of [Location],x,,,,
+crystal system,x,[Material]'s crystal system is [Crystal System],,,,,
+maintenance,,,,maintained by,,,
+major works,,,,work,,,
+maker,,,,,product,,
+makes,,,,product,,,
+treats,x,[Medication] is for [Disease],,,,,
+man,,,,male,,,
+managed by,,,,maintained by,,,
+manufactured by,,,,manufacturer,,,
+used for treatment,x,[Medicine] is for [Disease],x,,,,
+manufactures,,,,,manufacturer,,
+married to,,,,marry,,,
+film genre,x,[Movie] is [Genre],,,,,
+libretto by,x,[Movie] is a libretto by [Person],,,,,
+master,,,,student of,,,
+spin-off,x,[Movie] is a spinoff of [Movie],x,,,,
+medals,,,,,,x,
+medium,,,,made from,,,
+member,,,,composed of,,,
+first broadcast,x,[Movie] is first broadcast on [Date],,,,,
+members,,,,composed of,,,
+universe,x,[Movie] is in the universe of [Art Work],,,,,
+mentor,,,,student of,,,
+film studio,x,[Movie] is produced by [Company],,,,,
+merged with,,,,merged into,,,
+screenplay by,x,[Movie] is screenplayed by [Person],,,,,
+mode,,,,,,x,
+film starring,x,[Movie] stars [Actor],,,,,
+month,,,,,,x,
+mother,,,,father,,,
+mother tongue,,,,native language,,,
+collection,x,[Museum] has [Collection of Work],x,,,,
+dedicated to,x,[Museum] is dedicated to [Person],,,,,
+movie director,,,,director,,,
+municipal council,,,,council,,,
+memorial to,x,[Museum] is memorial to [Event],,,,,
+murderer,,,,murdered by,,,
+genre,x,[Music Artist] is [Genre],,,,,
+music genre,,,,genre,,,
+music group,,,,music by,,,
+musical artist,,,,music by,,,
+musical score by,,,,music by,,,
+musician,,,,music by,,,
+anthem,x,[Music] is an anthem of [Country],x,,,,
+music by,x,[Music] is made by [Artist],,,,,
+studio,x,[Music] is recorded at [Studio],,,,,
+nation,,,,city,,,
+national holiday,,,,public holiday,,,
+released,x,[Music] is released on [Date],,,,,
+official residence,x,[Occupation] lives in [Location],,,,,
+native to,,,,culture,,,
+network,,,,broadcast on,,,
+ideology,x,[Organization]'s ideology is [Ideology],,,,,
+noble family,,,,,,x,
+nominated for,,,,award,,,
+nominee for,,,,award,,,
+not to be confused with,,,,,,x,
+notable work,,,,,author,,
+occupation,,,,job,,,
+occupied by,,,,maintained by,,,
+of team,,,,played for,,,
+office held,,,,,,x,
+CPU,x,[PC]'s cpu is [CPU],,,,,
+marry,x,[Person] and [Person] are married,,,,,
+political party,x,[Person] belongs to [Political Party],,,,,
+on the shore of,,,,on the coast of,,,
+opened,,,,starting,,,
+operated,,,,operated by,,,
+record label,x,[Person] belongs to [Record Label],,,,,
+operates,,,,operated by,,,
+builder,x,[Person] built [Artifact],x,,,,
+operator,,,,operated by,,,
+war,x,[Person] causes [War],,,,,
+orbits,,,,orbit,,,
+choreographer,x,[Person] choreographs [Play],x,,,,
+organizer,,,,maintained by,,,
+original language,,,,native language,,,
+originates from,,,,,,x,
+OS,,,,operating system,,,
+coined by,x,[Person] coins [Artifact],x,,,,
+outlet,,,,outflow,,,
+overlies,,,,,,x,
+owned by,,,,,divisions,,
+owner,,,,,divisions,,
+owner of,,,,divisions,,,
+work,x,[Person] creates [Work],,,,,
+died in,x,[Person] dies at [Location],,,,,
+parent,,,,father,,,
+parent company,,,,,divisions,,
+parent company of,,,,divisions,,,
+died on,x,[Person] dies on [Date],,,,,
+part of,,,,is in,,,
+participant,,,,player,,,
+participant in,,,,,player,,
+participant of,,,,,player,,
+partner,,,,,husband,,
+parts,,,,is in,,,
+party,,,,,,x,
+passed by,,,,signatory,,,
+disappeared,x,[Person] disappears on [Date],,,,,
+first described by,x,[Person] first finds [Disease],x,,,,
+patron saint of,,,,,patron saint,,
+pendant of,,,,,,x,
+disorder,x,[Person] has [Diseases],,,,,
+period,,,,era,,,
+house,x,[Person] has a house in [Location],,,,,
+place of birth,,,,birth place,,,
+place of death,,,,died in,,,
+job,x,[Person] is [Occupation],,,,,
+rank,x,[Person] is [Rank],,,,,
+platforms,,,,platform,,,
+male,x,[Person] is [Sex],,,,,
+candidate,x,[Person] is a candidate of [Election],x,,,,
+chancellor,x,[Person] is a chancellor of [Country],x,,,,
+chief executive,x,[Person] is a chief executive of [Company],x,,,,
+plays for,,,,played for,,,
+father,x,[Person] is a child of [Person],,,,,
+portrait of,,,,painting of,,,
+portrayed by,,,,,work,,
+cinematographer,x,[Person] is a cinematographer of [Movie],x,,,,
+predecessor,,,,,successor,,
+head coach,x,[Person] is a coach of [Sport Team],x,,,,
+premiere,,,,first performance,,,
+commander of,x,[Person] is a commander of [Army],,,,,
+concubine,x,[Person] is a concubine of [Person],x,,,,
+consort,x,[Person] is a consort of [Person],,,,,
+husband,x,[Person] is a husband of [Person],,,,,
+team manager,x,[Person] is a manager of [Sport Team],x,,,,
+composed of,x,[Person] is a member of [Music Group],x,,,,
+produces,,,,producer,,,
+mistress,x,[Person] is a mistress of [Person],x,,,,
+production company,,,,film studio,,,
+production designer,,,,,work,,
+production house,,,,film studio,,,
+profession,,,,job,,,
+professor,,,,student of,,,
+programmer,,,,,,x,
+patron,x,[Person] is a patron of [Person],,,,,
+promotion to,,,,,,x,
+protection,,,,,,x,
+protocol,,,,,,x,
+premier,x,[Person] is a premier of [Group],x,,,,
+public office,,,,,,x,
+presenter,x,[Person] is a presenter of [TV show],x,,,,
+student of,x,[Person] is a student of [Person],,,,,
+active in,x,[Person] is active in [Location],,,,,
+publishing house,,,,publisher,,,
+pupil,,,,student of,,,
+pupil of,,,,student of,,,
+pupils,,,,student of,,,
+purpose,,,,use,,,
+award,x,[Person] is awarded by [Award],,,,,
+race,,,,nationality,,,
+nationality,x,[Person] is born in [Country],,,,,
+radio station,,,,radio network,,,
+birth place,x,[Person] is born in [Location],,,,,
+birth date,x,[Person] is born on [Date],,,,,
+tomb,x,[Person] is buried at [Tomb],,,,,
+record producer,,,,producer,,,
+regulatory body,,,,,,x,
+relative,,,,,,x,
+convicted of,x,[Person] is convicted of [Crime],,,,,
+relegation to,,,,,,x,
+drafted by,x,[Person] is drafted by [Group],,,,,
+era,x,[Person] is from [Era],,,,,
+replaced,,,,,successor,,
+replaced by,,,,,successor,,
+representative body,,,,,,x,
+represented by,,,,head of government,,,
+prison,x,[Person] is in [Prison],,,,,
+resident in,,,,residence,,,
+resting place,,,,tomb,,,
+murdered by,x,[Person] is killed by [Person],,,,,
+result of,,,,,result,,
+played for,x,[Person] is played at [Group],,,,,
+royal house,,,,dynasty,,,
+runs on,,,,,,x,
+chairman,x,[Person] is the chair of [Group],x,,,,
+school,,,,studied at,,,
+emperor,x,[Person] is the emperor of [Country],x,,,,
+screenwriter,,,,screenplay by,,,
+script,,,,alphabet,,,
+scriptwriter,,,,,,x,
+head of government,x,[Person] is the head of government of [Country],x,,,,
+seat,,,,county seat,,,
+king,x,[Person] is the king of [Country],x,,,,
+dynasty,x,[Person] is the leader of [Dynasty],,,,,
+senior coach,,,,head coach,,,
+separated from,,,,,,x,
+sequel of,,,,series,,,
+mayor,x,[Person] is the mayor of [City],x,,,,
+set of,,,,,,x,
+sex,,,,male,,,
+monarch,x,[Person] is the monarch of [Country],x,,,,
+shareholder,,,,,,x,
+shares border with,,,,next to,,,
+president,x,[Person] is the president of [Country],x,,,,
+signed by,,,,signatory,,,
+significant works,,,,work,,,
+prime minister,x,[Person] is the prime minister of [Country],x,,,,
+singer,,,,music by,,,
+sister city,,,,twin city,,,
+solid solution series with,,,,,,x,
+son,,,,,father,,
+songwriter,,,,author,,,
+sons,,,,,father,,
+queen,x,[Person] is the queen of [Country] ,x,,,,
+residence,x,[Person] live in [Location],,,,,
+played by,x,[Person] plays [Fictional Character],x,,,,
+plays,x,[Person] plays [Instrument],,,,,
+sports,,,,sport,,,
+voice actor,x,[Person] plays in [Movie],,,,,
+spouse,,,,,husband,,
+square,,,,is in,,,
+stadium,,,,home field,,,
+starring,,,,film starring,,,
+team,x,[Person] plays in [Sport Team],,,,,
+native language,x,[Person] speaks [Language],,,,,
+stock exchange,,,,,,x,
+discipline,x,[Person] studies [Academic Subject],,,,,
+student,,,,,student of,,
+studied at,x,[Person] studies at [School],,,,,
+students,,,,,student of,,
+works at,x,[Person] works at [Group],,,,,
+studied under,,,,student of,,,
+pet,x,[Pet] is a pet of [Person],x,,,,
+satellite,x,[Planet] is a satellite of [Planet],x,,,,
+subclass of,,,,is a,,,
+subdivided into,,,,,city,,
+subject,,,,,,x,
+subject of,,,,,,x,
+subsidiary,,,,divisions,,,
+subsidiary company,,,,divisions,,,
+subsidiary of,,,,divisions,,,
+subsystem of,,,,,,x,
+orbit,x,[Planet] is in the orbit of [Orbit],,,,,
+honours,,,,studied at,,,
+type species,x,[Plant] is [Type],,,,,
+first performance,x,[Play] is first performed on [Date],,,,,
+takes place in,,,,,,x,
+taxonomic rank,,,,,,x,
+teacher,,,,student of,,,
+teacher of,,,,,student of,,
+performer,x,[Play] is performed by [Person],,,,,
+radio network,x,[Radio Program] is broadcasted on [Radio Channel],,,,,
+teams played for,,,,played for,,,
+teleplay by,,,,,,x,
+television channel,,,,broadcast on,,,
+tenant,,,,,,x,
+railway line,x,[Railway] is in [Location],x,,,,
+territory claimed by,,,,claimed by,,,
+theme song,,,,music by,,,
+denomination,x,[Religion] is a denomination by [Artifact],,,,,
+drain,x,[River] drains [Location],,,,,
+tributary of,x,[River] is a tributary of [River],,,,,
+tonality,,,,,,x,
+outflow,x,[River] outflows to [Location],,,,,
+took part in,,,,took part,,,
+developed by,x,[Software] is developed by [Company],,,,,
+license,x,[Software] is under license of [License],,,,,
+use,x,[Software] is used for [Purpose],,,,,
+treatment,,,,treats,,,
+programming language,x,[Software] is written in [Programming Language],,,,,
+tributary,,,,,tributary of,,
+affiliate of,x,[Sport Team] is an affiliate of [Sport Team],,,,,
+tunnel under,,,,bridge over,,,
+TV channel,,,,broadcast on,,,
+sports league,x,[Sport Team] plays at [Competition],,,,,
+player,x,[Sport Team] plays in [Competition],x,,,,
+champion,x,[Sport Team] wins [Competition],x,,,,
+unit of,,,,,,x,
+home field,x,[Sport Team]'s home field is [Location],,,,,
+constellation,x,[Star] is a [Constellation],,,,,
+state,x,[State] is a state of [Country],,,,,
+used for,,,,use,,,
+terminus,x,[Station] is the terminus of [Railway],x,,,,
+used in,,,,use,,,
+user,,,,,,x,
+uses,,,,,use,,
+utility,,,,use,,,
+venue,,,,,took part,,
+street,x,[Street] is in the street of [Location],,,,,
+system of,x,[System] is a system in [Artifact],,,,,
+voiced by,,,,voice actor,,,
+time zone,x,[Timezone] is a timezon in [Country],x,,,,
+town,x,[Town] is in [Location],,,,,
+wife,,,,,husband,,
+winner,,,,champion,,,
+winners,,,,champion,,,
+woman,,,,male,,,
+won by,,,,champion,,,
+words by,,,,author,,,
+broadcast on,x,[TV Series] is broadcasted on [TV Channel],,,,,
+working for,,,,works at,,,
+works,,,,work,,,
+weapon,x,[Weapon] is used by [Person],x,,,,
+writer,,,,author,,,
+writing system,,,,alphabet,,,
+written by,,,,author,,,
+year,,,,time,,,

process.py

@@ -1,33 +1,87 @@
-"""
+""" Process raw T-Rex file.
+mkdir data_raw
+cd data_raw
 wget https://figshare.com/ndownloader/files/8760241
 unzip 8760241
+cd ../
 """
 
 import json
+import string
+import re
 import os
 from glob import glob
 from tqdm import tqdm
-# from random import shuffle, seed
-
-os.makedirs('data', exist_ok=True)
-f_writer = open('data/t_rex.raw.jsonl', 'w')
-for i in tqdm(glob("*.json")):
-    with open(i) as f:
-        data = json.load(f)
-    for _data in data:
-        for triple in _data['triples']:
-            p = triple['predicate']['surfaceform']
-            o = triple['object']['surfaceform']
-            s = triple['subject']['surfaceform']
-            if p is None or o is None or s is None:
-                continue
-            out = {"predicate": p, "object": o, "subject": s, "title": _data["title"], "text": _data["text"]}
-            f_writer.write(json.dumps(out) + "\n")
-f_writer.close()
-
-
-with open('data/t_rex.raw.jsonl') as f:
+
+import pandas as pd
+
+# process raw data
+if not os.path.exists('data/t_rex.raw.jsonl'):
+    os.makedirs('data', exist_ok=True)
+    f_writer = open('data/t_rex.raw.jsonl', 'w')
+    for i in tqdm(glob("data_raw/*.json")):
+        with open(i) as f:
+            data = json.load(f)
+        for _data in data:
+            for triple in _data['triples']:
+                p = triple['predicate']['surfaceform']
+                o = triple['object']['surfaceform']
+                s = triple['subject']['surfaceform']
+                if p is None or o is None or s is None:
+                    continue
+                out = {"predicate": p, "object": o, "subject": s, "title": _data["title"], "text": _data["text"]}
+                f_writer.write(json.dumps(out) + "\n")
+    f_writer.close()
+
+# apply filtering to remove noisy instances
+stopwords = ["he", "she", "they", "it"]
+list_alnum = string.ascii_lowercase + '0123456789 '
+
+
+def filtering(entry):
+
+    def _subfilter(token):
+        if len(re.findall(rf'[^{list_alnum}]+', token)) != 0:
+            return False
+        if token in stopwords:
+            return False
+        if token.startswith("www"):
+            return False
+        if token.startswith("."):
+            return False
+        if token.startswith(","):
+            return False
+        if token.startswith("$"):
+            return False
+        if token.startswith("+"):
+            return False
+        if token.startswith("#"):
+            return False
+        return True
+
+    if not _subfilter(entry["object"].lower()):
+        return False
+    if not _subfilter(entry["subject"].lower()):
+        return False
+
+    if entry['object'].islower() and entry['subject'].islower():
+        return False
+
+    return True
+
+
+with open(f"data/t_rex.raw.jsonl") as f:
     data = [json.loads(i) for i in f.read().split('\n') if len(i) > 0]
-s = {i['subject'] for i in data}
-o = {i['object'] for i in data}
-s.update(o)
+print(f"[before]: {len(data)}")
+data = [i for i in data if filtering(i)]
+df = pd.DataFrame(data)
+count = df.groupby("predicate")['title'].count()
+df = df[[count[p] >= 3 for p in df['predicate']]]
+df = df.drop_duplicates()
+print(f"[after] : {len(df)}")
+print(f"[entity]: {len(set(df['object'].unique().tolist() + df['subject'].unique().tolist()))}")
+_df = df[['object', 'subject']]
+data = [i.to_dict() for _, i in df.iterrows()]
+
+with open(f"data/t_rex.filter.jsonl", 'w') as f:
+    f.write('\n'.join([json.dumps(i) for i in data]))

t_rex.py

@@ -5,7 +5,7 @@ import datasets
 logger = datasets.logging.get_logger(__name__)
 _DESCRIPTION = """T-Rex dataset."""
 _NAME = "t_rex"
-_VERSION = "0.0.3"
+_VERSION = "0.0.4"
 _CITATION = """
 @inproceedings{elsahar2018t,
   title={T-rex: A large scale alignment of natural language with knowledge base triples},
@@ -17,8 +17,9 @@ _CITATION = """
 
 _HOME_PAGE = "https://github.com/asahi417/relbert"
 _URL = f'https://huggingface.co/datasets/relbert/{_NAME}/resolve/main/data'
-_TYPES = ["raw", "filter"]
-_URLS = {i: {str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.jsonl']} if i in ['raw', 'filter'] else {
+_TYPES = ["raw", "filter", "filter_unified"]
+_NON_SPLITS = ["raw", "filter", "filter_unified"]
+_URLS = {i: {str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.jsonl']} if i in _NON_SPLITS else {
     str(datasets.Split.TRAIN): [f'{_URL}/t_rex.{i}.train.jsonl'],
     str(datasets.Split.VALIDATION): [f'{_URL}/t_rex.{i}.validation.jsonl'],
     str(datasets.Split.TEST): [f'{_URL}/t_rex.{i}.test.jsonl']}
@@ -46,7 +47,7 @@ class TREX(datasets.GeneratorBasedBuilder):
 
     def _split_generators(self, dl_manager):
         downloaded_file = dl_manager.download_and_extract(_URLS[self.config.name])
-        if self.config.name in ['raw', 'filter']:
+        if self.config.name in _NON_SPLITS:
             return [datasets.SplitGenerator(
                 name=datasets.Split.TRAIN,
                 gen_kwargs={"filepaths": downloaded_file[str(datasets.Split.TRAIN)]})]

unify_predicate.py

@@ -0,0 +1,69 @@
+# import json
+# import pandas as pd
+#
+# with open("data/t_rex.filter.jsonl") as f:
+#     data = pd.DataFrame([json.loads(i) for i in f.read().split('\n') if len(i) > 0])
+# freq = data.groupby("predicate").count()['title']
+# data['freq'] = [freq.loc[i] for i in data['predicate']]
+# data = data[data['freq'] >= 3]
+# tmp = data.groupby("predicate").sample(10, replace=True)
+# tmp = tmp.drop_duplicates()
+# tmp.to_csv("data/t_rex.filter.predicate_check_sample.csv", index=False)
+
+import json
+import pandas as pd
+
+# load manual check sheet
+df_predicate = pd.read_csv('predicate_manual_check.csv')
+df_predicate = df_predicate[df_predicate['remove (noisy)'] != 'x']
+df_predicate = df_predicate[df_predicate['remove (too vague)'] != 'x']
+predicate_main = df_predicate[df_predicate['ok'] == 'x']['unique predicates'].tolist()
+df_sub = df_predicate[df_predicate['ok'] != 'x']
+df_sub_same = df_sub[['unique predicates', 'same as']].dropna()
+df_sub_same = df_sub_same[[i in predicate_main for i in df_sub_same['same as']]]
+df_sub_same.index = df_sub_same.pop('unique predicates')
+sub_same = df_sub_same['same as'].to_dict()
+df_sub_rev = df_sub[['unique predicates', 'reverse of']].dropna()
+df_sub_rev = df_sub_rev[[i in predicate_main for i in df_sub_rev['reverse of']]]
+df_sub_rev.index = df_sub_rev.pop('unique predicates')
+sub_rev = df_sub_rev['reverse of'].to_dict()
+
+# load data and filter based on manual predicate check sheet
+with open(f"data/t_rex.filter.jsonl") as f:
+    data = pd.DataFrame([json.loads(i) for i in f.read().split('\n') if len(i) > 0])
+data['predicate'] = [sub_same[i] if i in sub_same else i for i in data['predicate']]
+data['reverse'] = [i in sub_rev for i in data['predicate']]
+data['predicate'] = [sub_rev[i] if i in sub_rev else i for i in data['predicate']]
+data_filter = data[[i in predicate_main for i in data['predicate']]]
+
+data_filter_rev = data_filter[data_filter['reverse']].copy()
+o = data_filter_rev.pop("object")
+s = data_filter_rev.pop("subject")
+data_filter_rev["subject"] = o
+data_filter_rev["object"] = s
+data_filter[data_filter['reverse']] = data_filter_rev[data_filter.columns]
+data_filter.pop("reverse")
+
+df_main = df_predicate[df_predicate['ok'] == 'x'][['unique predicates', 'pretty relation name', 'pretty relation name is reverse']]
+df_main['reverse'] = [i == 'x' for i in df_main.pop('pretty relation name is reverse')]
+df_main['predicate'] = df_main.pop('unique predicates')
+
+data_filter_join = data_filter.merge(df_main, how='inner', on='predicate')
+data_filter_join_rev = data_filter_join[data_filter_join['reverse']].copy()
+o = data_filter_join_rev.pop("object")
+s = data_filter_join_rev.pop("subject")
+data_filter_join_rev["subject"] = o
+data_filter_join_rev["object"] = s
+data_filter_join[data_filter_join['reverse']] = data_filter_join_rev[data_filter_join.columns]
+data_filter_join.pop("reverse")
+data_filter_join.pop("predicate")
+data_filter_join['predicate'] = data_filter_join.pop("pretty relation name")
+
+print(f"[after] : {len(data_filter_join)}")
+print(f"[entity]: {len(set(data_filter_join['object'].unique().tolist() + data_filter_join['subject'].unique().tolist()))}")
+print(f"[predicate]: {len(data_filter_join['predicate'].unique())}")
+
+
+data = [i.to_dict() for _, i in data_filter_join.iterrows()]
+with open(f"data/t_rex.filter_unified.jsonl", 'w') as f:
+    f.write('\n'.join([json.dumps(i) for i in data]))