jisungyoon / Structure-of-Science

Science of Science, Collective Intelligence
2 stars 0 forks source link

language link check #35

Closed jisungyoon closed 5 years ago

jisungyoon commented 5 years ago

우선 새로추가한 위키 + 한국어 위키 대상으로 langlinks 갯수의 heatmap 을 그려보았습니다.

스크린샷 2019-03-18 오후 7 36 21

재밌네요 zh_yue, be_x_old 두 위키의 경우 나가는 랭기지링크는 있어도 들어오는 위키가 없네요... 데이터가 임포팅이 안된 건 아닙니다. 그러면 반대의 상황이 됬었어야.. 생각보다 lang_links는 서로 잘 연결 되어 있는 것 같습니다. (신기하게도... 누가 이런 걸 다 이어 놓는지) 우선 저 두 위키는 버리는게 맞을까요?

이제 어느정도 처리한다음 널모델(random bipartite network) 에 비해 얼마나 연결이 있는지 체크 해보겠습니다.

jisungyoon commented 5 years ago

저기 파란줄 tewiki는 거의 10000정도의 링크가 연결되어 있네요

jisungyoon commented 5 years ago

가장 작은 링크는 te to vo 2099 개 정도네요. 이런 위키도 제외해야할 듯합니다.

bluekura commented 5 years ago

이거 말고 분포도 볼 수 있을까요?

jisungyoon commented 5 years ago

이거 말고 분포도 볼 수 있을까요?

어떤 분포 말씀하시는건가요?

wsjung77 commented 5 years ago

number of langlinks 의 분포요?

bluekura commented 5 years ago

(신기하게도... 누가 이런 걸 다 이어 놓는지)

Bot이 먼저 하고, 사람이 건드리는 방식일겁니다... 현재는 아마도...

이거 말고 분포도 볼 수 있을까요?

어떤 분포 말씀하시는건가요?

저 count 자체의 분포를 좀 보고싶어요. 어느정도 선이면 "낮다" 라고 판단 가능한지

jisungyoon commented 5 years ago
스크린샷 2019-03-18 오후 8 12 15

이정도 입니다.

bluekura commented 5 years ago

혹시 rankplot으로 볼 수 있을까요? ccdf나...

wsjung77 commented 5 years ago

cumulative보다는 rank로 보는 게 나을 거 같긴 한데, 일단 둘 다 그려주시면 되겠네요. :)

jisungyoon commented 5 years ago

카운팅 코드에 버그가 있었네요. 여러 element들이 중복해서 들어가있던.. 제대로 된 피규어들입니다. hist

스크린샷 2019-03-18 오후 9 06 12

rank plot

스크린샷 2019-03-18 오후 9 06 06

ccdf

스크린샷 2019-03-18 오후 9 10 38
bluekura commented 5 years ago

그리신게 ccdf가 아니라 cdf인것 같긴 하지만... 일단 그리신 그림들을 종합해보면 제 생각에 평균적으로 만개 이하면 충분히 "작다" 라는 생각이 들긴 하네요 들어오는/나가는 링크 기준 평균적으로 만개 이하인 위키들은 일단 다 지울까요?

jisungyoon commented 5 years ago

그리신게 ccdf가 아니라 cdf인것 같긴 하지만...

아 .. 졸려서 잘못 생각했군요..

일단 그리신 그림들을 종합해보면 제 생각에 평균적으로 만개 이하면 충분히 "작다" 라는 생각이 들긴 하네요 들어오는/나가는 링크 기준 평균적으로 만개 이하인 위키들은 일단 다 지울까요?

들어오는 기준, 나가는 기준 둘다 (and condition) 적용해서 지우도록 하겠습니다. 아마 거의 symmetry해서 비슷하긴 할 것 같습니다. 우선은 내일 출근해서 자른 결과 공유하도록 하겠습니다!

wsjung77 commented 5 years ago

이제 뭔가 많이 보던 그림의 형태이군요. :) 만 개면 충분하겠네요.

jisungyoon commented 5 years ago

생각보다 많이 짤리지는 않군요 index|wiki_name|source_lang_links_mean|target_lang_links_mean| 입니다.

39 zh_yue 26581.2241379 0.0 41 te 8181.18965517 8387.25862069 52 be_x_old 30243.6551724 0.0

우선 이 3개 정도만 잘라버릴까요?

jisungyoon commented 5 years ago

물론 비교할 수 있는 lang_links는 이것보다 더 작을 수도 있습니다. 양쪽링크가 S&T network에 다 들어와야 하기때문에..

bluekura commented 5 years ago

생각보다 많이 짤리지는 않군요 index|wiki_name|source_lang_links_mean|target_lang_links_mean| 입니다.

39 zh_yue 26581.2241379 0.0 41 te 8181.18965517 8387.25862069 52 be_x_old 30243.6551724 0.0

우선 이 3개 정도만 잘라버릴까요?

네 이정도로 하지요...

bluekura commented 5 years ago

생각보다 많이 짤리지는 않군요 index|wiki_name|source_lang_links_mean|target_lang_links_mean| 입니다. 39 zh_yue 26581.2241379 0.0 41 te 8181.18965517 8387.25862069 52 be_x_old 30243.6551724 0.0 우선 이 3개 정도만 잘라버릴까요?

네 이정도로 하지요...

3가지를 보니 3가지가 다 특정 국가의 방언계통이네요 전체 지역의 공용어가 아닌 언어를 모두 지울까요?

bluekura commented 5 years ago

생각보다 많이 짤리지는 않군요 index|wiki_name|source_lang_links_mean|target_lang_links_mean| 입니다. 39 zh_yue 26581.2241379 0.0 41 te 8181.18965517 8387.25862069 52 be_x_old 30243.6551724 0.0 우선 이 3개 정도만 잘라버릴까요?

네 이정도로 하지요...

3가지를 보니 3가지가 다 특정 국가의 방언계통이네요 전체 지역의 공용어가 아닌 언어를 모두 지울까요?

생각해보니 이렇게 되면 Cebuano가 지워지겠네요... 얘도 조금 이상했죠 behavior가? (유저대비 글 수라던지...)

jisungyoon commented 5 years ago

넵 Cebuano도 매우 이상하긴 했습니다 하지만 그런 논리라면 언어 - 국가 매칭이 되어야...

jisungyoon commented 5 years ago

20000개 기준으로 끊으면

23 mr 17522.7241379 18109.0344828 28 ml 18810.5172414 19263.3275862 39 zh_yue 26581.2241379 0.0 41 te 8181.18965517 8387.25862069 46 jv 16355.0862069 16956.6896552 52 be_x_old 30243.6551724 0.0 55 ga 18106.9482759 18716.9482759

정도를 날릴 수 있습니다.

jisungyoon commented 5 years ago

사실 절대적 수치보다는 원래 몸뚱이 기준 몇개나 연결되어 있을까가 좀 더 의미 있을 수 도 있을 것 같네요

bluekura commented 5 years ago

20000의 경우도 아일랜드어 말고는 다 방언이군요...

jisungyoon commented 5 years ago

20000의 경우도 아일랜드어 말고는 다 방언이군요...

음 일단 원래 몸뚱이 기준 얼마나 있는지도 한번 데이터를 뽑아 와보겠습니다!

jisungyoon commented 5 years ago
스크린샷 2019-03-19 오후 7 21 54

이 피규어 기준으로 거를 수 있을 것 같은 느낌이드네요. 한칸 한칸의 값은 다음과 같습니다 lang i, lang j 가 있을때 element(i,j) = (number_of_lang_links_i_to_j) /((number_of_nodes_in_network_i) * (number_of_nodes_in_network_j)) 입니다. 그리고 네트워크가 없으면 분모가 0이기 때문에 bad 처리하였습니다.

따라서 회색 줄은 네트워크가 없거나 lang_link가 없을때 나타나게 되며 하얀색 줄은 원래 네트워크가 이상하게 만들어졌을 경우 (네트워크 구축 실패의 케이스) 입니다. threshold를 정하기전에 저런 애들을 없애고 우선 보는게 어떨까요?

jisungyoon commented 5 years ago

궁금해 하실까바 각 네트워크별 크기는 다음과 같습니다.

{'af': 79479, 'ast': 42691, 'az': 193388, 'azb': 63306, 'be': 225706, 'be_x_old': 98902, 'bg': 303886, 'bn': 115403, 'br': 5062, 'bs': 114643, 'ca': 674575, 'ceb': 0, 'cy': 124742, 'da': 291853, 'el': 205223, 'eo': 308239, 'et': 197503, 'eu': 354017, 'ga': 54126, 'gl': 206363, 'he': 299628, 'hi': 138195, 'hr': 199983, 'hy': 303787, 'id': 597187, 'jv': 58891, 'ka': 164691, 'kk': 226944, 'ko': 678624, 'la': 160051, 'lb': 0, 'lt': 210819, 'lv': 112990, 'mk': 0, 'ml': 81660, 'mr': 62800, 'ms': 333024, 'nn': 201295, 'no': 678167, 'ro': 507569, 'sco': 85369, 'sh': 460908, 'simple': 178693, 'sk': 286875, 'sl': 209773, 'sq': 3, 'sr': 613879, 'sv': 4879164, 'ta': 0, 'te': 47, 'tl': 86841, 'tr': 515182, 'uk': 1057032, 'ur': 464973, 'uz': 201094, 'vi': 1416588, 'vo': 0, 'zh_yue': 77024}

bluekura commented 5 years ago

이 피규어 기준으로 거를 수 있을 것 같은 느낌이드네요. 한칸 한칸의 값은 다음과 같습니다 lang i, lang j 가 있을때 element(i,j) = (number_of_lang_links_i_to_j) /((number_of_nodes_in_network_i) * (number_of_nodes_in_network_j)) 입니다. 그리고 네트워크가 없으면 분모가 0이기 때문에 bad 처리하였습니다.

여기서의 network라 하심은...? 어떤 네트워크를 의미하시나요.

jisungyoon commented 5 years ago

이 피규어 기준으로 거를 수 있을 것 같은 느낌이드네요. 한칸 한칸의 값은 다음과 같습니다 lang i, lang j 가 있을때 element(i,j) = (number_of_lang_links_i_to_j) /((number_of_nodes_in_network_i) * (number_of_nodes_in_network_j)) 입니다. 그리고 네트워크가 없으면 분모가 0이기 때문에 bad 처리하였습니다.

여기서의 network라 하심은...? 어떤 네트워크를 의미하시나요.

아 만들어진 S&T network 입니다. 저희 방식대로 만들었던

bluekura commented 5 years ago

세부어가 과학기술 링크가 하나도 없나요? 신기하네요 여튼 알겠습니다... 근데 생각해보니 한쪽에만 링크가 있는건 좀 이상하니 undirected로 일단 projection을 해야 하는게 아닌가 싶기도 하고...

jisungyoon commented 5 years ago

세부어가 과학기술 링크가 하나도 없나요? 신기하네요 여튼 알겠습니다...

여기서 0인 경우는 저번에도 말씀드렸다 싶이 (과학 혹은 기술) 노드로 연결되어 있는 노드가 없는 경우 입니다. 둘중에 하나라도 없으며 네트워크를 못만들었으니 0입니다.

그리고 0 초과의 매우 작은 노드갯수를 가지는 경우는 보통 과학기술이 잘못 링크가 되어 있으면 나타나는 현상중 하나입니다

근데 생각해보니 한쪽에만 링크가 있는건 좀 이상하니 undirected로 일단 projection을 해야 하는게 아닌가 싶기도 하고...

projection을 하면 아마 저기에서 광동어랑 be-x-old도 살릴수 있겠으나 한방향의 데이터만을 가지고 만드는게 말이되나 싶기도 하네요

bluekura commented 5 years ago

오케이 그럼 뭐 일단 저기서 에러처럼 보이면 다 날립시다.

jisungyoon commented 5 years ago
스크린샷 2019-03-20 오전 10 07 28

에러를 제외하고 봤을 때 이렇게 정도 나오는 것 같습니다.

jisungyoon commented 5 years ago

저기에 파란 줄 같은경우에는 스웨덴 위키의 네트워크가 워낙 커서 4,879,164 인것 같고 하얀 줄 같은 경우에는 br 네트워크의 크기가 5000이라 그런 것 같습니다. br도 날려볼까 싶은데 어떠신가요?

bluekura commented 5 years ago

브류타뉴어 자체가 프랑스에서 쓰는 소수어라서 잘라도 큰 문제는 안될 것 같습니다 (일전의 공식어가 아닌 것들에 대한 문제 포함...)

2019년 3월 20일 (수) 오후 11:11, balla2081 notifications@github.com님이 작성:

저기에 파란 줄 같은경우에는 스웨덴 위키의 네트워크가 워낙 커서 4,879,164 인것 같고 하얀 줄 같은 경우에는 br 네트워크의 크기가 5000이라 그런 것 같습니다. br도 날려볼까 싶은데 어떠신가요?

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/balla2081/Structure-of-Science/issues/35#issuecomment-474846634, or mute the thread https://github.com/notifications/unsubscribe-auth/AZDpXEdhRU-vMDfAyJfrFnVul31uRQfoks5vYkGXgaJpZM4b6_Oi .

-- Jinhyuk Yun, Ph. D. Dept. of Scientometric Research, Future Information Research Center Korea Institute of Science and Technology Information TEL: +82 2 3299 6061 MAIL: jinhyuk.yun@kisti.re.kr

jisungyoon commented 5 years ago

전체 셋은

"ga" "th" "pl" "en" "hi" "ar" "bn" "vi" "sk" "eo" "fi" "af" "bg" "kk" "zh" "fa" "simple" "it" "fr" "no" "tl" "la" "ro" "azb" "cs" "ko" "el" "es" "uz" "id" "sv" "eu" "ca" "et" "sco" "uk" "hy" "gl" "be" "ast" "de" "nn" "lt" "da" "hr" "sh" "ru" "ur" "sr" "hu" "az" "ja" "ml" "sl" "ms" "he" "tr" "jv" "mr" "pt" "nl" "cy" "lv" "bs" "ka"

으로 정해졌고, language_link parsing에 들어 갔습니다. 아마 데이터베이스를 많이 건드려야 하는 작업이라 하루정도 소요될 것으로 예상 됩니다. :)

bluekura commented 5 years ago

고생하십니다 :>

jisungyoon commented 5 years ago

결론을 먼저 말씀드리자면 azb, ast에서 나가는 링크들이 수상한 것 같습니다.

스크린샷 2019-03-21 오후 6 39 25

이 피규어는 랭귀지랭크 파싱 중 빠지게 되는 숫자를 표시한 피규어입니다.

A -> B 로 링크를 쏠때 파싱이 안되는 이유는 명료합니다.

  1. A에 있는 문서가 B의 네트워크에 없는 경우
  2. 이상한 B의 문서에 없는 경우

만약 1번 문제가 주류였다면 azt, ast는 십자가 처럼 줄이 처져있어야 합니다. 하지만 그림을 보면 azb, ast 에서 나가는 것만 문제가 있어보입니다. 아마 이상 현상이지 않을까 싶고 (2번 이유) 링크를 이상하게 건 것 같다는 의심이 드는군요

jisungyoon commented 5 years ago
스크린샷 2019-03-21 오후 6 39 36

이 피규어는 성공한 lang_links의 갯수를 나타낸 것이며

스크린샷 2019-03-21 오후 6 39 42

이 피규어는 실패한 비율을 나타낸 것 입니다. 역시 말하고 있는 바는 위에 언급한것과 같은 것 같습니다.

bluekura commented 5 years ago

찾아보니 두개 다 국가 공용어가 아닌 소규모 지역에서만 사용된 언어군요. 빼 두고 왜 뺐는지를 논문에 잘 적어둡시다 :)

2019년 3월 22일 (금) 오후 1:00, balla2081 notifications@github.com님이 작성:

[image: 스크린샷 2019-03-21 오후 6 39 36] https://user-images.githubusercontent.com/10040328/54799842-4e2c1400-4c35-11e9-876e-2cb9b971e9d2.png 이 피규어는 성공한 lang_links의 갯수를 나타낸 것이며

[image: 스크린샷 2019-03-21 오후 6 39 42] https://user-images.githubusercontent.com/10040328/54799851-5ab06c80-4c35-11e9-9217-725cdce63dcb.png 이 피규어는 실패한 비율을 나타낸 것 입니다. 역시 말하고 있는 바는 위에 언급한것과 같은 것 같습니다.

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/balla2081/Structure-of-Science/issues/35#issuecomment-475486359, or mute the thread https://github.com/notifications/unsubscribe-auth/AZDpXEtgoQKUlqyp9GeLQeKCyttdNziEks5vZFVJgaJpZM4b6_Oi .

jisungyoon commented 5 years ago

zh_yue와 be_x_old으로 들어가는 langlinks 가 왜 없는지 찾았습니다. ㅜㅜ mysql 데이터베이스 이름을 정할 때 '-'가 금지되어 있어서 ''로 치환해놨었는데, batch code를 짤 때 이 케이스를 잊어서 ㅜㅜ 금방 복구 시켜놓겠습니다.

jisungyoon commented 5 years ago
스크린샷 2019-03-22 오전 9 40 42

이 피규어가 자를 수 있는 기준인 것 같습니다. 총 7개의 위키가 짤렸으며 이유는 https://github.com/balla2081/Structure-of-Science/wiki/Data-set-record 여기에 저장 해놓았습니다. 앞으로도 이 곳에 업데이트 하도록 하겠습니다.

jisungyoon commented 5 years ago

lang_links 파싱이 끝났습니다. 총 68개의 위키 입니다. Mistarget이 되는 경우는 세 가지 경우로 압축 될 수 있습니다.

  1. A->B로 링크를 쏘았는데, B가 없는 경우 (연결이 잘못됨)
  2. B가 위키에는 있는데, S&T network에는 없는 경우
  3. (ambiguous) 처럼 테그가 달려져있는 경우
jisungyoon commented 5 years ago

Mistargeted_Langlinks

jisungyoon commented 5 years ago

success_Langlinks

jisungyoon commented 5 years ago

Mistargeted_Langlinks_ratio

jisungyoon commented 5 years ago

Ratio 피규어를 보면 세로 하얀 줄들을 볼 수 있는데, 이는 incoming langlinks 들이 잘못 된 경우입니다. 이런 위키 같은 경우는 의심군으로 분류해두었고, 후속 데이터 분석을 진행 후 이상하면 제거하는 방향으로 진행할 예정입니다.

bluekura commented 5 years ago

Good! 고생하셨습니다.

  • A->B로 링크를 쏘았는데, B가 없는 경우 (연결이 잘못됨)

이거 혹시 인코딩 이슈일 가능성은 없나요? 제 기억에 source는 id인데, target은 article name이었던 기억이... ^^;

jisungyoon commented 5 years ago

Good! 고생하셨습니다.

  • A->B로 링크를 쏘았는데, B가 없는 경우 (연결이 잘못됨)

이거 혹시 인코딩 이슈일 가능성은 없나요? 제 기억에 source는 id인데, target은 article name이었던 기억이... ^^;

인코딩 이슈는 다 확인하였으며 (이 경우는 거의 모든 문서가 fail 하기 때문에 확인가능합니다.) articlename 이라서 생기는 이슈 중에 대부분은 앞에 '분류'같은 prefix 가 붙기 때문인데, 이 문제에서도 손수 prefix를 다 찾아서 처리 하엿습니다. (거의 수작업..)

jisungyoon commented 5 years ago

prefix_dict['en'] = 'Category:' prefix_dict['de'] = 'Kategorie:' prefix_dict['ru'] = 'Категория:' prefix_dict['es'] = 'Categoría:' prefix_dict['ja'] = 'Category:' prefix_dict['zh'] = 'Category:' prefix_dict['fr'] = 'Catégorie:' prefix_dict['it'] = 'Categoria:' prefix_dict['pt'] = 'Categoria:' prefix_dict['pl'] = 'Kategoria:' prefix_dict['nl'] = 'Categorie:' prefix_dict['ar'] = 'تصنيف:' prefix_dict['ko'] = '분류:' prefix_dict['cs'] = 'Kategorie:' prefix_dict['fa'] = 'رده:' prefix_dict['he'] = 'קטגוריה:' prefix_dict['hu'] = 'Kategória:' prefix_dict['th'] = 'หมวดหมู่:' prefix_dict['fi'] = 'Luokka' prefix_dict['bg'] = 'Категория:' prefix_dict['uk'] = 'Категорія:' prefix_dict['no'] = 'Kategori:' prefix_dict['eo'] = 'Kategorio:' prefix_dict['bs'] = 'Kategorija:' prefix_dict['la'] = 'Categoria:' prefix_dict['da'] = 'Kategori:' prefix_dict['ast'] = 'Categoría:' prefix_dict['tr'] = 'Kategori:' prefix_dict['sr'] = 'Категорија:' prefix_dict['uz'] = 'Turkum:' prefix_dict['sco'] = 'Category:' prefix_dict['lv'] = 'Kategorija:' prefix_dict['el'] = 'Κατηγορία:' prefix_dict['et'] = 'Kategooria:' prefix_dict['mr'] = 'वर्ग:' prefix_dict['vi'] = 'Thể loại:' prefix_dict['af'] = 'Kategorie:' prefix_dict['hi'] = 'श्रेणी:' prefix_dict['ms'] = 'Kategori:' prefix_dict['az'] = 'Kateqoriya:' prefix_dict['be'] = 'Катэгорыя:' prefix_dict['sv'] = 'Kategori:' prefix_dict['ca'] = 'Categoria:' prefix_dict['gl'] = 'Categoría:' prefix_dict['lt'] = 'Kategorija:' prefix_dict['eu'] = 'Kategoria:' prefix_dict['tl'] = 'Kategorya:' prefix_dict['nn'] = 'Kategori:' prefix_dict['sk'] = 'Kategória:' prefix_dict['sh'] = 'Kategorija:' prefix_dict['ro'] = 'Categorie:' prefix_dict['hr'] = 'Kategorija:' prefix_dict['ga'] = 'Catagóir:' prefix_dict['hy'] = 'Կատեգորիա:' prefix_dict['azb'] = 'بؤلمه:' prefix_dict['simple'] = 'Category:' prefix_dict['kk'] = 'Санат:' prefix_dict['cy'] = 'Categori:' prefix_dict['jv'] = 'Kategori:' prefix_dict['sl'] = 'Kategorija:' prefix_dict['id'] = 'Kategori:' prefix_dict['ml'] = 'വർഗ്ഗം:' prefix_dict['ur'] = 'زمرہ:' prefix_dict['ka'] = 'კატეგორია:' prefix_dict['bn'] = 'বিষয়শ্রেণী:' prefix_dict['zh_yue'] = 'Category:' prefix_dict['be_x_old'] = 'Катэгорыя:' prefix_dict['ta'] = 'பகுப்பு:'

이렇게 구성되어 있고, 파싱을 할때 코드를 통해서 띄어주었습니다.

jisungyoon commented 5 years ago

최종 머징 프로세스까지 마쳤습니다. 그런데 문제가 하나 있네요. 이렇게 엄청난 many to many 관계가 관측이 됩니다. sh_mr_counter.pkl (9, 13) eo_mr_counter.pkl (9, 13) id_mr_counter.pkl (9, 13) ga_mr_counter.pkl (8, 13) la_mr_counter.pkl (9, 13) fr_mr_counter.pkl (9, 13) zh_yue_mr_counter.pkl (9, 13) sv_pt_counter.pkl (21, 17) sv_sr_counter.pkl (21, 16) en_sv_counter.pkl (1, 14) en_sv_counter.pkl (3, 21) en_sv_counter.pkl (1, 25) en_sv_counter.pkl (1, 56) en_sv_counter.pkl (1, 89) en_sv_counter.pkl (2, 11) fr_sv_counter.pkl (2, 21) fa_mr_counter.pkl (9, 13) af_mr_counter.pkl (9, 13) el_mr_counter.pkl (9, 13) pl_fr_counter.pkl (12, 1) jv_mr_counter.pkl (5, 13) pl_en_counter.pkl (12, 1) cs_mr_counter.pkl (9, 13) he_mr_counter.pkl (9, 13) ru_mr_counter.pkl (9, 13) de_mr_counter.pkl (9, 13) ru_pt_counter.pkl (1, 13) en_vi_counter.pkl (1, 13) en_vi_counter.pkl (26, 1) en_vi_counter.pkl (1, 14) ro_mr_counter.pkl (9, 13) gl_mr_counter.pkl (9, 13) en_ka_counter.pkl (12, 1) en_ja_counter.pkl (12, 1) sv_mr_counter.pkl (9, 13) en_hi_counter.pkl (16, 62) en_hi_counter.pkl (15, 15) en_hi_counter.pkl (15, 16) hi_fi_counter.pkl (18, 18) vi_mr_counter.pkl (9, 13) fi_mr_counter.pkl (9, 13) ja_mr_counter.pkl (9, 13) hi_sr_counter.pkl (17, 18) id_ms_counter.pkl (5, 11) id_ms_counter.pkl (13, 14) en_nl_counter.pkl (22, 1) az_mr_counter.pkl (9, 13) pl_sv_counter.pkl (1, 89) es_mr_counter.pkl (9, 13) th_mr_counter.pkl (9, 13) tr_mr_counter.pkl (9, 13) en_tl_counter.pkl (29, 30) en_tl_counter.pkl (2, 11) en_tl_counter.pkl (24, 1) en_tl_counter.pkl (9, 12) pl_sl_counter.pkl (12, 1) mr_lv_counter.pkl (13, 9) ml_mr_counter.pkl (9, 13) mr_pt_counter.pkl (13, 9) uk_mr_counter.pkl (9, 13) pl_mr_counter.pkl (9, 13) it_mr_counter.pkl (9, 13) kk_ko_counter.pkl (12, 1) en_ms_counter.pkl (73, 100) en_ru_counter.pkl (26, 1) hr_mr_counter.pkl (9, 13) en_pt_counter.pkl (1, 13) mr_cy_counter.pkl (13, 9) ar_mr_counter.pkl (9, 13) ms_mr_counter.pkl (9, 13) vi_sv_counter.pkl (1, 21) da_mr_counter.pkl (9, 13) sr_mr_counter.pkl (9, 13) simple_mr_counter.pkl (8, 13) nn_mr_counter.pkl (9, 13) hi_vi_counter.pkl (18, 18) sl_mr_counter.pkl (9, 13) ca_mr_counter.pkl (9, 13) no_mr_counter.pkl (9, 13) en_ml_counter.pkl (1, 12) en_de_counter.pkl (15, 1) en_de_counter.pkl (26, 1) en_de_counter.pkl (14, 1) en_de_counter.pkl (11, 1) bn_mr_counter.pkl (9, 13) en_mr_counter.pkl (9, 13) hu_mr_counter.pkl (9, 13) ko_mr_counter.pkl (9, 13) bg_mr_counter.pkl (9, 13) en_zh_counter.pkl (14, 14) en_zh_counter.pkl (1, 24) lt_mr_counter.pkl (9, 13) mr_nl_counter.pkl (13, 9) eu_mr_counter.pkl (9, 13) hy_mr_counter.pkl (4, 13) sk_mr_counter.pkl (9, 13) et_mr_counter.pkl (9, 13) be_mr_counter.pkl (9, 13) zh_mr_counter.pkl (9, 13)

이정도의 관계까지 파싱이 되는걸 확인했습니다. 사실 이 친구들이 그렇게 큰 영향을 미치지는 않을 텐데, 이 애들을 빼야하는지 고민이 되네요

bluekura commented 5 years ago

일단 이런 것을 알아만 두고 이건 그대로 진행하죠... :)

2019년 3월 26일 (화) 오전 4:48, Jisung Yoon notifications@github.com님이 작성:

관계가 10이 넘어가는 쌍을 확인했을 때 모두 de to tr...

— You are receiving this because you were assigned. Reply to this email directly, view it on GitHub https://github.com/balla2081/Structure-of-Science/issues/35#issuecomment-476350601, or mute the thread https://github.com/notifications/unsubscribe-auth/AZDpXIvfgBpFzjdSruCN5R4gpko7vOTjks5vaSgNgaJpZM4b6_Oi .