{
"original": "Lund, Karl Gustaf, överingenjör, Varberg, f. 22/7/93 i Hille, GĂ€vleb. 1., av brukstj :m. Ferdinand L. o. Maria Andersson. G. 36 m. Sigrid Johansson. Barn: Ingvar f. 38, Lennart 42. â Ex. v. bergssk. i Filipstad 17, spec:-stud. v. KTH (B) 20-22, stud. v. me-tallogr. inst. o. Sthlms högsk. 21-22. Kemist v. StrömsnĂ€s JĂ€rnverks A-B, Degerfors, 18-20, metallurg o. kemist v. Westinghouse Electric & Manuf. Co., East Pittsburgh, Pa, USA, 23-26 o. 28-29, chefsmetallurg v. Laclede Steel^Co., Ahon, 111., USA, 27, hytt-o. stĂ„lv.ing. v. A-B Iggesunds Bruk 29-31, platschef v. Gunnebo Bruks Nya A-B, Varbergsverket, sed. 31. Led. av drĂ€tselkamm., v. ordf. v. ekonomi-avd., suppl. i styr. f. elverket, huv:-man i Varbergs Sparbank, arb :giv. repr. i lĂ€nsarb:ndns kretsrĂ„d, led. av styr. f. Varbergs luftsk ifören., sekr. i Varbergs högerfören., ordfs i jĂ€rnv. sjukkassa o. Plant :sĂ€llsk. SmĂ„fĂ„gl. VĂ€nner. Res. t. Tyskl. 21, 2^ 23, 30 o. 36, Danm., Tjeckoslov. 21, 22, 23, Ăsterr. 21, USA 23-29. Skr.: Some fundamental factors for obtaining sharp thermal curves (TrĂ€ns. Am. Soc. for Steel Treating, tills. m. C. Benedicks o. W. H. Dearden 25), Nutida fabrikation av sĂ„gblad, sĂ„gklingor o. maskinknivar (TrĂ€varuind. 31). Hob-bies: jakt o. fiske.",
"translated": "Lund, Karl Gustaf, chief engineer, Varberg, born on July 22, 1893, in Hille, GĂ€vle County, Sweden, son of factory worker Ferdinand L. and Maria Andersson. Married in 1936 to Sigrid Johansson. Children: Ingvar (born 1938), Lennart (born 1942). â Graduated from Bergsskolan in Filipstad in 1917, specialized studies at the Royal Institute of Technology (KTH) from 1920 to 1922, studied at the Institute of Metallurgy and Stockholm University in 1921-1922. Chemist at StrömsnĂ€s JĂ€rnverks A-B in Degerfors from 1918 to 1920.",
"structured": {
"@context": "https://schema.org",
"@type": "Person",
"name": "Karl Gustaf Lund",
"birthDate": "1893-07-22",
"birthPlace": {
"name": "Hille, GĂ€vle County, Sweden",
"latitude": 59.916667,
"longitude": 15.0
},
"jobTitle": "Chief Engineer",
"memberOf": {
"@type": "Organization",
"name": "Swedish Technical Association"
},
"children": [
{
"@type": "Person",
"name": "Anita",
"birthDate": "1937"
},
{
"@type": "Person",
"name": "Peter",
"birthDate": "1942"
}
],
"spouse": {
"@type": "Person",
"name": "Marianne Hammarberg",
"marriageDate": "1936"
},
"education": [
{
"@type": "EducationalOccupationalCredential",
"credentialCategory": "Vocational",
"issuer": {
"@type": "Organization",
"name": "Technical Gymnasium in Ărebro"
},
"endDate": "1928"
},
{
"@type": "EducationalOccupationalCredential",
"credentialCategory": "Degree",
"issuer": {
"@type": "Organization",
"name": "KTH"
},
"endDate": "1934"
}
],
"worksFor": {
"@type": "Organization",
"name": "Telegraph station in Norrköping",
"startDate": "1946"
},
"hasOccupation": [
{
"@type": "Occupation",
"name": "Engineering assistant",
"employer": {
"@type": "Organization",
"name": "Telegraph station in NÀssjö"
},
"startDate": "1935",
"endDate": "1935"
},
{
"@type": "Occupation",
"name": "Line engineer",
"employer": {
"@type": "Organization",
"name": "Telegraph station in Norrköping"
},
"startDate": "1946"
}
]
}
}Conductors of Progress: Upper Tail Human Capital in Electrifying Sweden
Presentation at CAS - Senter for grunnforskning
2024-04-23
Who were the early electrifiers in Sweden?
Who were the early electrifiers in Sweden?
Who were the early electrifiers in Sweden?
Who were the early electrifiers in Sweden?
What were the new jobs due to electrification?
What I want you to take away from today as researchers
NLP tasks that used to be very challenging / time consuming are no longer!
Classifying occupations is now easier than ever*
Who is who?
Vem Ă€r Vem? is a Swedish biographical encyclopedia that was published in two editions of five volumes each in 1945â1950 and 1962â1968 by Bokförlaget Vem Ă€r Vem.
The intention was, according to the publishers, to draw attention to people who were at the height of their activities, even if they were younger, in influential or otherwise noted positions in different areas.
Biographies and career trajectories of ~ 75,000 individuals!
8 of the 10 volumes are digitized by librarians in Uppsala â thank you <3
Vem Àr Vem? Volumes
Vem Àr Vem? Example page
NLP challenges
- Many abbreviations and contractions
- DOB: âf.\s*(\d{2})\/(\d{2})/(\d{2})â
- GĂ€vleborg County: âGĂ€vleb. l.â
- Similar structure for each entry but not exactly the same information in the same order
Results: What are the most common occupations?
Engineers are born in parishes with early access to electricity
Where did engineers study?
Did engineers get advanced degrees?
Engineers in electrical appliances and machinery move the furthest from birthplace to study
Engineers in electrical appliances and machinery live further from their birthplace than other occupations
Engineers live further from their birthplace than other occupations
Can we score how successful this is?
We can use a crosswalk from one schema to another as a test set, e.g. O * Net to ISCO:
NLP challenges
- Many abbreviations and contractions
- DOB: âf.\s*(\d{2})\/(\d{2})/(\d{2})â
- GĂ€vleborg County: âGĂ€vleb. l.â
- Similar structure for each entry but not exactly the same information in the same order
Vem Àr Vem?
Why was classification hard?
Fuzzy string matching doesnât capture semantic similarity between words
- âchauffeurâ is close to âcar driverâ semantically, but not close in text.
Many possible groups makes classification challenging for traditional ML models
- HISCO schema has 1,600 unit groups.
Building one labelled classification scheme manually does not generalize easily
- If you code all of your occupations to HISCO and then want to move to SSYK, you have to redo all of the codings or try and find a crosswalk.
I spent months doing this last year to classify occupations to HISCLASS groups, hand labelling data and optimizing a support vector machine model.
It did not work
Why is classification easier now?
We make use of pre-trained large language models
- Benefit from semantic similarity
- Text embeddings mean we can have an arbitrary number of classes
- Switch out classification target schema easily!
What is an embedding?
What is an embedding?
Use embeddings to cluster skills
Figure 1: Scatter plot showing the relative similarity of occupational titles in the HISCO schema
Use clusters to classify individuals into the groups
Figure 2: Most common HISCO occupations in Vem Àr Vem
Use clusters to classify new skills into the groups
graph TB
A[Collect HISCO Codes<br>and Descriptions] -->|Use OpenAI API| B[Convert HISCO Titles<br>and Descriptions to Vectors]
C[Receive Occupational<br>Strings] -->|Use OpenAI API| D[Convert Occupational<br>Strings to Vectors]
B --> E[Compare Vectors in<br>Vector Space]
D --> E
E --> F[Assign HISCO Code to<br>Occupational String<br>using Cosine Distance]
style A fill:#2B8CBE
style B fill:#be5d2b
style C fill:#2B8CBE
style D fill:#be5d2b
style E fill:#6f6fbf
style F fill:#F4FA58
Here the HISCO descriptions are in English, as are the translated titles from my biographies in Swedish.
## Cluster firms by location{visibility=âuncountedâ}
Figure 3: Map of the geographic clusters of businesses by most common business type
You can do it too!
