Working paper Open Access

Identifying AI talents among LinkedIn members, A machine learning approach

Thomas Roca

JSON Export

  "files": [
      "links": {
        "self": ""
      "checksum": "md5:cb98d32029907209e38510fa9f32990d", 
      "bucket": "3967dea9-0aa3-4274-821d-a53b8ac49e84", 
      "key": "AI in the Labour Force  Data for Policy June 2019.pdf", 
      "type": "pdf", 
      "size": 3583975
  "owners": [
  "doi": "10.5281/zenodo.3240963", 
  "stats": {
    "version_unique_downloads": 552.0, 
    "unique_views": 310.0, 
    "views": 343.0, 
    "version_views": 535.0, 
    "unique_downloads": 439.0, 
    "version_unique_views": 480.0, 
    "volume": 1702388125.0, 
    "version_downloads": 607.0, 
    "downloads": 475.0, 
    "version_volume": 2166873685.0
  "links": {
    "doi": "", 
    "conceptdoi": "", 
    "bucket": "", 
    "conceptbadge": "", 
    "html": "", 
    "latest_html": "", 
    "badge": "", 
    "latest": ""
  "conceptdoi": "10.5281/zenodo.2649207", 
  "created": "2019-06-07T07:31:29.301714+00:00", 
  "updated": "2020-01-20T17:26:59.939409+00:00", 
  "conceptrecid": "2649207", 
  "revision": 4, 
  "id": 3240963, 
  "metadata": {
    "access_right_category": "success", 
    "doi": "10.5281/zenodo.3240963", 
    "description": "<p>How to identify specific profiles among the&nbsp;hundred of millions gathered in LinkedIn?&nbsp;LinkedIn Economic Graph thrives on skills,<br>\naround 50 thousand of them are listed by&nbsp;LinkedIn and constitute one of the main signals&nbsp;to identify professions or trends. Artificial&nbsp;Intelligence (AI) skills, for example, can be&nbsp;used to identify the diffusion of AI in industries&nbsp;[16]. But the noise can be loud around&nbsp;skills for which the demand is high. Some&nbsp;users may add &quot;trendy&quot; skills on their profiles&nbsp;without having work experience or training&nbsp;related to them. On the other hand, some&nbsp;people may work in the broad AI ecosystem&nbsp;(e.g. AI recruiters, AI sales&nbsp;representatives,&nbsp;etc.), without being the AI practitioners we&nbsp;are looking for. Searching for keywords in profiles&#39;&nbsp;sections can lead to mis-identification of&nbsp;certain profiles, especially for those related to&nbsp;a field rather than an occupation. This is the<br>\ncase for Artificial Intelligence.&nbsp;In this paper, we propose a machine learning&nbsp;approach to identify such profiles, and suggest<br>\nto train a binary text-classifier using job offers&nbsp;posted on the platform rather than actual profiles.<br>\nWe suggest this approach allows to avoid&nbsp;manually labeling the training dataset, granted&nbsp;the assumption that job profiles posted by recruiters&nbsp;are more &quot;ideal-typical&quot; or simply provide&nbsp;a more consistent triptych &quot;job title, job&nbsp;description, associated skills&quot; than the ones&nbsp;that can be found among member&#39;s profiles.</p>", 
    "language": "eng", 
    "title": "Identifying AI talents among LinkedIn members, A machine learning approach", 
    "license": {
      "id": "CC-BY-4.0"
    "relations": {
      "version": [
          "count": 2, 
          "index": 1, 
          "parent": {
            "pid_type": "recid", 
            "pid_value": "2649207"
          "is_last": true, 
          "last_child": {
            "pid_type": "recid", 
            "pid_value": "3240963"
    "communities": [
        "id": "dfp17"
    "keywords": [
      "Artificial Intelligence", 
      "Machine learning", 
      "Natural Language Processing", 
      "Big Data"
    "publication_date": "2019-04-23", 
    "creators": [
        "affiliation": "Microsoft, Linkedin", 
        "name": "Thomas Roca"
    "access_right": "open", 
    "resource_type": {
      "subtype": "workingpaper", 
      "type": "publication", 
      "title": "Working paper"
    "related_identifiers": [
        "scheme": "doi", 
        "identifier": "10.5281/zenodo.2649207", 
        "relation": "isVersionOf"
All versions This version
Views 535343
Downloads 607475
Data volume 2.2 GB1.7 GB
Unique views 480310
Unique downloads 552439


Cite as