Candidate Matching
A key part of the recruitment process is identifying and ranking candidates to meet a job specification.
Free text search is the traditional way of identifying the relevant skills and experience within a candidate profile and candidate profiles are ranked based on the number of results found. Whilst this method has been the dominant method for decades, it has significant drawbacks:
- Keywords might not identify the candidate. For instance, searching profiles for the job title “Director” would also find “reporting to the Director”, giving a false positive result
- Search terms can have contextual meaning. For instance “Windows” refers to both an architectural feature and a software operating system. There is no confusion about the term within the context of the candidate profile but without that context it has dual meaning
- Ranking is based on the number of search terms found, rather than the overall relevance of the candidate profile.
So how can Symilarity help?
Symilarity’s Helix product provides AI/Machine Learning based tools to improve the traditional search process.
Relevancy
Helix uses unsupervised machine learning to build a model of all the candidate profiles. This process (called Vector Space Modelling) creates a statistical association between all the words (terms) found in the documents and how strongly they are related. This means search terms are more than the words themselves, they also represent the related terms.
The more relevant a document is to the search term depends not only on the appearance of the term itself but also upon the related terms. This means that the ranking of the document is based on the relevance of the document, not just the number of occurrences of the search term.
Location
Whilst location may be less relevant with the increasing use of home working, some organisations still wish their staff to be within a certain geographic area. Candidates may demonstrate meeting these requirements based on their home location or through the location of their previous clients. This requires the recruiter to understand the relevant geography.
When ingesting candidate profiles, Helix can identify places referenced in the profile (the candidates home location or work locations) and add location coordinates (i.e. Latitude and longitude). Helix enables candidate profiles to be found based upon their geographic location.
Enhanced Free Text Search
Where free text searching is used, Helix provides additional functionality that can improve the search results.
For instance, search terms can accommodate misspellings by specifying the degree of deviation allowed. This is often referred to as “Fuzzy Matching”.
Search results can be based on the proximity of two search terms by specifying the maximum number of words between the two terms.
Also, within a set of search terms, it is possible to make some of those terms more important than others. This “boosts” the ranking of documents which include those more important terms.
Screenshots
