=Paper=
{{Paper
|id=Vol-2571/CSP2019_paper_19
|storemode=property
|title=Emplobot - Design of the System (short paper)
|pdfUrl=https://ceur-ws.org/Vol-2571/CSP2019_paper_19.pdf
|volume=Vol-2571
|authors=Paweł Drozda,Arkadiusz Talun,Leszek Bukowski
|dblpUrl=https://dblp.org/rec/conf/csp/DrozdaTB19
}}
==Emplobot - Design of the System (short paper)==
https://ceur-ws.org/Vol-2571/CSP2019_paper_19.pdf
Emplobot – Design of the System
Paweł Drozda1,2 and Arkadiusz Talun2 and Leszek Bukowski2
1 Faculty of Mathematics and Computer Science, University of Warmia and Mazury in Olsztyn,
Poland, pdrozda@matman.uwm.edu.pl
2 Emplocity Ltd., Warszawa, Poland
Abstract. This paper presents general overview of Emplobot recruiting system,
designed by Emplocity Ltd. as part of the project „Development of autonomous
artificial intelligence using the learning of deep neural networks with strength-
ening, automating recruitment processes”. The main goal of the described sys-
tem is to facilitate, accelerate and automate recruiting process. For this propose
the chatbot was implemented which can communicate with potential job candi-
dates preparing their virtual CV. Thereafter, it can perfectly match the candidate
profiles with jobs adds in completely automatic way.
Keywords: Recruiting System, Chatbot, Deep Learning.
1 Introduction
Looking for a job in almost any field of the economy is associated with the tedious
and often long-lasting process of browsing many on-line job portals, comparing re-
quirements, expectations and benefits for the desired position. This often causes a
reluctance of many people to look for their dream job or change current position for
the better one. On the other hand, searching perfect employee or even matching the
position requirements by recruiters in the traditional way is very hard and time con-
suming process. Hiring the right employee is often associated with many recruitment
interviews, analysis of candidate profiles, which for a potential employer is involves
long-term process and high costs. Additionally, in many cases, especially in the areas
of the employee market, recruitment processes fail.
The problems described above resulted in the development of automatic processes
for combining candidate CVs with available job offers. In [1], authors proposed to tag
CVs and job offers with semantic annotations using shared ontology. In enables to
match the candidates with adequate job offer. Other approach in [2] was described.
Authors introduced idea of joining GitHub developer profiles to job advertisements.
They extracted concepts from job adds and GitHub profiles and then determined the
similarity between vectors of concepts. Since it is based only on GitHub profile, it
may be applied with a very limited range. Next paper [3] concerning automatic job
adds and candidate profile matching uses natural language processing techniques with
the use of Human Resources Domain. Another approach was presented in [4]. The
authors used recommendation system with collaborative filtering. Proposed system
Copyright © 2019 for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0).
differentiates two databases of jobs: unskilled, low-paid and highly qualified. The last
two the state of the art solutions [5], [6] focus on the use of machine learning tech-
niques when combining job advertisements with job seekers.
All of the cited solutions consist of trying to match the job offers with candidate
CVs for limited databases of candidates and job offers without the possibility of mak-
ing further recruitment steps. It significantly limits the usefulness of these tools.
1.1 Emplocity system
The solution proposed in this paper is a significant innovation in relation to above
mentioned solutions, due to possible scalability. Database of job offers are daily ex-
panded by parsers working on the largest online portals with job advertisements in
Poland. On the other hand, virtual CVs are created through automatic chatbot conver-
sations with potential candidates. Chatbot is available in the Facebook messenger,
which allows you to potentially reach about 15 million people of working age in Po-
land. Moreover, Emplocity system allows further handling of the recruitment process.
2 Design of the System
One of the most important assumptions of the system is the automatic support of the
recruitment process 7 days a week, 24 hours a day with the possibility of servicing
thousands of potential candidates at the same time.
The main idea of event flow of the system and it’s main components is presented
on fig. 1.
50
45
40 Data A
35 Data B
30
25
20
15
10
5
0
0 5 10 15 20 25 30
Fig. 1. Design of the Emplocity System.
System consists of the following components: chatbot, which should be considered
the main one. He talks to potential candidates and prepares virtual CV. Thereafter, it
pass prepared CVs to matching component.
Next component is responsible for gathering the job adds from the biggest and the
most commonly used job offer web systems in Poland (as for example pracuj.pl). It
crawls thousands of job offers every day and feeds the system database.
Other component integrates the whole system. With the use of the artificial intelli-
gence algorithms (deep neural networks, etc.), this element combines job offers with
profiles of individual candidates and presents an anonymous recommendation taking
into account previous experiences and skills of candidates for the further recruitment
processes.
The last element of the system supports recruiting companies with the reports of
the scale and effectiveness of chatbot activities. It presents all recruitment processes
in one place along with the number of matched candidates and the status of their invi-
tation. Moreover, with the analytical panel of the system it is possible to monitor can-
didates’ preferred work location, number of candidates invited to further steps, aver-
age time of chat conversation, traffic on career website, percent of visitors which start
chat conversation.
3 Recruiting scenario
In order to show how the system works, an example scenario will be presented.
Let us suppose that certain company is looking to employ new senior Python de-
veloper. If it decide to use Emplocity system, it can add the address for Emplobot to
all promotional materials, websites, etc. as well as reach potential target group of
candidates by Facebook Ads Manager setting restrictions on:
• target country: Poland
• age: 25+
• job title: Python developer, senior Python developer, etc.
• degree: master in Computer Science, etc.
When the candidates start to talk with emplobot it will create virtual CVs containing
such information as work experience, interests, skills, education, etc. Thereafter, CVs
will be compared with job advertisements and the most appropriate candidates will be
proposed for a company.
The recruiter from the company can choose the best candidates to finish the whole
recruiting process. Moreover, he can check the statistics of chatbot working such as
number of candidates matches, average time of chat conversation, etc.
References
1. Yahiaoui, L., Boufaida, Z., Prie, Y.: Semantic Annotation of Documents Applied to e-
recruitment. Semantic Web Applications and Perspectives, Proceedings of the 3rd Italian
Semantic Web Workshop (2006).
2. Hauff, C., Gousios, G.: Matching GitHub Developer Profiles to Job AdvertisementsPro-
ceedings of the 2015 IEEE/ACM 12th Working Conference on Mining Software Reposito-
ries (MSR), p. 362 – 366, (2015).
3. Martinez-Gil, J, Paoletti, A.L., Schewe, K.D.: A Smart Approach for Matching, Learning
and Querying Information from the Human Resources Domain. New Trends in Databases
and Information Systems: ADBIS Short Papers and Workshops, Prague, Czech Republic
(2016).
4. Schmitt, T., Gonard, F., Caillou, P., Sebag, M.: Language Modelling for Collaborative Fil-
tering: Application to Job Applicant Matching. In IEEE International Conference on Tools
with Artificial Intelligence (2017).
5. Maheshwary, S., Misra, H.: Matching resumes to jobs via deep siamese network. InCom-
panion Proceedings of the The Web Conference 2018, WWW ’18, p. 87–88 (2018).
6. Yiou Lin, Hang Lei, Prince Clement Addo, Xiaoyu Li: Machine Learned Resume-Job
Matching Solution. CoRR abs/1607.07657 (2016).