skills to generate the recommendations, our initial results in-
dicate that an improved and better set of job recommenda-
tions are obtained on a locally established and previously
published dataset.
In addition, the paper proposed a generalizable ensemble
method for skill extraction from unstructured text of resumes
as well as JDs. Our ensemble consists of sub-modules such
as a Watson-driven NLU system (for extracting concepts,
keywords and entities), a PoS tagging system whose out-
put PoS patterns were mapped for skill identification and
an expandable dictionary whose base dictionary was seeded
from several online open source knowledge bases. The per-
formance of the proposed ensemble method was compared
with that of the manual annotators – an accuracy of more
than 0.90 and F1-score of more than 0.83 was obtained on
two different job datasets. Moreover, several scoring algo-
rithms were explored for matching skills extracted from re-
sumes with (explicit and implicit) skills extracted from JDs.
Due to non-availability of standard large open source
dataset for job recommendation task, we evaluated our sys-
tem on a dataset used in Maheshwary and Misra (2018).
Though our results on this dataset are better than the ones re-
ported in the original paper, our immediate next step would
involve using a more diverse dataset, a stronger evaluation
of the system by including more resumes and leveraging
additional techniques for skill identification and extraction.
Our future work in this space will involve generating ranked
recommendations on different career path options that opti-
mally utilize the accumulated skill and experience of a can-
didate. We also intend to use skill graph for inferring profes-
sional growth of a user and leveraging that for better recom-
mendations. This could help a professional in understanding
where he/she stands when compared to his/her peers. Appli-
cation of skill graph for professional growth inference could
also help in comparing two organizations in terms of profes-
sional growth of their employees. We propose using these
skill graphs to infer Skill-Gap in a candidate profile and use
this as an additional recommendation to the user. Addition-
ally the system can be used to analyze cost of acquiring a
skill and recommend better skills on which to get trained.
7 Acknowledgments
This work was guided by our colleague Danish Contractor
who provided expertise that greatly assisted this research.
The deployment work was supported by our colleagues Pad-
manabha Venkatagiri Seshadri and Vivek Sharma.
We are also immensely grateful to Renuka Sindhgatta and
Bikram Sengupta for their comments on an earlier versions
of the manuscript and support for this work.
References
Al-Otaibi, S. T., and Ykhlef, M. 2012. A survey of job
recommender systems. International Journal of Physical
Sciences 7(29):5127–5142.
Bastian, M.; Hayes, M.; Vaughan, W.; Shah, S.; Skomoroch,
P.; Kim, H.; Uryasev, S.; and Lloyd, C. 2014. Linkedin
skills: large-scale topic extraction and inference. In Proceed-
ings of the 8th ACM Conference on Recommender systems,
1–8. ACM.
Colucci, S.; Di Noia, T.; Di Sciascio, E.; Donini, F. M.;
Mongiello, M.; and Mottola, M. 2003. A formal approach
to ontology-based semantic match of skills descriptions. J.
UCS 9(12):1437–1454.
G
¨
olec¸, A., and Kahya, E. 2007. A fuzzy model for
competency-based employee evaluation and selection. Com-
puters & Industrial Engineering 52:143–161.
Gugnani, A.; Kasireddy, V. K. R.; and Ponnalagu, K. 2018.
Generating unified candidate skill graph for career path rec-
ommendation. In 2018 IEEE International Conference on
Data Mining Workshops (ICDMW), 328–333. IEEE.
Javed, F.; Hoang, P.; Mahoney, T.; and McNair, M. 2017.
Large-scale occupational skills normalization for online re-
cruitment. In AAAI, 4627–4634.
Kivim
¨
aki, I.; Panchenko, A.; Dessy, A.; Verdegem, D.;
Francq, P.; Bersini, H.; and Saerens, M. 2013. A graph-
based approach to skill extraction from text. Proceedings of
TextGraphs-8 Graph-based Methods for Natural Language
Processing 79–87.
Lau, T., and Sure, Y. 2002. Introducing ontology-based
skills management at a large insurance company. In In Pro-
ceedings of the Modellierung 2002, 123–134.
Lin, Y.; Lei, H.; Clement Addo, P.; and Li, X. 2016. Machine
Learned Resume-Job Matching Solution. ArXiv e-prints.
Liu, K.; Shi, X.; Kumar, A.; Zhu, L.; and Natarajan, P. 2016.
Temporal learning and sequence modeling for a job recom-
mender system. In Proceedings of the Recommender Sys-
tems Challenge, 7. ACM.
Maheshwary, S., and Misra, H. 2018. Matching resumes to
jobs via deep siamese network. In Companion of the The
Web Conference 2018 on The Web Conference 2018, 87–
88. International World Wide Web Conferences Steering
Committee.
Malinowski, J.; Keim, T.; Wendt, O.; and Weitzel, T. 2006.
Matching people and jobs: A bilateral recommendation ap-
proach. In HICSS.
Mikolov, T.; Sutskever, I.; Chen, K.; Corrado, G. S.; and
Dean, J. 2013. Distributed representations of words and
phrases and their compositionality. In Advances in neural
information processing systems, 3111–3119.
Mimno, D., and McCallum, A. 2008. Modeling career path
trajectories.
Paparrizos, I. K.; Cambazoglu, B. B.; and Gionis, A. 2011.
Machine learned job recommendation. In RecSys.
Salton, G., and Buckley, C. 1988. Term-weighting ap-
proaches in automatic text retrieval. Information processing
& management 24(5):513–523.
Zhao, M.; Javed, F.; Jacob, F.; and McNair, M. 2015. Skill:
A system for skill identification and normalization. In AAAI,
4012–4018.
13293