Report, 27 October 2021
In the last few years, the natural language processing community witnessed the advances
in neural representations of free texts with transformer-based language models
(LMs). Given the importance of knowledge available in relational tables, recent
research efforts extend LMs by developing neural representations for tabular data. In
this work, we present the first survey that analyzes these efforts. We first categorize
the downstream tasks where the models are successfully utilized. The alternative solutions
are then characterized and compared in terms of training data, input pre-processing,
model training, and output representation. Finally, we provide insights on potential future
work directions.
Type:
Report
Date:
2021-10-27
Department:
Data Science
Eurecom Ref:
6721
Copyright:
© EURECOM. Personal use of this material is permitted. The definitive version of this paper was published in Report, 27 October 2021 and is available at :
See also:
