Several web services for various natural language processing (NLP) tasks (‘‘NLP-as-a-service” or NLPaaS) have recently been made publicly available. However, despite their similar functionality these services often differ in the protocols they use, thus complicating the development of clients accessing them. A survey of currently available NLPaaS services suggests that it may be possible to identify a minimal application layer protocol that can be shared by NLPaaS services without sacrificing functionality or convenience, while at the same time simplifying the development of clients for these services. In this paper, we hope to raise awareness of the interoperability problems caused by the variety of existing web service protocols, and describe an effort to identify a set of best practices for NLPaaS protocol design. To that end, we survey and compare protocols used by NLPaaS services and suggest how these protocols may be further aligned to reduce variation.
[1]
Robert Leaman,et al.
PubTator central: automated concept annotation for biomedical full text articles
,
2019,
Nucleic Acids Res..
[2]
Mihai Surdeanu,et al.
The Stanford CoreNLP Natural Language Processing Toolkit
,
2014,
ACL.
[3]
Walter Goralski.
Chapter 26 – Hypertext Transfer Protocol
,
2017
.
[4]
Erhard W. Hinrichs,et al.
WebLicht: Web-Based LRT Services for German
,
2010,
ACL.
[5]
Roy Fielding,et al.
Architectural Styles and the Design of Network-based Software Architectures"; Doctoral dissertation
,
2000
.
[6]
Maciej Piasecki.
User-driven Language Technology Infrastructure - the Case of CLARIN-PL
,
2014
.
[7]
Yue Wang,et al.
Open Agile text mining for bioinformatics: the PubAnnotation ecosystem
,
2019,
Bioinform..
[8]
D Manyu,et al.
Hypertext transfer protocol
,
2009
.