We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.AI

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Computation and Language

Title: A Data-Centric Framework for Composable NLP Workflows

Abstract: Empirical natural language processing (NLP) systems in application domains (e.g., healthcare, finance, education) involve interoperation among multiple components, ranging from data ingestion, human annotation, to text retrieval, analysis, generation, and visualization. We establish a unified open-source framework to support fast development of such sophisticated NLP workflows in a composable manner. The framework introduces a uniform data representation to encode heterogeneous results by a wide range of NLP tasks. It offers a large repository of processors for NLP tasks, visualization, and annotation, which can be easily assembled with full interoperability under the unified representation. The highly extensible framework allows plugging in custom processors from external off-the-shelf NLP and deep learning libraries. The whole framework is delivered through two modularized yet integratable open-source projects, namely Forte (for workflow infrastructure and NLP function processors) and Stave (for user interaction, visualization, and annotation).
Comments: 8 pages, 4 figures, EMNLP 2020
Subjects: Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
Cite as: arXiv:2103.01834 [cs.CL]
  (or arXiv:2103.01834v4 [cs.CL] for this version)

Submission history

From: Zhengzhong Liu [view email]
[v1] Tue, 2 Mar 2021 16:19:44 GMT (8749kb,D)
[v2] Wed, 3 Mar 2021 02:57:35 GMT (8749kb,D)
[v3] Tue, 31 Aug 2021 19:00:40 GMT (8780kb,D)
[v4] Thu, 2 Sep 2021 03:10:43 GMT (8780kb,D)

Link back to: arXiv, form interface, contact.