Teanga: A Linked Data based platform for Natural Language Processing

Teanga: A Linked Data based platform for Natural Language Processing Housam Ziad, John P. McCrae, Paul Buitelaar Insight Centre for Data Analytics, National University of Ireland, Galway [firstname.lastname]@insight-centre.org Abstract In this paper, we describe Teanga, a linked data based platform for natural language processing (NLP). Teanga enables the use of many NLP services from a single interface, whether the need was to use a single service or multiple services in a pipeline. Teanga focuses on the problem of NLP services interoperability by using linked data to define the types of services input and output. Teanga’s strengths include being easy to install and run, easy to use, able to run multiple NLP tasks from one interface and helping users to build a pipeline of tasks through a graphical user interface. Keywords: Natural Language Processing, Linked Data, NLP Architecture, NLP Framework 1. Introduction Natural language processing (NLP) tasks typically consist of many individual components that need to be used to- gether in order to solve real-world problems. However, it is frequently the case that these components are developed independently and thus far integration of these services is far from trivial. The installation of these services can act as a significant barrier to entry for NLP developers and even once developed these pipelines can be opaque and brittle. These issues are of course endemic to software develop- ment and until recently could only be solved by integrating all components within a single development model, for ex- ample, such as integrating all NLP tools using the Python language as has been done by NLTK(Bird, 2006). An al- ternative model has arisen in the form of Web services that provide integration between multiple components through clear and defined protocols such as REST. However, Web services have generally not been adopted by researchers or industry, in part due to the fact that the remote nature of the computation can lead to issues with the availability of ser- vices (as external services are often down) and the speed of these services (as sending requests to servers creates signif- icant bottlenecks). In this paper, we propose a new platform called Teanga1, which aims to achieve the best of both worlds. We use Web services combined by means of novel linked data stan- dards, in particular, JSON-LD, to provide interoperabil- ity between services without the need to have particular programming or framework. We also use containerization technology, in particular, Docker, to ensure that these ser- vices and the pipelines that are generated by these tools are highly portable and can easily be used at scale. Further- more, Teanga has an easy to use UI that allows users to visualise their pipelines as well as the progress (or failure) of each service individually. 2. NLP Services NLP services cover a broad spectrum of tasks, including term extraction, taxonomy extraction, machine transla- tion, sentiment analysis, suggestion mining, automatic summarisation, entity recognition and text classification, 1‘teanga’ ["tjaNg@] means ‘language’ in Irish which, along with many others, are used widely in many application …

Content truncated to 500 words