Comparative Study of Annotation Tools and Techniques
CHAPTER ONE
Objective
- Study and analyse different existing annotation tools and
- Test the discovered tools and
- Compare the studied annotation tools and
- Make a comprehensive conclusion and suggestion on the best tools suitable for different context of use.
Chapter Two
Literature review
This chapter presents some basic terminologies on the subject matter with a review of works done related to the research.
What is annotation?
Annotation has been one of important topics under research for the past years and it has played a considerable role in data management.
Annotation is a way of interpreting a document or commenting on a document.it adds extra information to it and makes it easier to browse, search, retrieve, categorize and analyse (Okunoye et al., 2010). It can also be understood as a way of creating semantic labels within a document (Pernelle & Nathalie, 2017).
Annotation structures data and facilitates easy storage and access of data. It helps in information sharing and knowledge elicitation by reading one’s perception on a particular document of interest.
Types of annotations
The formats of the document to be annotated determines the type of annotation. The annotated document can be of different formats: text, image, videos.
Text annotation
Text annotation is an act of adding extra note to a text. This can be achieved by adding marks, putting footnotes, highlighting or underlining a text of interested. Annotating a text adds value to it, makes it more informative and allow user own to integrate his/her interpretation with the text (Gosal, 2015).
Image annotation
Image annotation is a process of adding descriptive captions or tags (location, time, etc.) or keywords to the image to make it more accessible and more descriptive so that it can easily be stored, searched, categorized, retrieved and recognized (Hanbury, 2008).
Video annotation
Video annotation is a process of adding captions or keywords that add extra information to the video in order to facilitate videos access and retrieval from a large video database (Dasiopoulou, Giannakidou, Litos, Malasioti, & Kompatsiaris, 2011). It also helps in storing, browsing, searching, categorization of videos.
Annotation techniques
Based on the type of data, time and annotation accuracy that one wants to achieve, one may choose among the three annotation techniques. Annotation can either be manual, semi-automatic or automatic.
Manual annotation
Manual annotation is a way of transforming the extant syntactic resources into interconnected structures buy putting extra information to a document or part of document which forms metadata (Pernelle & Nathalie, 2017). Manual annotation is more accurate than automatic annotation but is expensive and does not take into consideration multiple perspectives (Pernelle & Nathalie, 2017).
Chapter Three
Discussion of annotation Tools
Many annotation tools have been developed and they are available to satisfy user needs. Most of the annotation tools are accessible on cloud and have free trial versions for testing purpose, whereas others require buying license and custom installation or configuration on your local infrastructure.
In this chapter six annotation tools are studied and tested. This testing focuses on some parameters that can be used to evaluate the performance of an annotation tool.
Machine learning and annotation
Machine learning models are sometimes employed in automatic annotation by some annotation tools, and in return annotation tools are used in the training phase of some machine learning models, this is usually found in supervised learning, where the inputs(X) and output(y) need to be labelled.
Categories of annotation tools
Constrained on the target data types that are to be annotated by an annotation tool, we can categorize the annotation tools into three main categories:
- Text annotation tools
- Image annotation tools
- Video annotation tools
Chapter Four
Comparative analysis
Features used for comparison
The study of the above selected annotation tools led to a comparative analysis between the studied tools. Different features and properties of each of the studied tools have been compared in order to evaluate an annotation tool’s performance measure.
Chapter Five
Conclusion and Recommendations
Conclusion
In this research, a study of six text annotation tools has been done. The features, functionalities and properties of the tools have been discussed in details and tested.
After studying and critically testing the annotation tools, a comparative table that shows the comparison of the tools was produced. The comparison primarily focused on different parameters and properties inherent in the text annotation tools studied. The advantages of using each annotation tool were figured out.
All the studied tools have different features and properties; and they differ from one tool to another. Thus, there is no annotation tool that can be taken to be better than others. Depending on the project requirements and specifications, the user should always be able to find a tool that complies with his/her need.
The comparative table will guide users in making their choices of annotation Tools depending on what they are looking for and what is provided or presented by an annotation tool.
Recommendations
For future work, this research study can further be improved by studying as many annotation tools as possible, including image annotation tools and videos annotation tools focusing more on those that support Machine Learning.
References
- Abahai, A. (2015). Automated Document Annotation for Effective Data Sharing. 8491, 4–7.
- AI-enabled Text Annotation Tool | PDF, Markdown, CSV, html, tweets, & many more types of Documents. (n.d.). Retrieved May 19, 2019, from https://www.tagtog.net/
- Bipboy, M. (2017). Using Social Annotation Tool to Enhance Learning : A Literature Review. 1–17.
- Bodain, Y., & Robert, J. M. (2007). Developing a robust authoring annotation system for the Semantic Web. Proceedings – The 7th IEEE International Conference on Advanced Learning Technologies, ICALT 2007, (Icalt), 391–395. https://doi.org/10.1109/ICALT.2007.119
- Brusilovsky, P. (2005). Efficient techniques for adaptive hypermedia. (June), 12–30. https://doi.org/10.1007/bfb0023957
- Cejuela, J. M., McQuilton, P., Ponting, L., Marygold, S. J., Stefancsik, R., Millburn, G. H., & Rost, B. (2014). Tagtog: Interactive and text-mining-assisted annotation of gene mentions in PLOS full-text articles. Database, 2014, 1–8. https://doi.org/10.1093/database/bau033
- CEUR workshop proceedings. (n.d.). Retrieved from https://gate.ac.uk/sale/tao
