In the biomedical domain, figures in the scientific articles attribute significantly in understanding the core concepts. But, these figures are always difficult to interpret by the humans as well as machines and thus, associated texts in the article are required to summarize figures.

The current paper proposes an unsupervised automatic summarization system for individual figures present in a scientific biomedical article where different quality measures capturing relevance of the sentences to the figure are simultaneously optimized using the search capability of a multi-objective optimization technique to obtain a good set of sentences in the summary.

Figure-Summarization: A Multiobjective optimization based approach N Saini, S Saha, V Potnuru, R Grover, P Bhattacharyya - IEEE Intelligent Systems, 2019

Our proposed system, namely MOOFigSum, obtains 5% and 11% improvements in terms of F1-measure metric over the unsupervised technique for both datasets, respectively, while in comparison to supervised techniques, MOOFigSum obtains 9% and 2% improvements over these datasets, respectively.

Scientific document clustering (SDC) is the process of partitioning a given set of scientific articles/documents into K-groups based on some similarity/dissimilarity criterion. This task has applications in many areas like text summarization, classification of documents, scope detection of conference/journal papers, etc.

In this paper, we have developed a granular self-organizing map (GSOM) based approach for SDC. GSOM was designed to overcome the limitations of self-organizing map (SOM), ie, to avoid the problem of overlapping patterns present in different clusters. Both GSOM and SOM are the types of artificial feed-forward neural networks.

Scientific Document Clustering using Granular Self-organizing Map N Saini, R Grover, S Saha, P Bhattacharyya

To represent the documents in the form of numeric vectors, various representation schemes, capable of capturing semantics of the documents like word2vec etc., are utilized in our approach. To show the potentiality of GSOM, various comparative approaches like K-means, hierarchical clustering algorithms and SOM are considered. The results are reported using an internal cluster validity index namely, Dunn index. It evaluates the performance using compactness within clusters and separation among clusters. Results show that GSOM is superior over other existing algorithms. Finally, results are validated using statistical significant t-test.