Search

Text summarization has become an important research area, especially in the biomedical domain, where information overload is a major problem. In this paper, we propose a novel biomedical text summarization system that combines two popular data mining techniques: clustering and frequent itemset mining. Biomedical paper is expressed as a set of biomedical concepts using the UMLS metathesaurus. The K-means algorithm is used to cluster similar sentences. Then, the Apriori algorithm is applied to discover the frequent itemsets among the clustered sentences. Finally, the salient sentences from each cluster are selected to build the summary using the discovered frequent itemsets. For the evaluation step, we selected randomly 100 biomedical papers from the BioMed Central database full-text, and we evaluated the performances of our system by comparing the resulting summaries with the abstracts of these papers using the ROUGE metrics in term of recall, precision, and F-measure. We also compared the obtained summaries with those achieved by five well-known summarizers: TextRank, TextTeaser, SweSum, ItemSet Based Summarizer, Microsoft AutoSummarize, and two baselines: summarization using only the frequent itemsets mining (FRQ-CL), and summarization using only the clustering (CL-FRQ). The results demonstrate that this combination can successfully enhance the summarization performances, and the proposed system outperforms other tested summarizers.

In this paper, we have proposed a novel word embedding-based biomedical text summarizer. Biomedical words are represented by real dense vectors. Sentences are represented by summing-up the word vectors that contain. The PageRank algorithm is applied to rank sentences using the cosine similarity as a distance measure between sentences vectors. The top N highly ranked sentences are selected to build the summary. For the evaluation, we created a corpus of 200 biomedical papers downloaded from the Biomed Central full-text database. We used a pre-trained Word2vec model of word vectors generated from a combination of PubMed, PMC, and recent English Wikipedia dump texts. We compared our method with four other summarizers using: ROUGE-1, ROUGE-2, ROUGE-3, and ROUGE-SU4 metrics by evaluating the generated summaries with the abstracts of papers. Our summarizer achieved an improvement of 3.48%, 7.68%, 9.76%, and 3.47% respectively against the second-ranked summarizer.

Vertical partitioning is the process of subdividing the attributes of a relation into groups, creating fragments. It represents an effective way of improving performance in the database systems where a significant percentage of query processing time is spent on the full scans of tables. Most of proposed approaches for vertical partitioning in databases use a pairwise affinity to cluster the attributes of a given relation. The affinity measures the frequency of accessing simultaneously a pair of attributes. The attributes having high affinity are clustered together so as to create fragments containing a maximum of attributes with a strong connectivity. However, such fragments can directly and efficiently be achieved by the use of maximal frequent itemsets. This technique of knowledge engineering reflects better the closeness or affinity when more than two attributes are involved. The partitioning process can be done faster and more accurately with the help of such knowledge discovery technique of data mining. In this paper, an approach based on maximal frequent itemsets to vertical partitioning is proposed to efficiently search for an optimized solution by judiciously pruning the potential search space. Moreover, we propose an analytical cost model to evaluate the produced partitions. Experimental studies show that the cost of the partitioning process can be substantially reduced using only a limited set of potential fragments. They also demonstrate the effectiveness of our approach in partitioning small and large tables.

International audience; In the last decade, OnLine Analytical Processing (OLAP) has taken an increasingly important role as a research field. Solutions, techniques and tools have been provided for both databases and data warehouses to focus mainly on numerical data. however these solutions are not suitable for textual data. Therefore recently, there has been a huge need for new tools and approaches that treat and manipulate textual data and aggregate it as well. Textual aggregation techniques emerge as a key tool to perform textual data analysis in OLAP for decision support systems. This paper aims at providing a structured and comprehensive overview of the literature in the field of OLAP Textual Aggregation. We provide a new classification framework in which the existing textual aggregation approaches are grouped into two main classes, namely approaches based on cube structure and approaches based on text mining. We discuss and synthesize also the potential of textual similarity metrics, and we provide a recent classification of them.

In this paper, we combined sentence clustering and frequent itemsets mining to build a single biomedical text summarization method. Biomedical documents are represented as a sets of UMLS concepts. Very generic concepts are discarded. The vector space model is used to represent sentences. The K-means clustering algorithm is applied to cluster semantically similar sentences. The most frequent itemsets are extracted among the global cluster. The generated frequent itemsets are used to calculate the score of sentences. The top N highly scoring sentences are selected to represent the final summary. The method is evaluated against three summarizers: TextRank, SweSum and Itemset based summarizer on a 50 randomly selected biomedical papers from the BioMed Central database full text. The evaluation process consists of comparing the generated summaries with the abstracts of these papers using the ROUGE toolkit. Our method achieved good results, it ranked first in ROUGE-1 and ROUGE-2 measures with an improvement of \(\sim \)3% than the Itemset based summarizer and it ranked second in ROUGE-SU4 measure with a diminution of \(\sim \)1% always against the Itemset based summarizer.

We present in this paper a system for textual aggregation from scientific documents in the online analytical processing (OLAP) context. The system extracts keywords automatically from a set of documents according to the lists compiled in the Microsoft Academia Search web site. It gives the user the possibility to choose their methods of aggregation among the implemented ones. That is TOP-Keywords, TOPIC, TUBE, TAG, BienCube and GOTA. The performance of the chosen methods, in terms of recall, precision, F-measure and runtime, is investigated with two real corpora ITINNOVATION and OHSUMED with 600 and 13,000 scientific articles respectively, other corpora can be integrated to the system by users.

Vertical partitioning is a technique used to reduce disk access, when executing a given set of queries, by minimizing the access to irrelevant instance variables. In this paper we use the FP-Max data mining algorithm, for extracting frequent item set attributes. The frequently accessed instance variables are, then, grouped as vertical class fragments. We study the application of this approach to sets of queries on large databases and data warehouses. We used two benchmarks with various minimum support levels and we compare our results with the results of the approach using the Apriori data mining technique. The partitioning solution obtained produces an improvement of 14% for large data bases and 19% for data warehouse compared to a solution without partitioning.