Multi-document summarization (MDS) is a pre-programmed process to excerpt data from various documents regarding similar topics. We aim to employ three techniques for generating summaries from various document collections on the same topic. The first approach is to calculate the importance score for each sentence using features including TF-IDF matrix as well as semantic and syntax similarity. We build our algorithm to sort the sentences by importance and add it to the summary. In the second approach, we use the k-means clustering algorithm for generating the summary. The third approach makes use of the Page Ranking algorithm wherein edges of the graph are formed between sentences that are syntactically similar but are not semantically similar. All these techniques have been used to generate 100–200 word summaries for the DUC 2004 dataset. We use ROUGE scores to evaluate the system-generated summaries with respect to the manually generated summaries.

Benchmarking Semantic, Centroid, and Graph-Based Approaches for Multi-document Summarization

Implementación de un dispositivo de datos leíbles por máquina para el apoyo de la adquisición de lenguaje en la infancia

Availability of data is not the foremost concern today; it is the extraction of relevant information from that data which requires the aid of technology. It is to help millions of users arrive at the desired information as quickly and effortlessly as possible. Document summarization and opinion-based document classification can effectively resolve the well-known problem of information overload on the Web. Summarization is about finding the perfect subset of data which holds the information of the entire set. In this paper, first we studied and evaluated three methods of generating summaries of multiple documents, namely, K-means clustering, novel-graph formulation method, and the stack decoder algorithm. The performance analysis emphasized on time, redundancy and coverage of the main content, was conducted along with the comparison between respective ROUGE scores. Next, hybrid architecture was proposed using a Stack decoder algorithm for creating automated summaries for multiple documents of similar kind, which were used as the dataset for analysis by a recursive neural tensor network to mine opinions of all the documents. The cross-validation of the generated summaries was done by comparing the polarity of summaries with their corresponding input documents. Finally, the results of opinion mining of each summary were compared with its corresponding documents and were found to be similar with few variations.

Sujal

Papers

Multi-document Summarization and Opinion Mining Using Stack Decoder Method and Neural Networks