Atif Khan

A framework for multi-document abstractive summarization based on semantic role labelling

We have proposed a framework for multi-document abstractive summarization based on semantic role labeling (SRL). To the best of our knowledge, SRL has not been employed for abstractive summarization.The integration of genetic algorithm with SRL based framework for abstractive summarization results gives improved summarization results.My study focus on two highlights and discussion is based on these two highlights. We propose a framework for abstractive summarization of multi-documents, which aims to select contents of summary not from the source document sentences but from the semantic representation of the source documents. In this framework, contents of the source documents are represented by predicate argument structures by employing semantic role labeling. Content selection for summary is made by ranking the predicate argument structures based on optimized features, and using language generation for generating sentences from predicate argument structures. Our proposed framework differs from other abstractive summarization approaches in a few aspects. First, it employs semantic role labeling for semantic representation of text. Secondly, it analyzes the source text semantically by utilizing semantic similarity measure in order to cluster semantically similar predicate argument structures across the text; and finally it ranks the predicate argument structures based on features weighted by genetic algorithm (GA). Experiment of this study is carried out using DUC-2002, a standard corpus for text summarization. Results indicate that the proposed approach performs better than other summarization systems.

The Power of Implicit Social Relation in Rating Prediction of Social Recommender Systems

The explosive growth of social networks in recent times has presented a powerful source of information to be utilized as an extra source for assisting in the social recommendation problems. The social recommendation methods that are based on probabilistic matrix factorization improved the recommendation accuracy and partly solved the cold-start and data sparsity problems. However, these methods only exploited the explicit social relations and almost completely ignored the implicit social relations. In this article, we firstly propose an algorithm to extract the implicit relation in the undirected graphs of social networks by exploiting the link prediction techniques. Furthermore, we propose a new probabilistic matrix factorization method to alleviate the data sparsity problem through incorporating explicit friendship and implicit friendship. We evaluate our proposed approach on two real datasets, Last.Fm and Douban. The experimental results show that our method performs much better than the state-of-the-art approaches, which indicates the importance of incorporating implicit social relations in the recommendation process to address the poor prediction accuracy.

Genetic semantic graph approach for multi-document abstractive summarization

The aim of automatic multi-document abstractive summarization is to create a compressed version of the source text and preserves the salient information. Existing graph based summarization methods treat sentence as bag of words, rely on content similarity measure and did not consider semantic relationships between sentences. These methods may fail in determining redundant sentences that are semantically equivalent. This paper introduces a genetic semantic graph based approach for multi-document abstractive summarization. Semantic graph from the document set is constructed in such a way that the graph nodes represent the predicate argument structures (PASs), extracted automatically by employing semantic role labeling (SRL); and the edges of graph correspond to semantic similarity weight determined from PAS-to-PAS semantic similarity, and PAS-to-document set relationship. The PAS-to-document set relationship is represented by different features, weighted and optimized by genetic algorithm. The salient graph nodes (PASs) are ranked based on modified graph based ranking algorithm. In order to reduce redundancy, we utilize maximal marginal relevance (MMR) to re-ranks the PASs and use language generation to generate summary sentences from the top ranked PASs. Experiment of this study is carried out using DUC-2002, a standard corpus for text summarization. Experimental results reveal that the proposed approach performs better than other summarization systems.

Clustered genetic semantic graph approach for multi-document abstractive summarization

Multi-document summarization aims to produce a compressed version of numerous online text documents and preserves the salient information. A particular challenge for multi-document summarization is that there is an inevitable overlap in the information stored in different documents. Thus, effective summarization methods that merge similar information across the documents are desirable. This paper introduces a clustered genetic semantic graph approach for multi-document abstractive summarization. The semantic graph from the document set is constructed in such a way that the graph vertices represent the predicate argument structures (PASs), extracted automatically by employing semantic role labeling (SRL); and the edges of graph correspond to semantic similarity weight determined from PAS-to-PAS semantic similarity, and PAS-to-document relationship. The PAS-to-document relationship is expressed by different features, weighted and optimized by genetic algorithm. The salient graph nodes (PASs) are ranked based on modified weighted graph based ranking algorithm. The clustering algorithm is performed to eliminate redundancy in such a way that representative PAS with the highest salience score from each cluster is chosen, and fed to language generation to generate summary sentences. Experiment of this study is performed using DUC-2002, a standard corpus for text summarization. Experimental results indicate that the proposed approach outperforms other summarization systems.

Enhanced lexicon based model for web forum answer detection

A Web forum is an online community that connects people with common interest together. Within the forum, members interact to share knowledge, expertise and resources. A major issue in detecting web forum answers is to establish a good relationship between the question and the candidate answer. This relationship is often established using lexical features. Web forum text, unlike news articles, is faced with noise challenges, and this hinders the performance of lexical features. In this paper, we investigate the effect of noise on most of the common lexical features used in mining web forum answers with a view of normalizing it to enhance the performance of the features. We propose 13 lexical features for exploration. These features belong to four different quality dimensions that can guarantee good answers. We empirically address the following questions in the paper. What category of noise is more rampant in web forum? What lexical mining features are more susceptible to noise? Will normalization of forum corpus enhance the performance of lexical features in detecting web forum answers? We used three publicly available datasets of varying technical degrees for the experiments. The experimental results revealed that proper normalization of web forum corpora can yield up to 9% increase in the performance of the lexical features.