Improving Text Summarization using Ensembled Approach based on Fuzzy with LSTM
- PDF / 2,341,342 Bytes
- 12 Pages / 595.276 x 790.866 pts Page_size
- 16 Downloads / 203 Views
RESEARCH ARTICLE-COMPUTER ENGINEERING AND COMPUTER SCIENCE
Improving Text Summarization using Ensembled Approach based on Fuzzy with LSTM Minakshi Tomer1,2 · Manoj Kumar3 Received: 24 October 2019 / Accepted: 23 July 2020 © King Fahd University of Petroleum & Minerals 2020
Abstract Abstractive text summarization using attentional recurrent neural network (sequence-to-sequence) models have proven to be very effective. In this paper, a novel hybrid approach is presented for generating abstractive text summaries by combining fuzzy logic rules (which selects extractive sentences) with bidirectional long short-term memory (Bi-LSTM) which further produces abstractive summary. Bi-LSTM uses attention mechanism and Adam optimizer for updating network weights. The proposed approach utilizes fuzzy measures and inference to extract textual information from the document to find the most relevant sentences. These relevant sentences are given as input to Bi-LSTM to produce an abstractive summary of the significant sentences. The proposed FLSTM model is evaluated using ROUGE toolkit. The experiment is performed on standard datasets (i.e., DUC and CNN/daily mail). Another salient feature of this work is merging of DUC 2003–2004, DUC 2006–2007 datasets to generate a larger dataset to achieve better results. The FLSTM model is compared with other state-of-the-art models, and the empirical results suggested that the proposed FLSTM model outperforms all other models. Keywords Automatic text summarization · Fuzzy rules · Feature extraction · Bi-LSTM
1 Introduction Text summarization (TS) is a procedure that produces the condensed text from the source document. The purpose of text summarization is to find out the major concept and producing a text of shorter length compared to source document. One of the most challenging problems in the area of NLP is to locate the significant parts present in input document [1]. Text summarization obtains a text of shorter length with all pertinent information present from an abundance of text sources. As the amount of online data has grown tremendously with superabundant information from different sources, it makes very time consuming and difficult for users to obtain relevant information. Due to this if anyone searches for some particular information, a lot of data is retrieved which is impossible for a person to read thoroughly. This brings an exponential
B
Minakshi Tomer [email protected]
1
USICT, Guru Gobind Singh Indraprastha University, New Delhi, India
2
Maharaja Surajmal Institute of Technology, New Delhi, India
3
Ambedkar Institute of Advanced Communication Technologies and Research, New Delhi, India
growth in the area of generating automatic text summary (ATS). ATS can be majorly grouped into two different categories: extractive and abstractive on the basis of output [2]. Extractive text summarization approach produces summaries by extracting words, set of words, phrases, set of phrases or sentences from the source document. It consists of three main steps: intermediate rep
Data Loading...
