@InCollection{Limbasiya2020, author = {Nivid Limbasiya and Prateek Agrawal}, booktitle = {Algorithms for Intelligent Systems}, publisher = {Springer Singapore}, title = {{Bidirectional Long Short-Term Memory-Based Spatio-Temporal in Community Question Answering}}, year = {2020}, month = jan, pages = {291--310}, abstract = {Community-based question answering (CQA) is an online-based crowdsourcing service that enables users to share and exchange information in the field of natural language processing. A major challenge of CQA service is to determine the high-quality answer with respect to the given question. The existing methods perform semantic matches between a single pair of a question and its relevant answer. In this paper, a Spatio-Temporal bidirectional Long Short-Term Memory (ST-BiLSTM) method is proposed to predict the semantic representation between the question–answer and answer–answer. ST-BiLSTM has two LSTM network instead of one LSTM network (i.e., forward and backward LSTM). The forward LSTM controls the spatial relationship and backward LSTM for examining the temporal interactions for accurate answer prediction. Hence, it captures both the past and future context by using two networks for accurate answer prediction based on the user query. Initially, preprocessing is carried out by name-entity recognition (NER), dependency parsing, tokenization, part of speech (POS) tagging, lemmatization, stemming, syntactic parsing, and stop word removal techniques to filter out the useless information. Then, a par2vec is applied to transform the distributed representation of question and answer into a fixed vector representation. Next, ST-BiLSTM cell learns the semantic relationship between question–answer and answer–answer to determine the relevant answer set for the given user question. The experiment performed on SemEval 2016 and Baidu Zhidao datasets shows that our proposed method outperforms than other state-of-the-art approaches.}, doi = {10.1007/978-981-15-1216-2_11}, keywords = {Answer quality prediction, BiLSTM, Community question answering, Deep learning, Par2vec, Spatio-Temporal}, url = {https://link.springer.com/chapter/10.1007/978-981-15-1216-2_11} }