Abstract
AbstractA successful keyphrase extractor extracts exclusive, relevant keyphrases that capture the summary about the topic of the data and help in a fast information process. There are various methods previously available for keyphrase extraction, especially from Twitter data. Despite their availability, performance enhancement of these methods is a challenging problem. Also extracting keyphrases from twitter can help in extracting an idea from tweets related to specific social issue, disaster, user reviews about a product which in turn help in various applications which use keyphrases in awareness making, law abiding and recommendation applications. Hence, focusing on enhancing the performance of the existing models and to extract relevant keyphrases, we propose a supervised keyphrase extraction model using Extractive BERT Summarizer (BERT SUM) with Bidirectional Long Short-Term Memory (Bi-LSTM) using Global vectors (GloVe) word embedding from the Twitter tweets. We implement the model on the real-time Twitter data of 6000 tweets. We use BERT SUM for summarizing the tweets and Bi-LSTM for classification and extraction of keywords. To evaluate the performance of the proposed architecture the performance metrics precision, recall and F1-score are used, which are accepted commonly by the previous keyphrase extraction works. The results from the experiment showed that the proposed architecture outperforms the already available state-of-the-art keyphrase extraction methods.KeywordsTwitterSupervised Keyphrase extractionTwitter analyticsInformation retrievalDeep learning
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have