Inaugural Speech Classification with Named Entities and Key Phrases

• Main Authors: Han, Hyoil, Lim, SeungJin
• Format: Conference Proceeding
• Language: English
• Subjects: Big Data; Conferences; Deep learning; key phrase; Manuals; named entity; political narratives; Supervised learning; supervised machine learning; Support vector machines; text mining; US presidential inaugural speeches
• ISSN: 2375-9356
• DOI: 10.1109/BigComp51126.2021.00029

The amount of online textual data is massive and is beyond homo sapiens' manual understanding. The techniques to analyze the massive textual data have been continuously developed. Machine learning techniques are providing excellent tools to help our understanding the huge amount of textual data by providing a succinct summary of documents, recommendation of articles, etc. Using named entities and key phrases can help better understanding the textual data and provides a smaller feature set than using unigram features. Our work shows that terms of named entities and key phrases actually provide truly meaningful results in classifying documents into target classes and hence provides the positive direction toward massive data classifications by using a smaller feature set. In this work, we classified inaugural speeches of the US presents by using named entities and key phrases into two parties: Republican and Democratic parties. We utilized a supervised learning approach (Support Vector Machines) and achieved 100% accuracy of performance in classifying inaugural speeches by using named entities and key phrases related to patriotic terms.