No open-access location is known for this record yet.
No open-access location is known for this record yet.
EXECUTIVE SUMMARY This health care reform debate regarding H.R. 3962 generated a considerable amount of text. This paper utilizes documents from debates in the U.S. House of Representatives to examine relevant trends and patterns. We propose text mining as a tool that automatically extracts knowledge from unstructured texts. The findings reveal characteristics beyond party affiliation that differentiate Democrats from Republicans. Keywords: Text mining, Health care, Knowledge discovery, Cluster analysis, H.R. 3962 INTRODUCTION From the start, the Obama administration positioned the health care reform issue as a priority. For at least a year, a great deal of debate took place all over the country but especially in Washington DC where Democrats and Republicans traded accusations. Even after the passage of the H.R. 3962 bill, the debate continues and will most likely continue during the next presidential election. These debates generate a considerable amount of text, and a study of this wealth of information should be interesting and beneficial. This type of study could provide each with a better understanding of the underlying message of the opponent in this debate. Text mining or knowledge discovery for text could provide the proper tools to analyze the data. Text analysis has been applied in a wide range of fields. For example, in the context of politics, Hart (2000) analyzed a collection of political documents that covered U.S. presidential campaigns from 1948 to 1996. His objective was to show how text analysis can be used to understand texts from different sources. His study uncovered patterns and trends within the texts of three groups (politicians, press, and public). In this paper, the authors will apply text mining to automatically extract knowledge from the debates' text. This is to uncover patterns in the text beyond party affiliation differences. These patterns are characteristics that differentiate Democrats from Republicans. This is possible as the debates' text is analyzed to extract relevant knowledge using tools such as text encoding and clustering. METHODOLOGY Data The current study uses a collection of statements by U.S. Representatives generated during the debate of H.R. 3962 in the U.S. House of Representatives, (Congressional Digest, 2009). The text corpus includes ten documents, one for each U.S. Representative. Table 1 shows a summary of the data. The original text in a PDF format was preprocessed to remove all errors and unwanted characters. Then each document was tagged according to its attributes such as author, party affiliation, and text size. The final corpus is a set of ten Word documents. A preliminary and manual scanning of the documents shows that Republicans use quantitative references more frequently than Democrats. As shown in Table 2 below, the average per document is seven for Democrats and twenty-five for Republicans. On the other hand, Democrats use more references to misinformation from the other side such as lies and distortion, false, and inaccurate. Analysis The first phase of the analysis starts with removing punctuation, filtering stop words, mapping words to their basic forms. Then, a set of words (bag-of-words) is extracted from the text. A term-by-document matrix (lexical table) is then constructed. The rows correspond to words or terms, and the columns correspond to documents. This approach is also the vector space model where each element of the vector represents a word. The element is set to one if the word exists in the document and to zero if the word is not. The text mining software tool used to automatically execute the steps just discussed is TermExtractor which is a free software package for terminology extraction. This software helps to extract and validate domain relevant terms by using a corpus of domain documents, parses the documents, and extracts a list of terms with high importance. …
Loading…
Imported from OpenAlex (CC0) — nothing here is generated.