Political Opinions in Literature: Identifying Themes in International Compositions

Political Opinions in Literature: Identifying Themes in International Compositions

1

Team POLITIC

Team Research Proposal

Team POLITIC

Political Opinions in Literature: Identifying Themes in International Compositions

Robert Cai, Matthew Carr, Adam Elrafei, Alexander Goniprow,

Adrian Hamins-Puertolas, Manpreet Khural, Andrew Li, Alexandra Winter,

Soumya Yanamandra, Dan Yang, and Kay Zhang

University of Maryland Gemstone Program

Mentor: Dr. Peter Mallios

Librarian: Timothy Hackman

and

The Maryland Institute for Technology in the Humanities

We pledge on our honor that we have not given or received any unauthorized assistance on this assignment.

Introduction

The United States was involved in numerous international conflicts throughout the 20th century. A prevalent theory suggests deeper public understanding of foreign cultures might have allowed the United States to avoid several of these conflicts, including the Iran Hostage Crisis and the Vietnam War (Li). Since the United States is a democracy, citizen perception of foreign countries has a direct relationship with foreign policies enacted. A thorough understanding of how the American public gathers its perceptions of foreign cultures is crucial to fullycomprehend American foreign policy andinternational relations. Foreign literature is one important medium that exposes the United States to the political and cultural ideologies of other countries (Griswold 1077). The American public reads novels by foreign authors togainanintimate perspective offoreign societies—views unavailable through domestic media. Readers can also connect to other cultures because novels create emotional ties by appealing to universal human themes (Aubry 27). At the same time, international and domestic political concerns guide the United States’ public interest in foreign literature. For instance, it is not a coincidence that the peaceful writings of Gandhi became important in the United States during the Civil Rights Movement(Mallios 10-19).
However, different foreignauthorsoften provide opposing viewpoints of their societies. The most popular works form a selective base of foreign literature that potentially accommodateselites’ self-serving political biases. Using experimental methods, Gilens asserts that the United States’ ignorance and misinformation “leads many [citizens] to hold political views different from those they would hold otherwise” (379). Therefore, understanding public intent and attitude requires knowing why certain novels and authors seem representative of a cultural canon. To become a better-informed political citizen of the United States, one must think critically about the uses of foreign literature.
Our study will investigate how publicly available United States media receivedforeign novels and authors and how these portrayals work toward social and political ends of government support and criticism (Mallios 10-19). Specifically, we will conduct a low-constraint case study of Russian literature to address the following question: Did the reception of Russian novels and authors in the United States and United States foreign policy toward Russia reflect each other from 1900-1923? We hypothesize that the reception of Russian literature in the United States significantly correlates with United States policies toward Russia, due to inherent ties between literary evaluation and political understanding. Scholars, politicians, and other government officials will likely take interest in our study.
We will use the portrayals of selected Russian novels and authors in nationally available print media to define the reception of Russian literature in the United States during this time period. We recognize scholars could investigate how alternative forms of media, such as pictures or political cartoons, influence public understanding. However, we chose print media because it is the easiest to quantitatively analyze. We will define United States foreign policy toward Russia through quantifiable measures such as foreign aid, military investment, and trade deals from 1900-1923. This will take the form of overarching topics that describe the types of policies enacted, such as interventionism and humanitarianism. Our analysis will include keyword searches relative to both literary reception and foreign policy. We will track how these themes have evolved over time using techniques of topic modeling.[1]
Our study does not seek to determine a relationship between political climates and messages found in novels, opinions held by authors, or motivations behind translators. Instead, we will determine the extent to which there is a relationship between media reception of Russian literaturein the United States and the political climate. Our research is distinguished from previous studies in two ways: it analyzes reception in United States media and not the intent of authors or translators, and we will accomplish our analysis through quantitative, not just qualitative, methods.
Throughout the rest of our proposal, we will summarize our literature review, outline our methodology, explain the limitations of our research, list confounding variables, and conclude with descriptions of our anticipated results, our budget, our timeline, and the statistical tools we will use throughout the project.

Literature Review

Introduction of Russian Literature in the Western World

Eugene-Melchoir de Vogue's Le Roman Russe (The Russian Novel) in 1886 represented the increasing interest in Russian literature in Western Europe and America. Many writers, including Isabel Hapgood and Constance Garnett, published English translations of Russian novels, short stories, and poems to critical acclaim in subsequent decades (Moser 431). In other words, the late nineteenth and early twentieth centuries marked the availability of Russian literature to US public and intellectuals.

Many studieshave sought to understand literary themes found in major Russian works. For example, Emerson analyzes Leo Tolstoy’s views on war through a close reading of his many texts (1855). However, only a few studies address Russian literary reception in the United States during the early twentieth century. One of these rarities is Goldfarb’s account of how a prominent literary critic, William Dean Howells, supported Tolstoy’s works in the United States during the twentieth century (318). However, this study is limited in that it only contemplates Russian literary reception through Howells’ and his critics’ views. We intend to expand on such studies by using comprehensive statistical tools to analyze a wider base of reception material.

Canon Formation and Politics

Political motivations shape a nation’s literary canon, which in turn projects that nation’s identity. The idea of a national literature emerged in the late eighteenth century as a way of proving cultural independence on an international level (Corse, Nationalism and Literature 7-14). Original research studies suggest canonical or high-culture literature does not reveal how citizens perceive themselves, but rather how elites want to envision their nation (ibid 74). These previous studies turn to college syllabi and literary prizes to define the most frequently appearing works as canonical or high-culture (Brown, 1; Corse, Nations and Novels 1279-82). Unlike bestsellers or popular culture novels, canonical texts differ greatly between countries, as they are symbolic in value and not simply “economic commodities.” Theories of canon formation state novels have to experience a conjunction of large sales and certain types of recognition to reach canonical status (Ohmann 206). This recognition refers to the critical reception of works found in publications that “carried special weight in forming cultural judgments,” such as the New York Times Book Review and the New Republic (204).However, scholars have never specified the ways in which elites have translated cross-cultural differences into literature.

Topic Modeling

Researchers use topic modeling to analyze large corpora of data. Topic modeling affirms “documents are mixtures of topics, where a topic is a probability distribution over words” (Steyvers 2). Furthermore, Latent Dirichlet Allocation (LDA), a more specific type of topic modeling, asserts each document from a larger corpus consists of a plurality of topics (Chaneyand Blei 2). In past studies, researchers have used topic modeling in general and LDA specifically to analyze large corpora of data. For example, a 100-topic LDA model generated word probabilities under each topic for all articles in the journal Science between 1880 and 2002 (ibid 4).

More complex versions of topic modeling, however,can gather more information from our Russian author database. For example, Topics over Time (TOT) models can account for the chronology of documents in a corpus (ibid 9). Since our documents are dynamic in that they change over time,LDA would confound the topics’ changes and lose any perceivable patterns. Xuerui Wang and Andrew McCallum explain the topic analysis of US Presidential State-of-the-Union addresses, where LDA “confounds Mexican-American War (1846-1848) with some aspects of World War I (1914-1918)” since it is “unaware of the 70-year separation between the two events” (1). Modeling topics over time serves to address this issue.

In Wang and McCallum’s study, they incorporated timestamps to help track “changes in the occurrence of the topics themselves” as a function of time (2). They tested their model on three data sets: “more than two centuries of U.S. Presidential State-of-the-Union addresses,” “17-year history of the NIPS [Neural Information Processing Systems] conference,” and “nine months of email archive” (ibid). The results of their study show the TOT model is able to predict the timestamps of documents and generates topics that are “more distinct from each other than LDA topics” (ibid 5). In our research, we will also use a TOT model on the databases we anticipate constructing to account for time.

Furthermore, modified versions of LDA can relate metadata to topics. Metadata is information about the documents we collect such as “author, title, geographic location, [and] links” (Blei 10). Therefore, we can also correlate influences such as the gender and ethnicityof the authors of the reception material to word probabilities found in topics in our corpus.

Sentiment Analysis

Sentiment analysis is also useful for sorting through large corpora of data. While topic modeling focuses on the subject of the data in question, sentiment analysis focuses on the opinion expressed about the subject matter of the data (Lee and Pang 1). Multiple methods can determine the sentiment of a piece of data. Lee and Pang compared three different algorithms used for sentiment analysis: the Naive Bayes, maximum entropy classification, and support vector machines (ibid 3). The Naive Bayes algorithm is a simplistic algorithm. It may not hold to high accuracy rates with complicated sets of data, but it “tends to perform surprisingly well” and is even the ideal algorithm for use with “problem classes with highly dependent features” (ibid). Maximum entropy classification and support vector machines are both much more sophisticated methods. Maximum entropy classification algorithms “make no assumptions about the relationships between features”, which will make it better than Naive Bayes with data that has little or no dependence on similar features (ibid 4). Support vector machines differ from both of the previous methods in that they do not focus on probability, which brings them much closer to traditional methods used for normal topic modeling adapted to work with sentiment analysis (ibid 4).

For our project, sentiment analysis methods will allow us to quickly categorize articles by gauging how American periodicals perceive and discuss Russian authors and novels during the time period of interest. In addition, incorporating a sentiment categorization into our database will allow future researchers to quickly add to and examine our data.

Foreign Policy Analysis

Political scientists have devised several models and theories to explain how foreign policy develops (Boyer 185). One such theory is the rational actor model, which states stimuli and immediate responses lead to the creation of foreign policy (Boyer 189). However, the political aspect of our study does not seek to determine how political leaders create foreign policy, but rather attempts to measure and quantify it. Many previous studies have determined United States foreign policy towards various nations by analyzing its components. For example, Rick Travis analyzes foreign policy towards Africa by focusing on foreign aid to the continent (798). Haslam focuses on direct foreign investment and the corresponding treaties to determine United States foreign policy toward other nations (1182). For our study, we will gather data on “exports, imports, investments, arms sales, and categories of foreign aid (bilateral, aggregate, and per capita)” between the United States and the Russian Empire (and later the Soviet Union) to define United States foreign policy (Watson 253).

Methodology

Our first tasks were to determine a time range and country to investigate, as outlined in the literature review. We selected an upper time bound of 1923, since all preceding publicationsare in the public domain and we can publicly release all collected data. We chose 1900 as our lower time bound to guaranteea significant number of periodicals will be available.[2] Time allowing, we may be able to expand the time period of interest, guaranteeing more articles for analysis.Wedecided to investigate Russian literature for several reasons. First,Russia was a focal point of the United States during the twentieth century. World War I, the Bolshevik Revolution, and the threat of communism led to increased public and governmental interest in Russia during our selected time period. Second, onlya relatively small number of significant Russian authors had works available in English at the time. A narrow range of Russian literary figures suggestsAmericanperiodicals interested in examining Russian literature had to invoke certain Russian literary figures and works frequently, leading to larger sample sizes for the selected authors. Subsequently,we will be able to construct a more exhaustive corpus[3] of Russian literature than of the more readily available literature from other countries, such as Britain or France.
To decide which literary figures to study, wecompiled a list Russian literary figures whose works had English translations during our time period of interest. Using that list, we cataloged the number of search results found in the Readers’ Guide Retrospective[4] for each literary figure of interest.[5]From this preliminary summary of the availability of periodicals in the United States specifically discussing Russian literary figures, we chose to investigate Dostoevsky and Tolstoy to maintain the feasibility of our study.We bringsome bias in our selection of literary figures, as we havechosen two of the most renowned Russian literary figures in the United States. Therefore, our data regarding the reception of selected Russian literary figures in the United States will not be representative of the entirety of Russian literary figures. We could add one or two minor Russian authors to our research to increase the external validity of our project if time permits.
We resolved to capture a large, representative sample of the body of articles that explicitly mention our selected Russian literary figures in periodicals popularin the United States between 1900 and 1923. We will construct a database containing these articlesusing the Readers’ Guide Retrospective index. The Retrospective’s emphasis on more popular periodicals fits well with our intent to gain an understanding of how the generalAmerican public perceived significant Russian literary figures in the early twentieth century. We will use a subject search of selected literary authors to explore the Readers’ Guide Retrospective and find articles appropriate for the constructed database.

Scanning

Since most articles in the Readers’ Guide are not digitized, wehave to digitize the physical or microfilm versions of articles thatfall within search parameters. We are currently scanning articles by using publicly available resources at the University of Maryland McKeldin Library. Therefore, our initial database construction will contain only articles available within the University of Maryland archive system. Should time permit, it may be feasible to explore other academic archives for articles from the Readers’ Guide Retrospective.

We have standardized scanning techniques to reduce preventable variations in image quality and size.[6] Systematic errors, including the presence of dust particles, stains, and other debris on the scanning glass, also contribute to poor image quality and complicate analysis of the database. We willtherefore wipe down the scanning glass with glass cleaner solution and a microfiber cloth before and after each scan to reduce this source of error.

Preservation of the scanned material is essential to data accuracy and reliability. During microfilm scanning, an auto-adjust function adjusts the brightness and scanned size of each page to produce an optimally clear image. Furthermore, we must adjust the resolution of the scanner up from the default 300 dots per inch (DPI) to the maximum setting of 600 DPI. Similar settings are also present on the non-print source scanners. Once saved, the file is left unmodified with the exception of cropping. We will not manipulate images after scanning to retain the original image data, quality, and integrity.