The proliferation of social media brings new opportunities to discover the ways in which we receive, process, and disseminate information — even information that seems confined to our imaginations. Mental imagery — those images we create in our imaginations as we read a text or watch a film — is not well understood. Netlytic, a Web–based system for automated text analysis, permitted the capture and analysis of online discussions relating to mental images of J.R.R. Tolkien’s and Peter Jackson’s The Lord of the Rings as text and as film adaptation, giving insight to our understanding of mental imagery as a form of human cognition and information processing. Furthermore, this study serves as a starting point for further development of academic research using Web–based text analysis systems and online communities.
Online data collection
Discussion and conclusions
Public reception of information has always been a tricky aspect of research. Interviews, focus groups, and surveys are complicated by the researcher’s presence and bias, while observation alone rarely affords the intimacy necessary to gauge opinion. Online public fora — Web sites, comments, blogs, message (or discussion) boards, and other related public online postings — have the benefit of being both public and personal as well as commonplace; people are increasingly comfortable interacting within this format (Gunter, 2009). One way of harnessing the ever–burgeoning flow of online public reception is by uploading public posts, blogs, feeds, tweets, etc., into an online text analysis tool such as Netlytic (www.netlytic.org/). This paper outlines the process of using Netlytic in an academic research setting to analyze discussions from a popular Tolkien fan Web site, TheOneRing.net (TORn, http://www.theonering.net/) illustrating how one can open a window onto more complex, nuanced topics contained within online discussions.
The inspiration for this research (Grek Martin, 2011) was a love for both J.R.R. Tolkien’s The Lord of the Rings and Peter Jackson’s cinematic adaptation of the story. Fascination with the retelling of a beloved story through filmic medium combined with Martin Barker’s (2006) study — “Envisaging ‘visualisation’: Some challenges from the international Lord of the Rings audience project” — fueled a desire to understand the interplay between the images we mentally conjure while reading a story, the images from the film adaptation, and the subsequent mental images inspired from seeing the film. Barker’s study draws attention to unsolicited audience interviewee connections between visualizations from Tolkien’s book and Jackson’s films; the research outlined below builds on that premise through analysis of online discussions via TheOneRing.net (TORn; 1999–2009a, 1999–2009b), a fan community Web site.
The interaction between text–based (ekphrastic) and film–based mental imagery in the minds of the reader/spectator is an interesting phenomenon that is not well understood. Mental imagery, or visualization, is a form of information processing that occurs when people read a text and imagine characters, places, or objects from the description. Film inspires visualization as well, using off–screen voices, framed shots, and movement to foment mental images of other places, objects, and characters not shown (Allen, 1993). But how do mental images and visual images, in this case from film, interact within the minds of the readers/spectators? Do filmic images simply take over? Do devoted fans of the original text superimpose their own ekphrastic imaginings? Can images from the two media work together? These questions were at the forefront of this research, but the real question remained — how could these questions be answered?
Online data collection
Online fora created by and for fans of a given work are inherently public on two levels: the sites are created by members of the general public who have a desire to share their views, and the majority of the sites are public (i.e., not password protected) and are therefore available to anyone with an Internet connection. Fans of The Lord of the Rings are no exception to the online forum/message board phenomenon and a multitude of Web sites abound for fans of J.R.R. Tolkien, Peter Jackson, and the world(s) they have created. One site stands out for being both all-encompassing and popular — TheOneRing.net. Founded in 1999, the site has attracted thousands of viewers and claims “to be the most comprehensive Tolkien fan site on the Web” (TheOneRing.net, 1999–2011a). As seen in Figure 1, the site includes news and information about the Hobbit and The Lord of the Rings in both films and books, as well as pages devoted to Peter Jackson, upcoming events, galleries, and most importantly for the purpose of this study, message boards.
Figure 1: Homepage of TheOneRing.net (2012) (www.theonering.net).
Discussion threads from two TORn message boards — Movie Discussion — LOTR and Reading Room — were collected and analyzed for content related to mental imagery formation and spatial/geographic description. Geographic description was important not only to this researcher’s interest, but for the premise that larger, more complex images take longer to visualize and may be more susceptible to alteration in the mind’s eye (Pylyshyn, 2007, 2003; Kosslyn, 1995; Kosslyn and Jolicoeur, 1980; Kosslyn, et al., 2006). As culture and geography are intertwined in Tolkien’s epic, descriptors of culture (race) and both general and specific locations were included in the selection process.
Selecting appropriate threads for inclusion in the study occurred over a period of four days for each message board. Selection was based on the following criteria: topic relevance (whether a thread in the Movie Discussion — LOTR board mentioned the book, or a thread in the Reading Room board mentioned the movie,) geography (if any location in the movie or book was named,) and number of posts (threads with fewer than 20 posts were not included unless the topic was particularly relevant, e.g., ROTK Geography? Where were the Rammas Echor?). The subject lines for each thread were scanned for topic fidelity and appropriate threads were recorded in a spreadsheet and added to the Netlytic dataset corresponding to the message board of origin. At the time of data collection, posts on the Reading Room discussion board numbered over 29,000, and the Movie Discussion — LOTR board, over 24,000 (TheOneRing.net, 1999–2009b). New messages are posted frequently on TheOneRing.net, therefore, thread selection for each message board took place over a relatively short time to minimize the impact of additions, beginning with the latest threads through to January 2007. Threads earlier than January 2007 are archived on TheOneRing.net, but in a format incompatible with Netlytic and were therefore not part of the study. For Movie Discussion — LOTR, 112 threads (3,098 individual posts) were imported into Netlytic on the dates 16, 20, 21, and 22 January 2011. For Reading Room, 82 threads (2,685 individual posts) were imported on 26 and 27 February 2011 and 1 and 2 March 2011.
Netlytic (formerly known as Internet Community Text Analyzer) is a Web–based system for automated text analysis and discovery of social networks from online communication (http://netlytic.org), giving the user the ability to examine public online posts, blogs, tweets, feeds, or other online textual media associated with any given topic. Automated text analysis techniques are used in conjunction with visualization techniques to highlight common themes, understand specific responses, or otherwise analyze the textual information inherent in the datasets (Gruzd, 2011). For the purposes of this study, an individual post within a thread constitutes a discrete unit of analysis of the dataset, even though the posts were uploaded to Netlytic in the context of a complete thread. A compelling function of Netlytic is that while it treats each post as a discrete element, it maintains all the connections that post possesses, including relationships with other posts and the poster’s relationships to other posters. This function allows not only a more traditional approach to text analysis as the researcher analyzes particular words or phrases, or adds new categories to the search, but relationships between posters can be examined and visualized as well.
Figure 2: The dapper conversion tool.
The process of importing threads into Netlytic was complicated somewhat by the format of the TORn message boards. Netlytic requires data be rendered in certain formats, e.g., RSS feed, or XML; therefore each thread had to be converted before it was imported. The first step in the process is to format the message board threads on the TORn Web site. To format a thread, the functions “View Flat Mode” and “Print Thread” (see Figure 2) were chosen, which performed a dual function: “Flat Mode” allows all the thread content to be seen at once and “Print Thread” eliminates everything except the information from the main content of the message board including the top and bottom content (e.g., advertisements, banners, animations), and the side content (e.g., site navigation.) This formatting streamlined the content that eventually would be uploaded into Netlytic. Using the online tool “dapper” (http://open.dapper.net/), a conversion mechanism that transformed formatted threads into XML was set up. Once the XML file is created, dapper creates a new url, and confirms the conversion process by displaying the url and preview of the file (see Figure 2). Preview not only allows the user to verify that the process was successfully completed, but it displays sufficient content to determine if the correct information was converted. Importing the formatted thread into Netlytic involves copying the new url and pasting into the box provided by Netlytic (see Figure 3). The entire process, beginning with selecting a relevant thread to uploading the formatted content into Netlytic, was performed for all 194 message board threads added to the dataset.
Figure 3: Importing data into Netlytic (screenshots from 2012 and 2013).
To allow for the possibility of comparison between the two message boards, two datasets were created in Netlytic. Threads collected from the Reading Room were imported to “LOTR_Book” resulting in 2,685 data elements and threads collected from the Movie Discussion — LOTR board were imported to the “LOTR_Movie” dataset resulting in 3,098 data elements. Once a dataset is established in Netlytic, the researcher may edit, clean, visualize, or delete it. “Cleaning” the data allows the researcher to remove common words or phrases that have been imported, but are not actually part of the content of the post — examples are poster/user names and “signatures” (words or phrases added to the end of every post created by a particular author.) The next step is analysis, which begins by choosing “Text Analysis” and at the next screen “Analyze.” Tag clouds — “lists” of frequently appearing words and phrases — are automatically generated using two separate processes: the Yahoo! Extractor and Netlytic’s own Local Extractor. The Yahoo! Extractor returns frequent words and phrases (Kraft, et al., 2005), while the Local Extractor returns common etymological roots — e.g., “hobbit*” for “hobbit,” “hobbits,” “Hobbiton.” Tagged words and phrases remain linked to the corresponding posts; therefore one can read the entire relevant post and the context in which the information was created. The user may also create categories, described below, into which Netlytic can separate the corresponding posts.
Finally, Netlytic is capable of tracking the posters themselves and how they relate to one another through their posts and responses through “Network Analysis.” Though not an integral part of this study, this feature could be extremely useful in identifying patterns of online interaction. Some posters might be confined to a single message board, or “follow” specific TORn members by regularly responding to their posts; some posters have more wide–ranging influence across several message boards. However, Network Analysis did prove useful during the course of the analysis in identifying common user names, which were later eliminated from the datasets using the cleaning process described above.
Once all relevant threads were identified, selected, reformatted, and imported into Netlytic, textual analysis using ideas and terms supported by literary, cinematic, and geographic research began. Beginning with words and phrases chosen to elicit descriptions of visualizations, book versus film imagery, or geographic and cultural locations, the vocabulary was augmented and refined according to the terminology used by TORn posters. Categories were created within Netlytic and used to discover patterns within the data corresponding to themes of visualization and location. Analysis of the patterns highlighted interesting results regarding creation of mental imagery, but also regarding the network and goals of the fan community.
While Netlytic as a tool for research is relatively new, the theoretical foundation for textual analysis is well established. Glaser and Strauss’ (1967) grounded theory is particularly suited for this study: it permits direct observation of the views of a globally diverse “microlevel” of society and it accommodates Netlytic’s functionality. The coding and constant comparison of codes required for a grounded theory analysis (Glaser and Holton, 2004) is easily attainable using Netlytic; the context of the data (i.e., the posters’ entire messages) is immediately accessible, as the terms are hyperlinked to their original messages. Trends and patterns are observable and can be used to refine searches and categorization. The combination of grounded theory with Netlytic is an extremely effective strategy for researching online communities, public reaction to news and events, and social networking at any scale.
The terms analyzed in this study came from the natural language used in the main body of the posts, although, depending on the research question, Netlytic can be configured to include terms from usernames, signatures, as well as other types of information such as avatars, linked images, and dates. Word tag clouds generated from the Yahoo! Extractor — no longer part of Netlytic functionality due to proprietary concerns — and the Local Extractor were added to the initial vocabulary list and formed the foundation of search terms, while a process of modifying and refining search terms and categories through Netlytic formed the basis for comparison and analysis. Terms used by both message board groups were noted, as well as terms favoured by one group or the other. Favoured terms were particularly helpful in providing insight into mental image formation from text (Reading Room board) and film (Movie Discussion — LOTR board.)
Figure 4: Term (concept) page revealed by clicking on a word in the tag cloud. The entire post can be accessed by clicking on the highlighted word for that post.
Netlytic generates statistics for each term (Figure 4) relating the number of members who used the term, the number of unique messages containing the terms, and the total number of instances the term is used through the dataset. In Figure 4 the term “elves” and the variant “elven” was used by 110 different posters in 462 separate messages, with a total of 727 instances in the dataset. In this way Netlytic was instrumental in the identification of common terms, as well as in the comparison and manipulation of terms to find trends and patterns. In addition to statistics, Netlytic provides the immediate context of the first usage of the term for each post, with the number of instances within that post and the full context available by clicking on the term (see Figure 4). Throughout the process of familiarization, refinement, and analysis the complete context of any given term was always immediately available. Returning to the above example, it would be a mistake to assume that the term “elves” was used in the same context in all 462 messages — in some instances, it is clearly a username (e.g., “Elven”). By examining the context, it is possible to refine the term search so that the context in which the term was used was similar in most or all of the cases. In the case of the variant “elven,” the number of times the term appeared as a username was insignificant compared to the frequency it was used to describe the culture/race of “Elves.” With the context of term usage always at hand, different combinations of words and phrases could be examined without danger of overstating the evidence of particular patterns or trends in usage.
Figure 5: The “Categories” page of Netlytic, showing the modified pre–existing category “agreement.” Other pre-existing categories are “certainty,” “evaluation,” “opinion,” “positive,” “reference,” “self,” “uncertainty,” and “us.”
The comparison of mental images generated from text and film is a complicated suggestion; discovering patterns that inform this suggestion required the use of Netlytic’s “Categories” function. With “Categories” one can group related terms and concepts, allowing for analysis within, and comparison between, groups of terms that have similar meaning or function. Netlytic has a number of pre–set categories that can be modified by the user, or the user can create new categories using any number of words or phrases (see Figure 5). For this study, several new categories were created, most notably “visualization” (see Figure 6), a category containing all words and phrases related to the creation of mental imagery. Additional categories reflected geographic and cultural aspects of The Lord of the Rings. To analyze and compare patterns of term use, categories can be switched on or off by clicking the red square “x” to the right of the term (see Figure 5). It should be noted that at no time is a term or a category permanently deleted in Netlytic; this feature may clutter the screen from time to time, but it is invaluable for keeping a record of terms and categories that worked or did not work. A full description of term refinement and category creation using Netlytic can be found in Appendix C of Grek Martin (2011).
Figure 6: Visualization: a user–created and user–defined category. This image shows all the terms in the first composition of the “Visualization” category and, as well, a “deleted” term.
Quite possibly one of the most interesting and exciting aspects of Netlytic, especially given the visual nature of this study, is the very literal ability to visualize data (see Figures 7 and 8). Comparing Figures 7 and 8 with Table 1 below, it is easy to identify terms used more frequently by the Reading Room members compared to the members posting on the Movie Discussion board. For example, the term “think” was used far more often by the Reading Room members than it was used by the Movie Discussion group, while “look” and “see” was used marginally more often by the Movie Discussion board. It is also clear from this example that the members of the Reading Room use more terms more frequently, while the members of the Movie Discussion board cleave to fewer terms overall. Moreover, Netlytic visualizations are interactive; clicking on a term will bring up the term window, complete with context (see Figure 9). This method of category creation, comparison and analysis through visualization of data, refinement of categories, and re–analysis led to the discovery of patterns of spatially focused mental imagery as discussed by the members of the TORn Web site. While the results of this study will be discussed in a forthcoming publication, it should be noted that a number of TORn members indicated an integration of both text–based and film–based mental imagery.
Table 1: A comparison of number of major term instances in the “visualization” category. Reading Room (2,685) Term Movie Discussion — LOTR (3,098) 1,023 50% “think” 1,116 36% 692 26% “see” 853 28% 521 19% “find” 360 12% 476 18% “make” 528 17% 443 16% “thought” 503 16% 420 16% “made” 492 16% 317 12% “look” 500 16% 167 6% “show” 231 7% 130 5% “image***” 185 6%
Figure 7: A visual breakdown of terms within the category “visualization” from the “LOTR_Book” dataset. There were 6,195 instances of terms used (out of 2,685 records).
Figure 8: A visual breakdown of terms within the category “visualization” from the “LOTR_Movie” dataset. There were 6,391 instances of terms used (out of 3,098 records).
Figure 9: “Image***” term screen and selected message.
Discussion and conclusions
This study highlights patterns of mental image formation observable in the TORn online fan community. The process of creating categories, finding patterns, and analyzing and refining the results, was the main strategy of textual analysis for this study. Through this strategy, several key connections were made in the relationship between text–based and film–based mental imagery, and insights gained in the successful operations of a well–organized and thoughtful online fan community — both of which will be discussed more thoroughly in a future paper. Posters identified the ability to integrate text–based imagery and film–based imagery, as well as recognized “essential qualities” from the text visually represented in the films. While it was not a primary focus of this study, I stumbled onto reasons why fans create and join online communities and aspects of the functionality of a long–standing fan–based Web site.
Netlytic was instrumental for this study. The process of amassing data is quick, especially in comparison to traditional content analysis where interviews are transcribed and terms are manually coded. The level of content analysis is sophisticated enough to isolate confidently complex and nuanced material; using the tools available at the time of this study, we were able to effectively limit the number of similar but contextually errant instances of term usage. Visualization aids not only in understanding and analysis, but it is very useful if you want to display the results of a study, as many find numbers in tables either confusing or distracting. The tool itself is not difficult to use and instructional notes are available on the Web site to guide the user through the process.
For research purposes, however, Netlytic would benefit from a few adjustments. The visualization tool is based on exact match to the selected term, however, if one is visualizing a term with many variants, (e.g., hobbit***: hobbit, hobbits, Hobbiton) then the visualization may be inaccurate. At the time of this study, Boolean logic was not a part of the category creation process, but this feature would be immensely useful in refining the context of selected terms. Finally, having the ability to create sub–categories would allow a more flexible approach to analyzing more complex but far–reaching topics. For example, comparing the mental images of “good” places versus “evil” places could have been aided with the sub–categories of specific places (Shire, Rivendell, Mordor) each with its own list of terms (hobbit***, elves, Sauron.) However, as Netlytic is developed for researchers of various subjects, disciplines, and depths, these adjustments may be better understood as recommendations for expanding Netlytic’s capability.
By describing the process of this study and the use of Netlytic as the primary tool of analysis, a way to perform in–depth research on online content is available to scholars and students. Netlytic can be used for a variety of online formats (Facebook, RSS feeds, blogs, message boards, etc.) and the topic of research is virtually unlimited. In the past, researchers have had a difficult time gauging public opinion, but increasingly people resort to online interaction to voice perspectives, file reactions, and disseminate information. With online content analysis tools, researchers in many fields may begin to understand what online voices are saying. Indubitably, what they are saying goes way beyond “like” and “dislike”.
About the authors
Jennifer Grek Martin received her MLIS from Dalhousie University; she also holds an M.Sc. in geography from the University of Wisconsin — Madison. She currently teaches at Dalhousie University and Saint Mary’s University, both in Halifax, Nova Scotia, Canada.
Direct comments to: jgrekmartin [at] dal [dot] ca
Dr. Anatoliy Gruzd is an Associate Professor in the School of Information Management and Director of the Social Media Lab (http://SocialMediaLab.ca) at Dalhousie University, Canada. His research initiatives explore how social media and other Web 2.0 technologies are changing the ways in which people disseminate knowledge and information and how these changes are impacting social, economic and political norms and structures of our modern society. Dr. Gruzd is also actively developing and testing new Web tools and apps for discovering and visualizing information and online social networks. The broad aim of his various research initiatives is to provide decision makers with additional knowledge and insights into the behaviors and relationships of online network members, and to understand how these interpersonal connections influence our personal choices and actions.
E–mail: gruzd [at] dal [dot] ca
Vivian Howard is associate professor in the School of Information Management and Associate Dean Academic of the Faculty of Management at Dalhousie University. Her research interests include barriers and motivators for pleasure reading, particularly for young readers; social reading initiatives; and, Atlantic Canadian literature for children and teens. She is the editor of the YA Hotline newsletter and is the principal investigator of a research team developing the Sea Stacks Web site (http://www.seastacks.ca), a resource devoted to Atlantic Canadian books for youth.
E–mail: vivian [dot] howard [at] dal [dot] ca
To the members of TheOneRing.net for their brilliant insights and for proving that an online fan–based community can be a site of engaging intellectual discussion and debate.
Richard Allen, 1993. “Representation, illusion, and the cinema,” Cinema Journal, volume 32, number 2, pp. 21–48.http://dx.doi.org/10.2307/1225603
Martin Barker, 2006. “Envisaging ‘visualisation’: Some challenges from the international Lord of the Rings audience project,” Film–Philosophy, volume 10, number 3, pp. 1–25, and at http://www.film-philosophy.com/2006v10n3/barker.pdf, accessed 22 April 2013.
Barney Glaser with the assistance of Judith Holton, 2004. “Remodeling grounded theory,” Forum: Qualitative Social Research, volume 5, number 2, at http://www.qualitative-research.net/index.php/fqs/article/view/607/1315, accessed 22 April 2013.
Barney Glaser and Anselm Strauss, 1967. The discovery of grounded theory: Strategies for qualitative research. Chicago: Aldine.
Jennifer Grek Martin, 2011. “Two roads to Middle–earth converge: Observing text–based and film–based mental images from TheOneRing.net fan community,” Master’s thesis, Dalhousie University, at http://hdl.handle.net/10222/14242, accessed 22 April 2013.
Anatoliy Gruzd, 2011. “Exploring virtual communities with the Internet Community Text Analyzer (ICTA),” In: Ben Kei Daniel (editor). Handbook of research on methods and techniques for studying virtual communities: Paradigms and phenomena. Hershey, Pa.: Information Science Reference, pp. 205–233.
Barrie Gunter, 2009. “Blogging — Private becomes public and public becomes personalized,” Aslib Proceedings, volume 61, number 2, pp. 120–126.http://dx.doi.org/10.1108/00012530910946875
Stephen Kosslyn, 1995. “Mental imagery,” In: Stephen Kosslyn and Daniel N. Osherson (editors). An invitation to cognitive science. Volume 2: Visual cognition. Cambridge, Mass.: MIT Press, pp. 267–296.
Stephen Kosslyn and Pierre Jolicoeur, 1980. “A theory–based approach to the study of individual differences in mental imagery,” In: Richard E. Snow, Pat–Anthony Federico, William E. Montague (editors). Aptitude, learning, and instruction: Volume 2: Cognitive process analyses of learning and problem solving. Hillsdale, N.J.: Erlbaum, pp. 139–175.
Stephen Kosslyn, William L. Thompson, and Giorgio Ganis, 2006. The case for mental imagery. Oxford: Oxford University Press.
Reiner Kraft, Farzin Maghoul, and Chi Chao Chang, 2005. “Y!Q: Contextual search at the point of inspiration,” CIKM ’05: Proceedings of the 14th ACM International Conference on Information and Knowledge Management, pp. 816–823.
Zenon Pylyshyn, 2007. Things and places: How the mind connects with the world. Cambridge, Mass.: MIT Press.
Zenon Pylyshyn, 2003. Seeing and visualizing: It’s not what you think. Cambridge, Mass.: MIT Press.
TheOneRing.net, 1999–2009a. “About TheOneRing.net,” at http://www.theonering.net/torwp/about/, accessed 22 April 2013.
TheOneRing.net, 1999–2009b. “TheOne Ring Forums,” at http://newboards.theonering.net/forum/gforum/perl/gforum.cgi, accessed 22 April 2013.
Received 1 March 2013; accepted 22 April 2013.
Copyright © 2013, First Monday.
Copyright © 2013, Jennifer M. Grek Martin, Anatoliy Gruzd, and Vivian Howard.
Navigating an imagined Middle–earth: Finding and analyzing text–based and film–based mental images of Middle–earth through TheOneRing.net online fan community
by Jennifer M. Grek Martin, Anatoliy Gruzd, and Vivian Howard.
First Monday, Volume 18, Number 5 - 6 May 2013