Program Overview: Diction 5.0 contains a series of built-in dictionaries that search text documents for 5 main semantic features (Activity, Optimism, Certainty, Realism and Commonality) and 35 sub-features (including tenacity, blame, ambivalence, motion, and communication). After the user’s text is analyzed, Diction compares the results for each of the 40 dictionary categories to a "normal range of scores" determined by running more than 20,000 texts through the program.  Users can compare their text to either a general normative profile of all 20,000-plus texts OR to any of 6 specific sub-categories of texts (business, daily life, entertainment, journalism, literature, politics, scholarship) that can be further divided into 36 distinct types (e.g., financial reports, computer chat lines, music lyrics, newspaper editorials, novels and short stories, political debates, social science scholarship).  Diction also outputs raw frequencies (in alphabetical order), percentages, and standardized scores; custom dictionaries can be created for additional analyses.

Aesthetic Description: Diction runs through Microsoft Windows and features a "classic" Windows look that is mouse ("point and click") driven and easy to use. 

Text Preparation Procedure: To analyze our two convention speeches, the only advance text preparation we needed to do was to to save each speech in text format (with a .txt extension).  Diction also recognizes alpha-numeric (e.g., #12345678) and descriptive identifiers (e.g., *Gore Convention Speech) placed at the top of texts, which can help users keep track of multiple texts ("cases") imported into a statistical analysis package such as SPSS.  We added alpha-numeric and descriptive identifiers to each of the sample texts.

Analysis Process: Instead of requiring users to combine all texts (cases) into a single file, Diction 5.0 works through a master "project" file into which individual texts are opened.  To analyze the convention speeches, we first started a new Diction project (named "speeches") and then opened each speech into the project.  Once in the project, the speeches could all be analyzed at once with a single mouse click.

Program Output Description: The Diction output screen is divided into three parts--the File View Window Pane, the Data View Window Pane, and the Numeric View Window Pane.  The File View Window Pane is a narrow window on the top right of the screen.  It contains the names of all texts run in the master project file.  The results for each text can be viewed one at a time by clicking the text name in the File View pane and looking at the Data View Window Pane, which is next to the File View pane and takes up most of the output screen.

The Data View Window Pane verbally reports the results of a given text analysis.  It includes the following information: (1) total words analyzed and average word size; (2) standard dictionary totals for the 35 sub-feature variables (i.e., each variable's frequency of occurrence, the percent of words analyzed that the variable represents, the normal range of scores for type of normative comparison chosen, and whether or not your text is out of range on the variable); (3) an alphabetical list of the most frequently occurring words; (4) the totals for the calculated variables (i.e., frequency, percent, normal range, and out of range; and (5) the totals for the master variables (i.e., frequency, normal range, and out of range).

Selected findings from the analyses conducted to evaluate Diction will be used to illustrate some of the above features.  Before looking at the results, the first thing we did was to change the normative score comparison from the default "all texts" to the more meaningful "political campaign speeches."  This allowed us to compare the Bush and Gore convention speeches to a sample of 2357 speeches delivered by Republican, Democratic, and third-party presidential candidates between 1948 and 1996.  The analyses showed that Bush was above the normal range of scores on satisfaction, inspiration, temporal terms, and denial, and below the normal range of scores on numerical terms, leveling terms, passivity, and familiarity.  Bush was also below the normal range on the calculated variable "complexity."  Gore, on the other hand, delivered a pretty "average" speech--he was only out of (above) the normal range of scores on one variable, self reference.  For more results, see the sample output section below.  

Finally, the Numeric View Window Pane, a narrow pane beneath the other two, contains a string of numbers for each text that is analyzed.  The values in each string correspond to the numeric results in each data view pane.  These number strings are set up for easy import into SPSS for further statistical analyses.  

Sample Output: Data View Window Pane Output for Bush speech; Gore speech

Comments: Our evaluation team rated Diction as one of the easiest programs to use.  It required only a few mouse clicks before performing all key functions at the same time.  Diction requires little advance text preparation, as mentioned--the only mandatory requirement is that each text document be saved in text format (with a ".txt" extension).

In order to make normative comparisons possible, the newest version of Diction analyzes 500 words at a time and averages the 500-word analyses together for the final output.  This is a significant improvement over previous versions of Diction, which would only analyze the first 500-words of long texts or break the texts into 500 word units, giving several separate analyses for a single text.  The averaging feature new to Diction 5.0 is clearly the best way around the "500-word" problem, but we encountered some difficulties with it.  Sometimes, Diction would stop in the middle of segmenting a text and only return the average for a 500 increment of the total word count.  For example, one of our convention speeches was 3,724 words, but Diction only analyzed 1,000 of those words initially.  After resaving the speech file in a different ".txt" format, Diction eventually analyzed the text completely.  But this strange quirk still remains, so users should beware and look at the total word count after running Diction to make sure it's not a 500 words, 1,000 words, or another 500 word increment, since that's a sign that Diction did not do a full analysis.  If only a partial analysis is done, try resaving the file in other .txt formats (e.g., ASCII DOS text).  Also, Diction still gives users the options of analyzing just the first 500 words of a text or the whole text in 500 word segments.  We're not sure why anyone would want to do this, however, so users should also be cautioned about these additional options, especially when strange things happen.

Diction’s built-in dictionaries offered an intriguing, easy way to analyze the speeches, and the “normal range of scores” output let us compare the Bush and Gore speeches to the speeches of past presidential candidates.  In previous versions of Diction, all normative comparisons were to political texts, but a major addition to Diction 5.0 is multiple types of comparisons, which makes Diction useful not only to political scholars but also to educators, journalists, literary scholars, and businesspeople, to name a few.  But here's another caution: the normative comparisons can be changed after running an analysis simply by clicking a button on the output screen, which immediately changes some of the information in the Data View Window Pane.  But the text must be re-analyzed each time the normative comparison is changed.  Otherwise, part of the report file will have the new comparison and part will have the original analysis results, which can result in strange, confusing and incorrect scores.

Overall, Diction performed some useful and unique analyses for a reasonable price.  If it seems to suit your text analysis needs,  we recommend trying the program by downloading the demo version on the Diction website, accessible by clicking the URL above.    

Program Overview:

Aesthetic Description: 

Text Preparation Procedure:

Analysis Process:

Program Output Description:

Sample Output: Alphabetical, Frequency

Comments: VBPro efficiently performed all of its key functions.  It required a minimal amount of text preparation before being used—each case (for our sample text, each major heading in the Unabomber Manifesto) was marked with a “#” followed by a number or string of numbers (e.g., #001, #002, etc.).  The text was then saved in .txt format and run through the cleaning and formatting function provided by the program author.  All of VBPro’s functions had to be run separately, and coding took about twenty minutes.  On a positive note, all cases could be combined into a single text file and analyzed at once, which makes VBPro an improvement over other, size-restricted programs. 

On an aesthetic note, VBPro has a rather clunky, keyboard-driven DOS interface that may take Windows users some time to get used to.  But VBPro does what it claims to do as well as many slicker, better-looking programs.  And unlike most other programs, it’s free.  VBPro has been the preferred program of the author of this book for years, for both academic and commercial consulting purposes.  It also seems to be the favored program of communication scientists in general; more scholarly papers and publications in the field of mass communication have been generated from VBPro than from other computer text analysis programs currently on the market.  VBPro’s functions are somewhat limited, and it may not perform certain analyses needed by linguists, psychologists, or individuals in other specialized disciplines.  Overall, however, VBPro is an fine program.

Back to The Content Analysis Guidebook Online HOME PAGE

Kimberly A. Neuendorf