Text Analytics Fun - Visualizing Sentiment Between General Conference Talks

I decided to have some more fun with the April 2014 General Conference talks. This time, I applied some basic text analytics and natural language processing techniques to estimate each talk's overall sentiment. In general, a talk with "positive sentiment" uses more optimistic type words and a talk with "negative sentiment" uses more pessimistic type words.

Below is a chart summarizing the sentiment differences for each talk, but only based on adjectives and adverbs a computer algorithm automatically identified for me. Adjectives and adverbs tend to be the types of words that convey sentiment. Surprisingly, Talk #17 -- Love—the Essence of the Gospel -- had the most negative sentiment! Go figure! Talk #16 -- Live True to the Faith -- had the most positive sentiment! (Cell phone users - click/tap image to expand and see fine text)

[1] A Priceless Heritage of Hope  [2] Are You Sleeping through the Restoration  [3] Be Strong and of a Good Courage        [4] Bear Up Their Burdens with Ease    [5] Christ the Redeemer        [6] Daughters in the Covenant  [7] Fear Not; I Am with Thee     [8] Following Up    [9] Grateful in Any Circumstances   [10] I Have Given You an Example   [11] If Ye Lack Wisdom [12] If Ye Love Me, Keep My Commandments    [13] Keeping Covenants Protects Us Prepares Us and Empowers Us [14] Let's Not Take the Wrong Way  [15] Let Your Faith Show  [16] Live True to the Faith    [17] Love—the Essence of the Gospel    [18] Obedience through Our Faithfulness

[19] Protection from Pornography—a Christ-Focused Home  [20] Roots and Branches   [21] Sisterhood Oh How We Need Each Other   [22] Spiritual Whirlwinds     [23] The Choice Generation  [24] The Cost—and Blessings—of Discipleship  [25] The Joyful Burden of Discipleship [26] The Keys and Authority of the Priesthood   [27] The Priesthood Man  [28] The Prophet Joseph Smith  [29] The Resurrection of Jesus Christ  [30] The Witness  [31] Wanted Hands and Hearts to Hasten the Work [32] What Are You Thinking [33] What Manner of Men   [34] Where Your Treasure Is  [35] Your Four Minute

I also made a second chart showing sentiment differences, but this time it was based on all words which the sentiment lexicon/dictionary could identify and was not limited to just adjectives and adverbs. With this method, Talk #5 -- Christ the Redeemer -- ended up having the most negative sentiment! Talk #13 -- Keeping Covenants Protects Us Prepares Us and Empowers Us -- had the most positive sentiment.  

Overall, this was a fun exercise to see how general conference talks vary sentiment-wise. The methodologies and computer algorithms utilized for this analysis are by no means perfect, but they are surprisingly useful to get additional insight and extract information from text.

***

Summary of methodology:

·         R was used for all analysis/visualization. Libraries tm, NLP, openNLP, proxy, RWeka, openNLPmodels.en, and stringr used.

·         Each conference talk was copy and pasted into its own text file to make a corpus of documents.

·         Harvard's Inquirer Lexicon/Dictionary was used to estimate sentiment, but with some modifications by myself. This dictionary can be accessed at these links: http://www.wjh.harvard.edu/~inquirer/homecat.htm  (general page) or directly at http://www.wjh.harvard.edu/~inquirer/spreadsheet_guide.htm

·         Sentiment was estimated for each talk individually by taking the total number of positive sentiment words and dividing it by the total number of all sentiment words (positive + negative) the Lexicon could identify.

• The mean sentiment for the group of 35 talks was 72%. I.e. on average, of all the sentiment type words identified, 72% of them were positive in the talks.