1. bible
2. lord
3. christ
4. gospel
5. psalm
6. god
7. psalms
8. corinthians
9. preach
10. shall
11. heaven
12. readings
13. church
14. spirit
15. righteous
16. verse
17. lectionary
18. verses
19. spiritual
20. ministry
21. pray
22. enemies
23. thou
24. tongue
25. creation
26. wisdom
27. deuteronomy
28. testament
29. strength
30. refuge
31. therefore
32. kingdom
33. romans
34. holy
35. thankful
36. thy
37. reading
38. rejoice
39. understanding
40. faithful
41. message
42. earth
43. blessed
44. exodus
45. deut
46. faith
47. wise
48. beginning
49. pastor
50. chapel
51. chapter
52. survey
53. anger
54. resurrection
55. risen
56. read
57. hearts
58. chronicles
59. salvation
60. flesh
61. servant
62. glory
63. praying
64. kings
65. sheep
66. praise
67. trust
68. prosperity
69. bless
70. heavens
71. deeds
72. toward
73. discussion
74. whoever
75. speaks
76. ye
77. hath
78. amen
79. teaching
80. thess
81. apostles
82. preparing
83. eph
84. eccl
85. path
86. fear
87. upon
88. presence
89. inspire
90. search
91. zechariah
92. seek
93. teach
94. wrath
95. commandments
96. believers
97. humility
98. spoke
99. thee
100. devo

Background

Extracting Bible references from text means identifying whether a given piece of text is referring to a Bible verse or something else. For example, the meaning of Acts 2 depends on context:

• Referring to Bible passage: Acts 2 recounts the early church.
• Not referring to Bible passage: She’s 5 years old but acts 2.

When you encounter a phrase that could be a Bible reference, you have to look at context to determine whether the phrase is a Bible reference. Humans can make this leap pretty easily, but computers need rigorous models and lots of training data to guess whether an ambiguous phrase is a Bible reference. In the above example, the phrase “early church” is a strong indicator that the phrase “Acts 2″ is a Bible reference, while the phrase “years old” is an indicator the other way.

Twitter, with its high volume of content and decent search engine, provides lots of training data.

Methodology

Using the Twitter Search API, I downloaded 30,000 tweets possibly containing Bible references (e.g., [john 3], [jeremiah 29]) and then categorized them by hand as referring to a Bible verse or not.

I then ran a Naive Bayes algorithm on the resulting tweets to produce the above list, which contains the words that most strongly indicate the presence of a Bible reference.

This list suffers from sample bias, of course: a different set of tweets would produce a different list. In addition, the list is Twitter-centric; the results may not carry over into blogs or other media. (People substitute the number “2” for the word “to” and “4” for “for” on Twitter more frequently than they do elsewhere, for example, which oversamples content like “I’m meeting Matthew 4 dinner.”)

See It in Action

Search for Bible references on Twitter. Use the relevant and not relevant buttons to improve the filtering. I haven’t formally announced this new feature of OpenBible.info yet; consider the link a preview.

The verses appear in descending order of popularity (as determined by an analysis of verses on this site), so if you’re only looking for 100 verses, you can just grab the first 100. There are 750 total to give you more than two years’ worth of daily verses, if you want.

Background

When I was putting together a list of daily memory verses a few years ago, I was surprised to discover that I couldn’t find a freely usable list of verses: I ended up combing through a bunch of books and combining various sources to produce a list of about 400 verses. If you’re doing something similar, why should you go through the same pain?

The goal is to see how well this kind of data tracks official marketshare figures. Unfortunately, I’m not aware of any publicly available marketshare figures. (If anyone knows of any, feel free to leave a link in the comments.) In principle, this chart should provide a leading indicator of sales—if you accept the idea that mindshare precedes sales.

The result more-or-less tracks the CBA Bible Bestseller List (pdf), with a couple of exceptions. The chart appears to undercount the NKJV and the HCSB, which means that either (1) there’s a data problem—for example, some NKJV results may appear as part of the KJV data; or (2) these translations’ sales in CBA bookstores overstate their prominence as translations—in other words, they sell well in CBA bookstores, but people tend not to search for them much in Google. I lean toward the first answer: it’s likely that my methodology is flawed somehow.

Technical Background

The chart blends data from Google Trends and the Google Keyword Tool to see how many searches people conduct for various English Bible translations. Data before July 2008 is from Google Trends exclusively, while more recent months include data using the methodology described in Bible Translation Mindshare.

Originally, I wanted to use only this latter methodology. However, the data is chaotic; the figures jump around quite a bit, and the Google Keyword Tool sometimes gives inconsistent results even for the same search in the same month. For this reason, I decided to smooth the data a bit with Google Trends, which provides more consistent data from month to month.

Finally, I should mention that I have lots of potential conflicts of interest in producing this data.

I thought I’d see how many people search for various Bible translations. The results:

Translations

The translations come from the CBA Bible Translation Bestseller List. (CBA is the Christian Booksellers Association, a Christian-retailing industry group.) This list counts Bible sales through most Christian bookstores in the US. However, it’s an imperfect measure of a translation’s popularity since sales through non-CBA stores are becoming more important to Bible publishers.

The imperfection of the CBA list (and the desire to quantify its rankings) was the inspiration for this (also-imperfect) project.

Methodology

Bible version abbreviations that also have other meanings were a little tricky. “HCSB” is also the name of a bank and a school board in Florida. “ESV” is also a Cadillac and the stock symbol for an oil-drilling company. To compensate for these other uses, I took the top three non-Bible keyword phrases and compared the numbers for the top-three Bible related keyword phrases.

For example, there were 90,500 searches for “esv” in June. The keywords “escalade esv,” “cadillac esv,” and “cadillac escalade esv” had 57,300 searches, while “esv bible,” “esv online,” and “esv study bible” had 13,300 searches. Multiplying 90,500 by 19% (the proportion of the top three Bible-related searches to the top three non-Bible-related searches) leads to the conclusion that about 17,049 of those 90,500 searches for “esv” were Bible-related. It’s definitely an approximation, but the alternative was to count acronym-only searches for some translations and not for others.

Besides the ESV and HCSB, other Bible-translation acronyms have a few other meanings, but they don’t suffer from nearly the level of acronym confusion as those two translations. The HCSB also goes by the name “CSB,” which makes things even harder. Nothing Bible-related showed up when I entered CSB into the keyword tool, so I didn’t count it.

A similar problem occurs for the phrase “king james”—with 301,000 searches in June, it was the single most-popular translation-related query. But are the searches for the Bible translation or for, say, King James I of England? Based on the popularity of the related keywords, it looks like 97% of “king james” searches are for the Bible. This query puts the KJV well ahead of the NIV in popularity.

The Message also probably suffers from undercounting, since it’s impossible to know how many people searching for “message” are looking for the Bible (not too many, based on the Keyword Tool’s data—but definitely some).

Here are the words and phrases I started with. I included other clearly relevant synonyms in the total.

• christian standard bible
• english standard version
• esv bible
• esv bibles
• esv translation
• hcsb bible
• hcsb bibles
• hcsb translation
• kjv
• kjv bible
• kjv bibles
• kjv translation
• message bible
• message bibles
• message translation
• msg bible
• msg bibles
• nasb
• nasb bible
• nasb bibles
• nasb translation
• new american standard
• new international version
• new king james
• new living translation
• niv
• niv bible
• niv bibles
• niv translation
• nkjv
• nkjv bible
• nkjv bibles
• nkjv translation
• nlt
• nlt bible
• nlt bibles
• nlt translation
• tniv
• tniv bible
• tniv bibles
• tniv translation

Collective intelligence (sometimes called “the wisdom of crowds”) is something that I think will become more important in biblical studies as it becomes more important in other disciplines. The vast scope of the Internet makes possible explorations of questions that have been impractical until now.

So, what’s a practical application of collective intelligence to biblical studies? Paragraph and section identification to help people study the Bible.

I paid 40 people on Amazon’s Mechanical Turk two cents each to add paragraphs to the book of Philemon in both the ESV and NLT translations. Total cost (including commission): $1.00.

Results

The thickness of the bar in the below table indicates how many people would start a paragraph at the given sentence. The ¶ symbols (not sent to the Turkers) indicate where the Bible translators themselves started paragraphs.

The table shows similarities between the two translations (except for some variation near the beginning) and also reveals some divergence with the translators’ paragraph ideas.

(If you’re reading this post in a feed reader, the below table won’t make much sense to you. You should click through to the original post if you want to see the styles.)

Let’s take this approach one step further, asking people to group the paragraphs into sections (pericopes), titling those sections, and then having other people vote on which section titles are best. Unfortunately, this part of the experiment is still in-process, so I don’t have anything interesting to show you.

Practical Implications for Bible Software

Purveyors of Bible software are in a unique position to benefit from this kind of collective intelligence. Although they don’t have the massive volume of users found on the Internet, they data they compile from Bible software users will be of higher quality than data compiled from the Internet. (For one, Bible software users probably aren’t out to attack Christianity, unlike a good number of people on the Internet.)

Further, Bible software publishers can create tools to implicitly shape collective intelligence to their users’ benefit. For example, they could create a tool that allows people to add their own paragraphs and sections to Bible texts. This tool could let you share this data you create anonymously with other users of the software. The goal isn’t to create the “one true paragraph and section divisions” for a passage, but merely to inform people of the choices others have made, indicating (if someone desires) how other people organize biblical texts.

Similarly, Bible software could help people anonymously share outlines of Bible books. I think you could also rank relevant resources (commentaries, say) for certain passages based on the resources people consult after reading those passages and the length of time they spend looking at the resources.

The idea behind this kind of collective intelligence is to collect implicit data—data based on things people are already doing—to improve the software. The more people who use the software, the better it becomes. You could try to come up with a slick algorithm to predict, say, the most relevant resources for a passage. But the beauty of a collective-intelligence approach is that the abundance of data means that even a simple ranking algorithm produces good results. Or, as Anand Rajaraman puts it, “more data usually beats better algorithms.”

I’m excited about applying collective intelligence to biblical studies. Implicit-data aggregation provides the most promising short-term possibilities, since a lot of data regarding user behavior in Bible software already exists. Pragmatically, I think Bible software publishers can get a lot of mileage just by analyzing this data and allowing it to inform their interface in limited ways.

But explicit collective-intelligence projects, such as the parable finder that Sean discusses in his post, could bear fruit in the long run. The key is to get lots of data—even noisy data will do—that can serve as the basis for analysis.