Beruflich Dokumente
Kultur Dokumente
Alwin J. Bethel
TABLE OF CONTENTS
Abstract ............................................................................................................................................3
Introduction ......................................................................................................................................4
Background ..................................................................................................................................4
Problem Statement .......................................................................................................................4
Purpose of Study ..........................................................................................................................4
Literature Review.............................................................................................................................5
The Precedent ...............................................................................................................................5
The Data Fluency Problem ...........................................................................................................7
Methodology ..................................................................................................................................10
Phase 1. Web Scraping & Data Collection ................................................................................11
Phase 2. Preparing the Model, Text Mining, & Integrating Natural Language Processing .......12
Phase 3: Experimentation: Introduction of New Data ...............................................................13
Evaluation: Inspecting the Precedent & Limitations of this Project ..........................................15
Evaluation: Implementing the Data Science Method as a Solution (Theoretical) .....................15
Experimental Findings, Evaluating Error, & Future Improvements..............................................16
Conclusion, Limitations, Interpretation, & Recommendation .......................................................19
Conclusion & Limitations ..........................................................................................................19
Interpretation ..............................................................................................................................20
Recommendations ......................................................................................................................21
References ......................................................................................................................................20
Appendix ........................................................................................................................................22
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 3
Abstract
The first recorded recognition of near-death experiences (NDEs) traverses history to the times of
Plato, and is acknowledged in his work “The Myth of Er” (IANDS, 2017). However, today the
study of near-death experiences is relegated as a quasi-science and categorized into the field of
metaphysics. When almost 19 million Americans, since 1992, have had a near-death experience
is it prudent or wise to consider near-death experiences anecdotal or consign the field as pseudo-
scientific (NDERF, n.d.)? My interest in near-death studies began with an unintentional click on
this click would burgeon an interest into a passion. Recently, a heated argument with a friend
pushed my interested further and I began defending the knowledge I gained over the past
months. My intention to apply the knowledge and skills I have gained in the Master of
personal interest in the subject, but additionally due to the profound lack of meaningful analytics
to the agglomeration of transcribed near-death experiences, prevalent today. I intend to mine the
NDE scope of the internet to harvest collected data on the subject, clean and manipulate the data,
analyze the data, and present my findings. I am seeking to consolidate large sums of personal
accounts and perform text analysis to determine trends and insight. Based on the findings in this
project, it is clear that near-death experiences contain common trends and experiences. This
knowledge coupled with the data science method can be expanded to the NDE field of research.
keywords: near-death experiences, analytics, data mining, data manipulation, data science, text
Chapter 1: Introduction
Background
consciousness apart from the body occurring at the time of actual or threatened imminent death”
physical body at a time during or close to clinical death. Very frequently, the physical body will
have medically died, and the near-death experiencer will be subject to induced resuscitation or
throughout history and have crossed societal and cultural barriers (IANDS, 2017). In this project,
I will use transcriptions of past near-death experiences and employ statistical and analytical tools
to perform analysis on this data. I hope to shed light to the field of near-death studies by
experiences are not simply a medical anomaly and experiencers are depicting verifiable
occurrences.
Problem Statement
Despite decades of research and the application of research methods to the field of near-
death studies, there exists no real application of analytics to this realm of study, which
death studies, which ostensibly is non-existent or lacking. I intention to use concepts of data
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 5
science, big data analytics, text mining, and natural language processing to discern trends within
raw data. I would like to prove the capacity of applying tools like Python and libraries like nltk
to analyze large sums of data, and how it can benefit the NDE field and the holistic general
population. The intention to perform a study like this on a seemingly ambiguous elusive field has
motivations which point to my own intrinsic interest in the field. I seek to complete this project
for my own personal gain, as I have doubts about what I’ve learned and intend on using my
findings to illuminate these uncertainties. Even though I do not have any associated
organizations or universities working with me, I hope to be able to present my findings should an
The Precedent
The two primary organizations, and their corresponding resources, I used to execute this project
are the International Association for Near-Death Studies (IANDS) and The Near-Death
Experience Research Foundation (NDERF). Each of these organizations seek to strengthen the
near-death experience field of study with thorough research and empower experiencers of such
phenomena with knowledge and relatability. Each organization has an online portal that collects
voluntary data through the means of online surveys and questionnaires that ask highly probing
personal questions about a subject’s respective near-death account. The procedure of research
entails employing inquiries concerning (not limited to): a subject’s religious views before and
after a near-death experience, the contingency of meeting and interacting with non-human beings
or entities, the presence of emotional feelings and experiences throughout stages of the near-
death experience, inter-spatial existence in accordance to the subject’s body and environment
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 6
outside the body, encounters with a reality of existence outside the human body, and/or
heightened awareness due to existence outside of human body. A link to the IANDS near-death
2017), and the NDERF near-death experience questionnaire can be found here,
organization founded and run by Dr. Jeffrey and Dr. Jody Long, a radiation oncologist and
organization that holds summits each year in order to bridge the gap between NDE awareness
and research. Each organization has surveyed and archived a vast collection of near-death
accounts, which are documented as the questions and answers to each of the online
can be accessed for analytical procurement or seemingly even employ an analytics division to
their organization. IANDS has documented and recorded statistics concerning near-death
experiences and I endeavor to use these finding to gain a better perspective over the course of the
study. IANDS, based on their statistical research, lists the 15 most common characteristics of a
near-death experience contains one or two of the following elements and in some cases all of
them:
“a sense of being outside one’s physical body, sometimes perceiving it from an outside
realm, which may include vividly memorable landscapes; encounters with deceased
loved ones, spiritual beings and/or religious figures; knowledge of the nature of the
a sense of having knowledge of the future; messages regarding life’s purpose” (IANDS,
2017).
duo. Consequently, the NDERF website attempts to bridge the gap between near-death
experiences being seen as a paranormal quasi-science and the application of scientific research
and medical resolve to answer common near-death experience questions. A typical answer to
near-death experiences involves relegating the notion of an NDE to a phenomenon of the brain
lacking oxygen (hypoxia) or dying brain cells (cerebral anoxia) (NDERF, n.d.). This project
seeks to provide a more informed answer as to the reasoning behind why a near-death experience
occurs and why a subject is able to “escape” death. Analytical methodology (as to what
corresponds in the latter mentioned section, Methodology Overview) will be employed to better
understand the pretense behind a near-death experience and common associations in having a
near-death experience by analyzing text from what is considered big data. Additionally,
analytical methodology will be employed to discover trends through big data analytics that may
or may not have been seen before. Statistical research methods tend to employ corresponding
sample-population based modeling, while big data modeling analyzes raw data that takes on the
characteristics of volume, velocity, and variety. Hence, this project endeavors to advance
The question must be asked, why is the field of near-death studies lacking analytical
thought and application? Data fluency is defined as “the exchange of ideas with data as a
medium” (Gemignani, Galentino, Schuermann, & Gemignani, 2015, Location No. 905). The
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 8
apparent lack of data fluency in society, particularly in the study of near-death experiences,
foments a clear lack of ability to consume such data effectively and efficiently. It may be
attributed as the reason behind the lack of analytics in near-death studies, a field that should be
driven on seeking trends and grounded upon analytical thought. Fluency of data demands
producers of data: understand the proper expression of data, are proficient and strategic in the
production of data, be involved in a data-centric organizational culture, discern what data tools to
use and when, and apply systems to properly acquire and manipulate data. (Location No. 861).
Gemignani et al. explores the need for data fluency within a field of study or organizational
culture and deeply explains the meaning and context of data fluency. Gemignani et al. contends
data fluency is imperative at both the personal and organizational level to be effective in gaining
insight, developing and actualizing trends, and consolidating data for consumption (Location No.
489).
In general, data mining is relatively simple but manipulating and formatting data in a way
studies, there currently exists no consolidated housing of data, predominantly due to several
dissociated organizations studying the same concept. Unfortunately, the data of near-death
experiences are relegated to data silos, as a result of the Balkanization of the field, illustrated by
Gemignani. (Location No. 1131). The field is abundant with people who can interpret and
understand the context of raw data, but lacks problem solvers and researchers who are able to
meaningfully consume and consolidate the data in a way that provides use to the field. “Data is
useful when people use it to tell stories, craft compelling visualizations, and construct thoughtful
analyses. People are the missing ingredient” (Gemignani et al., 2015, Location No. 331).
Leaving the role of data fluency to a specific group of people promotes data elitism by relegating
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 9
data into a silo. Instead, Gemignani et al. states the data elite should empower groups and fields
of study by building bridges and groundwork and promoting a shared data culture that essentially
allow others to create meaningful data themselves (Location No. 1185). This concept is what is
intrinsically missing from the near-death experience arena. Instead of a shared culture of
Carande, Lipinski & Gusher (2017) of the Harvard Business Review assert that effective
data and analytics (D&A) start at the top (p. 3). They contend leadership should be profoundly
involved in delineating intentions across the culture of an organization, and in this case, a field of
study. They also assert that relegating contingent decisions to a small few can create intellectual
silos that produce competitive technologies, multiple goals, altering truths and analytical
recession within one a group. (Carande, Lipinski & Gusher, 2017, p. 3). The study goes on to
mention how certain organizations have potential to drive data and analytics, but limit data
interpretation and manipulation to a certain department. They assert that some organizations are
too reliant on software tools or smalls groups of people of higher expertise. This type of
modeling prevents organizations from producing transformative data and analytics. The studies
continue to intimate that IT professionals remain lacking specialized skills in data science, due to
executives disregarding and miscalculating the implication of need for quality data and analytics
producers. When this occurs enthusiasm for data and analytics can flounder, and essentially the
acumen and agility needed to propel the organization forward can also flounder, preventing goals
from being reached and deliverables from being established. (Carande, Lipinski & Gusher, 2017,
p. 2). Specifically, near-death experience researchers tend to be sole researchers and are often
experience research organization often exists as a smaller organization that lacks data fluent
Chapter 3: Methodology
Overview
I followed a loosely scientific approach for this project, using step-by-step trial and error
hypothesis, I used the initial problem statement posed in my introduction as a guide. I resolved to
experiment with what would work, what wouldn’t work, and why. I continued finding solutions
and improving my model as the project progressed. Initially, I sought to find a viable solution by
adhering to online tutorials and guides. Ultimately, I applied strategies in text mining and natural
language processing, and a number of different tools for acquiring a viable solution to the
proposed course of planning for this project. The mid-semester mark was a low point for this
project, and I asked my mentor for other potential directions for this project to head to, but after
modifying the project I continued to follow my initial plan for this project. Aforementioned, this
project is of personal interest and will continue beyond the scope of the Capstone in Informatics
course but serves as a cornerstone to further development. My goal for this project is to perform
big data analysis on a large database of unstructured text. According to the staggering figures,
aforementioned, a large percent of the population has undergone a near-death experience, and
transcribed full-text versions of these experiences are recounted and posted online. I would like
to use text mining and natural language processing to analyze this raw text data and discover
potential trends that are seemingly common throughout these experiences. I will establish
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 11
commonalities by looking for frequently occurring words in these experiences. As the project is
iterated into different phases, the following sub-sections describe the methods used - from
acquiring the data to applying text mining strategies to actualize solutions to my problem.
improvement benchmark and would be used as a means to discover trends in text. I realized later
that a sentiment analysis is used for measuring the general feeling (sentiment) of something,
commonly depicting a person’s opinion about a topic. In the context of this study, I felt that a
sentiment analysis may not be the appropriate method for achieving my goal. The data source
that I was using exists as narratives and personal accounts of near-death experiences, and one of
the goals of my project was to employ text analytics for analysis of these narratives, actualizing
trends in the text. I was not sure how to go about doing this, as I had never done any text
analytics in my degree, and so the search for the right tool and method of achieving this was a
tedious process. I specifically looked for open-source tools and researched a number of different
options and watched videos on how to operate these tools for text analytics. I eventually decided
to stick with what I knew – Python. My goal was to analyze the text from my data sources by
individual words (opposed to certain phrases). First, I had to determine how I could extract plain
text versions of the text I was trying to analyze. I searched endlessly, and the best tool I could
find was something called Textise (https://www.textise.net), which allows for a user to paste a
link to the targeted website of choice and return a plain text version. However, a challenge that I
faced was trying to find a solution to automate mining a website AND its directories within that
specific link. For example: the Near-Death Experience Research Foundation (NDERF) website
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 12
contains links to other pages with full accounts of NDEs. I wanted to find a way download the
summation of this information into a plain text format. I conducted further research and found
the tool, Octoparse, “…a modern visual web data extraction software. Both experienced and
inexperienced users would find it easy to use Octoparse to bulk extract information from
websites, for most of scraping tasks no coding needed… Octoparse simulates human operation to
interact with web pages” (Octoparse, n.d.) This tool allowed me to enter the URL of a website
and allow primary and subsequent plain-text downloads of the main website and its respective
directories, output in HTML format. I was able to download 189 HTML pages of plain text from
the Exceptional NDEs page of the NDERF website. I then imported the contents of these pages
into Microsoft Word, where I was able to compile and conglomerate this information and output
it into a text file. In a Word document with 1-inch margins, the complete contents include 1,179
pages of information containing 967,134 raw words of text. I output a text file,
Phase 2. Preparing the Model, Text Mining, & Integrating Natural Language Processing
Next, I installed the stopwords using the nltk.download() function (Line 55).
The NLTK downloader menu appears, and under the Corpora menu, choose stopwords by
stopwords.words and applied the English parameter (Line 56). The stopwords
function works by removing commonly used parts of speech, like articles (“the”, “a”) or
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 13
conjunctions (“and”, “because”), etc. from a variety of different languages (Natural Language
Toolkit, n.d.). The replacement of words created null strings in the dataframe, so the
df.replace function was used to replace null strings with np.nan (Line 57). Then
df.dropna was used to remove all NaN values from the dataframe (Line 59). Additionally,
df.drop was further employed to remove a redundant column (Line 60). From there,
counter from collections was used to return the 50 most_common words in the text
./:;<=>?@[\\]^_`{}~', were removed from the dataframe (Line 63). In Line 65, I
employed the lower() method to normalize capitalization anomalies amongst the text
(example: “Time” to “time”). By completion of these steps, the raw data that was collected was
cleaned of common stop words, punctuation marks, and capitalization anomalies, and should
The last phase of my experiment included the introduction of new data to the model and
comparison. From the cleaned text in the previous phase, I changed the dataframe into a list,
iwordzmostcommon, ordering the words by frequency (Line 66). Despite the name of this
list, it contained all words in the dataframe, and not just the most common. I originally used the
200 most common words, but changed it back to the full list to include the full scope of words. I,
however, kept the variable name the same in hopes of preserving the model and preventing it
from “crashing” due to repetitive variable name changes. I treated this list as a sort of “ground
truth” and used it as the measure to compare all new text introductions to the model. This list of
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 14
words (over 900,000 words) originated from the Exceptional NDEs page, and contained data
from near-death experiences as early as the 1950s and as late as 2018. The last phase of my
experiment involved introducing seven new sets of text, obtained from the NDERF website and
comparing it to the iwordzmostcommon list. These sets of texts were randomly selected and
originate from a range of different dates. I wanted to be intentional about selecting texts from a
range of different years, to see if there were any differences in common words depending on the
year the near-death experience occurred. Iteration 1 was an NDE that occurred in [undisclosed
month] 2012, Iteration 2’s experience occurred in August 1971, Iteration 3 occurred in
November 2001, Iteration 4 was an NDE that occurred in June 1978, Iteration 5 occurred in
April 2005, Iteration 6’s NDE occurred in [summer] of 1957, and finally Iteration 7 was an NDE
that occurred in an undisclosed month of 1988. For each iteration, I began with loading a new
text file (first iteration in Line 67) called newintroduction.txt (append number to end of file name
for each iteration). I then converted the text into a Pandas dataframe, removed all stopwords,
replaced nulls, cleaned up punctuation, and removed capitalization anomalies. I took the 200
most common words from each iteration (inewwordzmostcommon – append number for each
iteration) and compared it to the original list iwordzmostcommon. I took the 0th column from
each list and converted each into a respective dataframe, df1 and df2. I compared each list by
merging (essentially an inner join) the two collections together and looking for anomalies. From
the sources I looked at, the most effective way of compiling two lists and determining
commonalities was to use the merge function (stackoverflow Finding, n.d.) (stackoverflow
Pandas, n.d.). I repeated this process seven times, and for the final step, combined each of the
most common lists and ordered by the 200 most common words.
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 15
To reiterate the past sections, I spent time parsing and cleansing the original source of
data (the WordFull.txt file) to ensure that any anomalous data (stopwords, punctuation,
capitalization) was removed. This process was to ensure that I have a full raw set of words for
comparison to the new data that was introduced. It was compiled and arranged by the most
commonly used words in 189 transcribed NDE accounts. The purpose of compiling this list was
to collect a large sum of data first and organize it for comparison to new introductions of text.
New text was then introduced and compared to the established list, looking for commonalities
based on frequency of word usage. I would like to note: before this project began, I reached out
to one of the leaders of IANDS, a professor at the University of North Texas, with no response. I
intentioned to share my findings and collaborate with the field of near-death studies in hopes of
applying analytical thought to this field. However, I lacked success in getting in touch with a
representative of this field, and consequently the true application of the data science method can
only occur through theoretical thought at this moment (outlined below). Potentially in the future,
as this project continues, I may be able to collaborate with either of the two organizations listed,
NDERF or IANDS.
The Data Science method consists of the application of big data analytics coupled with
the adoption of a data fluent culture. This application of big data analytics and analytical tools, as
university. Per this project, the Data Science method is the key to the advancement of the field of
near-death studies as, in my opinion, big data analysis and trend actualization on a large scale is
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 16
the missing element to true analysis of near-death experiences (compared to analysis on separate
cases or without analytical tools like Python and its respective libraries involved). This project
can be seen as a stepping stone to the potential future applications of data science within the field
of near-death studies, as it bridges the gap between manual human analysis (potentially time-
consuming and lacks efficiency if performed in tool like Excel) and machine learning. Using a
Python library like nltk can prove extremely useful in parsing text information and providing
analytical thought processes to the field. Moreover, employing a data scientist or even a team of
analysts can further the research sector of this field as it provides a means of actualizing trends.
Compared to one person manually parsing, cleaning, and reducing raw transcription data,
application of data science and analytics has the potential to further this field immensely. In my
opinion, analytics is the key to reversing the commonly designated misconception that the field
The intention to text mine and apply concepts of natural language processing to the
original and subsequent text data was to discern trends within the data. The seven iterations of
the model output seven dataframes containing the most frequently used words in each set of new
texts (see Figure 1). Unsurprisingly, the word “experience” was commonly found in the list of
most frequent words. This is clearly due to the topic being near-death experiences and the word
experience being a common theme amongst these text transcriptions. Other frequently used
words include irrelevant parts of speech like the word “yes”, “it”, and “I”. The nltk stopwords
function was employed to remove words like these, but these words may or may not have been
included in the list. Additionally, towards the end of this project I inferred that a potential reason
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 17
why these certain words were not removed was due to the stopwords function being applied to
the dataset before cleaning it. This is important, as anomalous data like capitalized words
(example, “I”) or words with punctuation attached (example: “it.”) may not be recognized by the
stopwords function’s list of words to remove. In a future iteration, I would normalize the data
before applying natural language tools like stopwords. I also noticed that one of the most
commonly used words returns a null string, and this is due to how the system parses a sentence.
Since sentences require spaces, the null values can’t be removed without intricate code additions,
and so it must be ignored for this project (may be applied to future projects).
The final phase of this project compiled the list of frequently used words in each of the
independent sets of text and parsed the dataset to find the most frequently used words within this
compiled list. Essentially, this step was implemented as a means of taking an average of the data.
I took the top 30 most significantly recurring words in this compilation, ignoring words that are
common parts of speech like “the”, “what”, “I”, etc. The list of top 30 recurring words
(predominantly nouns and verbs) includes: time, life, body, feel (and derivatives), remember,
things, way, light, see (and derivatives), everything, know (and derivatives), people, thought,
beings, ever, unearthly, religion, part, tunnel, future, past, scenes, experience, love, god, and
information. Out of this top 30 list, 20% of these words occurs in all 7 sets of text (time, life,
body, feel (and derivatives), remember, and things) and 93% of words occur in at least 6 out of 7
of these sets of texts (see Figure 2). This data (from text that was collected completely randomly)
is indicative of showing trends. I predict a future implementation of this project that can analyze
a large number of new introductions will show nearly the same information.
Some things that may be improved upon in future iterations of this project are following.
I noticed toward the end of this project, while parsing through some of the raw data, that
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 18
reasoning for some commonalities between text sets may be attributed to the website’s listing of
common questions for each NDE page, such as “Did you pass into or through a tunnel?” or “Did
you see any beings in your experience?”. However, this implication only affects the data
minimally as most of the words used repeatedly in the question section of each NDE page, is
only repeated a few times. For example, in the first iteration, significant words like life and light
were used 26 and 24 times, respectively, throughout the full text. In comparison to the frequency
of most of these words, words in the question sections only account for a fraction of the times
they were used throughout the complete text. An exception would be the word “experience”,
which was used repeatedly and potentially is accounted for just by being used in the question
sections. In future iterations, I could implement additional code to remove unwanted words,
especially those that contribute to anomalous data and words that the stopwords function failed
The intention of this study was to comprehensively assess how data science and analytics
could potentially benefit the field of near-death studies. Extensive research in the literature
review section resolves these claims made in this project report. Methods and experimental
findings demonstrate how potential analytical manifestations may improve this field, taking into
context the limitations of this project. Consider that this project was designed, coordinated, and
implemented over the course of a few months (a university semester) by only one person
dedicating a few hours a week towards the goals of this project. Conceptualize the profound
effect that adopting the practices and tools outlined in this project may have on this field of
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 19
research. I found through the introduction of raw text data, several commonalities were found
within the transcripts of each of these near-death experiences. To sum up my findings, I included
a wordcloud that was used depict the compilation of words found in my WordFull.txt document,
towards the beginning of this project. This wordcloud displays the most emphasized words in the
document and reflect my findings towards the latter phases of this project (see Figure 4).
Interpretation
The question must be asked: how could seven randomly selected instances of near-death
experiences, ranging back 61 years, show commonality with words in these transcripts indicating
emphasis on words like: time, life, light, beings, religion, tunnel, love, god, and information? In
the context of near-death studies, the tunnel and the light are extremely common sights that are
experienced. Experiencers depict their consciousness moving from a physical body into a
perceived tunnel, and then approaching and becoming one with the energy of a light that
overwhelms them with feelings of love, happiness, acceptance, and elation, no matter what
religion or background they are from. Aforementioned, many relegate the idea of the light and
the tunnel as attributed to hypoxia but fail to realize that the overwhelming majority of
experiencers see sights and sounds that eclipse this claim. How can large populations of near-
death experiencers have clear recollection of events like having a life-review or interacting with
deceased relatives? (NDERF, n.d.) In my opinion, further applications of data science and
Recommendations
The methods and findings detailed in this report indicate positive potential to near-death
studies. Future implementations of this project may begin with analyzing more near-death
experiences, taking into account the improvements listed in Chapter 4. I believe that branching
out from single words and including phrases like “life review” (common feature of NDEs), may
prove useful analytically. Future iterations and improvements on this project may include
additional manifestations including, but not limited to: a consolidated database of information
spanning several organizations, text mining from non-text sources like YouTube videos, and/or
References
Carande, C., Lipinski, P., & Gusher, T. (2017, June 23). How to Integrate Data and Analytics
into Every Part of Your Organization. Retrieved September 20, 2018, from
https://hbr.org/2017/06/how-to-integrate-data-and-analytics-into-every-part-of-your-
organization.
Gemignani, Z., Galentino, R., Schuermann, P., & Gemignani, C. (2015). Data Fluency:
Han, Jiawei, et al. Data Mining: Concepts and Techniques. Morgan Kaufmann Publishers, 2012.
International Association for Near Death Studies (IANDS). (2017, August 29). Key Facts about
facts21.html?start=1.
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 21
Müller Andreas C., and Sarah Guido. Introduction to Machine Learning with Python: a Guide for
Natural Language Toolkit. “Accessing Text Corpora and Lexical Resources.” NLTK Book,
Near-Death Experience Research Foundation (NDERF), & Long, J. (n.d.). Frequently Asked
Near-Death Experience Research Foundation. (n.d.) “How Many NDEs Occur in the United
www.nderf.org/NDERF/Research/number_nde_usa.htm.
Octoparse. “Octoparse Tutorial.” Web Scraping Tool & Free Web Crawlers for Data Extraction |
rows-intersection-in-two-pandas-dataframes.
stackoverflow. “Pandas - Intersection of Two Data Frames Based on Column Entries.” Stack
of-two-data-frames-based-on-column-entries.
van Lommel, Pim. “About the Continuity of Our Conciousness.” Near-Death Experience
Appendix
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 23
Figure 1. (above) Compiled list of most frequently used words in each of the seven iterations of
raw text transcripts, arranged by frequency. The higher the item is on the list, the more
frequently it appears in the respective text.
Figure 2. The most frequently occurring words in the final compiled list of all seven iterations of
raw text, and each word’s respective frequency. Essentially shows an average of the data, and
indicates that 93% of significant words occur in at least six of the seven sets of text.
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 24
Figure 3. An illustration of the nltk downloader tool that appears when employing the command,
nltk.download(). To install stopwords, select the Corpora tab at the top, scroll down to
stopwords, click on it, and click download. Once it is finished, close this downloader tool and
resume in your code.
THE ANALYTICS OF NEAR-DEATH EXPERIENCES 25