Nursing Ghost Stories as told through Natural Language Processing
Language comes natural and easy to humans but is difficult for machines because of its structure and ambiguity. Ghost stories perhaps showcase that ambiguity as paranormal experiences are difficult to reconcile or explain within known scientific principles or inferences. As a project, we wondered how might how might machine learning make sense of these kinds of experiences
Natural Language Processing (NLP) is a field of computing that enables computers to analyze, understand and communicate human language. Today natural language processing powers several technologies such as: speech recognition assistants; language translation; sentiment analysis; entity and relationship recognition; as well as text summation, parsing and analysis
The Natural Language Toolkit (NLTK) is a platform of libraries and programs for natural language processing written in the Python programming language. NLTK was developed at the University of Pennsylvania and first released in 2001
Word clouds are visual representations of a text, where the sizing of words displayed reflects their prominence or emphasis within the text. The implementation of the word_cloud application used here was developed with elements of NLTK and other Python modules
Word clouds provide high-level analysis of themes associated with a corpora (body) of text. In business, they can be used to highlight pain points from customer feedback. For this effort, word clouds were applied to a collection of ghost experiences as told by nurses
Allnurses.com hosts a long-running discussion thread called "Nursing Ghost Stories” (NGS). The NGS collection spans over a decade (2005-2017) amounting to 199 pages as of the time of this writing. The NGS archive contains a mixture of first and second hand accounts along with commentary
Two corpora were developed from the NGS collection. One corpus contained plain text and another corpus was tokenized (tagged) by sentences, words and parts of speech. A word cloud was subsequently generated from the plain text corpus that is displayed above. Common stop words, for example commonly-used prepositions, were filtered out prior to the generation of the display
The word cloud is interesting for what it does and does not emphasize. For example, the word haunted and its variations are not prominent in the display. Hauntings involve recurrent paranormal experiences commonly experienced in the form of "imitative noises" and in more elevated forms through apparitions
However, many NGS discussions and accounts involved apparitional and/or sensed presence types of ghost experiences that were “one” time or exceptional encounters; in fact, “one” was the most common word and was added to the set of stopwords to generate the above display.
Among the apparitional encounters, were a preponderance of nearing death awareness (NDA) type experiences as said or told by patients, many of whom were in long-term care or hospice settings, to nurses under their supervision, or involved NDA behaviors on the part of patients that were observed by nurses.
In their final phases, terminally-ill patients often perceive welcoming apparitions or visitations from deceased relatives or loved ones. Terminal patients experiencing death-bed visions will also appear to hold conversations with persons who are not physically present in their rooms
It has been long known that ghost encounters or more broadly spontaneous experiences (psi functioning) can occur at any time and this is reflected in the word cloud. The word "day" is almost as emphasized as the word "night" in the collection
As would also be expected, terms of reference associated with the medical profession are most prominent in the word cloud. However these terms could be filtered from future word clouds to potentially obtain deeper insights on the experiences, but for now they are central for understanding the body of stories
In the near-term, NGS corpora can be used to develop sentiment analysis. Insofar as NDA involve apparitional encounters of family members and friends, the experiences are not likely be characterized in negative terms
Another exploratory effort could use topic discovery models to characterize and refine types of encounters within the NGS collection. Word embedding models could likewise explore semantic relationships in terms of topic distances, rankings and similarities.
What emerges from machine learning is that the reported encounters do not typically involve haunted hospitals, but rather primarily reflect end-of-life experiences
REFERENCES
Alvarado, C. S., & Zingrone, N. L. (1995). Characteristics of hauntings with and without apparitions: An analysis of published cases. Journal of the Society for Psychical Research.
Bird, S., Klein, E., & Loper, E. (2009). Natural language processing with Python: analyzing text with the natural language toolkit. " O'Reilly Media, Inc.".
Gauld, A., & Cornell, A. D. (1979). Poltergeists. Routledge Kegan & Paul.
Kircher, P. and Callanan, M. (2017, Dec 14). NDEs and Nearing Death Awareness in the Terminally Ill. International Association for Near Death Studies (IANDS).
Natural Language Toolkit: NLTK 3.2.5 documentation. (2017, Sep 24). NLTK Project
Pearson, P. (2014). Opening Heaven’s Door: What the Dying May be Trying to Tell Us about where They’re Going. Random House Canada. Sponsored
What's Your Best Nursing Ghost Story? (2017, Oct 30). AllNurses.com
IMAGE
Wordcloud from NGS Corpus. (2018, Feb 19). (c) Maryland Paranormal Research ®. All rights reserved.
