Creating a Comprehensive Semantic Search Engine for Parapsychology and
Anomalistics: A Proof of Concept
Michael Tremmel
Bender Institute of Neuroimaging, Justus Liebig University Giessen
Giessen, Germany
—
The initial goal of the project presented was to create a chatbot specializing in parapsychology and anomalistics. In the last couple of years, chatbots powered by “large language models” (LLM, that is, machine learning models that enable processing of natural language and generating of natural-sounding replies to queries), such as Bing, ChatGPT, or Gemini, have become more advanced and widely adopted. However, even popular chatbots may output superficial replies, in particular when asked for knowledge at an academic level, or worse, replies that are partly or wholly false (so-called “hallucinations”) while potentially sounding plausible (Huang et al., 2025; Xu et al., 2025). A way to counteract these shortcomings is to use so-called “retrieval augmented generation” (RAG). In RAG, an LLM may be complemented with a so-called “text embedding model” (TEM). A TEM is a model that enables the retrieval of exact quotes (transformed into vector embeddings, see below) from a text corpus database that can function as replies to a query. One way to realize RAG is the following: First, a query is made and processed. With the help of a TEM, quotes relevant to the query and references of the quotes are identified and retrieved from a text database. Next, with the help of an LLM, the query, quotes, and references are processed, and a reply to the query is generated, based on the LLM’s data as well as the quotes extracted. The output consists of the reply complemented by the quotes and corresponding references.
RAG (using TEM plus LLM) can improve replies given by an LLM, but cannot prevent the occurrence of hallucinations altogether (Huang et al., 2025). On the other hand, retrieved quotes and their references are a trustworthy, hallucination-free source of information and can facilitate academic information search on their own. Thus, the goal of the project changed to creating a semantic search engine that can retrieve such quotes and their references (using TEM on its own), and the use of an LLM became an optional add-on.
—
Michael Tremmel is a psychologist and a doctoral student at the Bender Institute of Neuroimaging at the University of Giessen. He is writing his thesis on adverse and unexpected experiences with meditation.
—
The Parapsychological Association is an international professional organization of scientists and scholars engaged in the study of psi (or ‘psychic’) experiences, such as telepathy, clairvoyance, psychokinesis, psychic healing, and precognition. The primary objective of the PA is to achieve a scientific understanding of these experiences.
Follow the PA: https://linktr.ee/parapsychologicalassociation
Donate to support the mission of the PA: https://donorbox.org/support-the-parapsychological-association
Published on October 16, 2026
