By means of current and emerging artificial-intelligence techniques, stories can be processed into embeddings, into high-dimensional points or trajectories, such that stories that were similar to one another would be mapped to points near to one another in story spaces.

A brief unfolding history of story embeddings is presented, below, and some of their envisioned applications are indicated.

In 2018, Fortuin, Weber, Schriber, Wotruba, and Gross developed a novel way to encode stories from natural language into sequences of different features and showed that it was possible to use sequence models, especially recurrent neural networks, to learn from these sequences.

In 2020, Lee and Jung, with Story2Vec, focused on social networks among story characters (character networks), extending substructure-based graph embedding techniques.

In 2021, Polceanu, Porteous, Lindsay, and Cavazza explored sequence-to-sequence architectures, encoder-decoder architectures, for story generation.

In 2021, Fang, Zeng, Liu, Bo, Dong, and Chen, towards controllable story generation, integrated latent representation vectors with a transformer-based pre-trained architecture.

In 2024, Benara, Singh, Morris, Antonello, Stoica, Huth, and Gao utilized large language models to ask targeted questions of texts while mapping responses to specific features.

In 2024, Hatzel and Biemann proposed a model, StoryEmb, which generated embeddings for stories such that similar stories, e.g., reformulations of the same story, resulted in similar embeddings.

In 2026, Mitka focused on the narrative comparison task using three narrative aspects: abstract theme, course of action, and outcome.

In 2026, Bigelow, Sarfati, Wurgaft, Lewis, McGrath, Merullo, Geiger, and Lubana explored mapping incrementally-processed stories into trajectories through conceptual belief spaces.

Story-based search involves inputting stories, instead of keywords, to retrieve resources. Users’ input stories could be processed into points or trajectories in story spaces so that resources could be retrieved, sorted, and presented.

As envisioned, dialogue systems could engage in conversations with users to improve their understandings of users’ input stories in order to enhance story-based search.

Mapping stories to points or trajectories in high-dimensional spaces would enable users to search for and retrieve stories by creating ostensive and extensional sets of stories, or story concepts.

Ostensive means defining a set by means of providing one or more examples. Extensional means defining a set by exhaustively listing all of its elements.

As envisioned, to create sets of stories, users would gather stories together via other search techniques. Users would be able to signal which kind of a set that they were creating, ostensive or extensional.

While users created sets of stories, artificial-intelligence systems could analyze their partially-completed sets to recommend or suggest other stories for users to make use of.

Additionally, artificial-intelligence systems could recommend or suggest existing, defined, known sets of stories by guessing at or completing those sets that users were in the process of creating.

Sets of stories, or story concepts, could also be operated upon and interrelated to one another. Users could create weighted fuzzy intersections, unions, and complements, and could perform conceptual blends. In these ways, more intricate sets of stories, or story concepts, could be created and utilized to retrieve search results.

The technologies under discussion could enable new, advanced forms of case-based reasoning and planning. Artificial-intelligence systems could retrieve resources from story spaces and store revised resources into such spaces.

Pieces of advice could be stored at one or more points in story spaces and could be retrieved when users’ input stories were sufficiently similar. Using such systems, users and artificial-intelligence agents could share pieces of advice with one another.

Users and artificial-intelligence agents could store wisdom materials at one or more points in story spaces and retrieve them using story-based search techniques.

Kinds of wisdom materials include, but are not limited to: allegories, anecdotes, aphorisms, fables, folklore, humor, lyrics, parables, poems, proverbs, quotations, songs, stories, and witticisms.

One could store historical stories as points or trajectories in story spaces and retrieve them using story-based search techniques, providing current events expressed in story form.

High-quality story embeddings will enable new kinds of search, retrieval, browsing, and discovery for library patrons.

Digital humanities and cultural analytics scholarship will advance as a result of high-quality story embeddings.

Corpus narratology is a set of empirical methods for the study of narrative by analyzing vast collections of stories, story corpora.

With vast story corpora, researchers could analyze patterns in and generalizations from collections of stories, for example the hero's journey or monomyth.

By mapping stories to points or trajectories in high-dimensional story spaces, how could multiple aesthetic measures be obtained and utilized on these representations of stories?

How could audience response be more readily predicted by mapping stories to points or trajectories in high-dimensional story spaces?

From story corpora containing certain kinds of stories, artificial-intelligence systems could learn and reason about cultural norms, values, and morals.

Representing stories as points or trajectories in story spaces will enable algorithms for the prediction, continuation, infilling, and completion of stories.

Users and artificial-intelligence agents could store laws, rules, and regulations at points in story spaces and retrieve them using story-based search techniques.

Artificial-intelligence agents’ situational contexts, i.e., their autobiographical episodes, could be of use for retrieving situationally-applicable laws, rules, and regulations.

Courses of action available to artificial-intelligence agents could be efficiently considered with pertinent laws, rules, and regulations loaded into working memory.

Benara, Vinamra, Chandan Singh, John X. Morris, Richard J. Antonello, Ion Stoica, Alexander G. Huth, and Jianfeng Gao. "Crafting interpretable embeddings for language neuroscience by asking LLMs questions." Advances in Neural Information Processing Systems 37 (2024): 124137.

Bigelow, Eric, Raphaël Sarfati, Daniel Wurgaft, Owen Lewis, Thomas McGrath, Jack Merullo, Atticus Geiger, and Ekdeep Singh Lubana. "Stories in space: In-context learning trajectories in conceptual belief space." arXiv preprint arXiv:2605.12412 (2026).

Fang, Le, Tao Zeng, Chaochun Liu, Liefeng Bo, Wen Dong, and Changyou Chen. "Transformer-based conditional variational autoencoder for controllable story generation." arXiv preprint arXiv:2101.00828 (2021).

Fortuin, Vincent, Romann Weber, Sasha Schriber, Diana Wotruba, and Markus Gross. "InspireMe: Learning sequence models for stories." In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 32, no. 1. 2018.

Hatzel, Hans Ole, and Chris Biemann. "Story embeddings - Narrative-focused representations of fictional stories." In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pp. 5931-5943. 2024.

Lee, O-Joun, and Jason J. Jung. "Story embedding: Learning distributed representations of stories based on character networks." Artificial Intelligence 281 (2020): 103235.

Mitka, Jan. "Disentangled representation learning for narrative similarity using synthetic supervision." (2026).

Polceanu, Mihai, Julie Porteous, Alan Lindsay, and Marc Cavazza. "Narrative plan generation with self-supervised learning." In Proceedings of the AAAI Conference on Artificial Intelligence, vol. 35, no. 7, pp. 5984-5992. 2021.