This is article is the first installment in the Workflow Patterns for AI-Assisted Development series. To get notified when a new article is published, subscribe for free via email or RSS.
Domain Language Mining
Discover jargon from an unfamiliar field using LLMs to help you write better software.
. The lack of vocabulary inhibits the knowledge acquisition, which is necessary for “ideas that lead to deep model insights” (ibid.)
When a domain expert uses your software, it will not align with their expectations. They may struggle to use it. They may judge it as amateurish. They may decide your software is not for them.
Additionally, you may model less effectively if you’re using the wrong terms. This can lead to developing a wrong solution.
▼ Therefore
Solution
Use LLMs to help you discover jargon the people who need your software use.
Actions
Choose one piece of jargon, domain language, or technical knowledge you are trying to find. This might be a concept you don’t know the name of, a tool or process you believe might exist, or a best practice you believe should exist but can’t find.
Craft a narrow prompt to try to extract that knowledge from the latent space encoded in the LLM.
Use few-shot prompting to increase your odds of success. Choose three or more problem domains you are familiar with. Create a corollary pair of questions and answers for each. Ensure your answers are not too similar to each other. Include those question-answer pairs, and then ask your question.
Adding examples to the prompt is known as few-shot prompting. Examples can be really useful when you explain things to people, and they’re even more useful when explaining things to LLMs. That’s because LLMs are great at picking up patterns in the prompt and continuing them in the completion. Therefore, you can use examples to show not only how exactly to interpret the question but also how exactly you want the LLM to give the answer.
Chapter 5: Prompt Content, Prompt Engineering for LLMs, by John Berryman and Albert Ziegler
Give the same prompt to least three different models. Varied responses means more opportunities to find what you seek. Overlap between responses indicates relevance.
Verify the candidate terms by searching dictionaries, especially domain-specific dictionaries. Search for the term in relevant academic writing, books, blogs, and forums. Continue verifying terms independently until you can qualify one as the likeliest to be correct.
Example
I am new to blogging, and I wanted a way to link multiple blog posts other than tags and categories. Specifically, I wanted to relate all of the posts that were in a series. I knew that a WordPress plugin would be useful for this, but I didn’t know what a blogger would usually call this series-oriented relationship.
I crafted the following prompt using few-shot learning. I used examples from problem domains I am familiar with. I included examples ending in -ship, -ment, and -tion to reduce bias.
Semantic nouns for relationship classes include:
- Author, Book => Authorship
- Club, Person => Membership
- Student, Course => Enrollment
- User, Role => Assignment
- Person, Event => Participation
- Researcher, Article => Contribution
Give me options for a semantic noun for a relationship class between Blog Post and Series.
- Blog Post, Series =>
I gave the same prompt to four different models. The responses gave me a broad series of options. It became clear that my lack of knowledge was hiding an important fact: the name of the relationship can change depending on why these blog posts are in a series.
Anthropic Claude Sonnet 4.5
Here are semantic noun options for the relationship between Blog Post and Series:
Top recommendations:
- Inclusion – emphasizes the post being included in the series
- Installment – suggests the post is one part of a serialized whole
- Entry – indicates the post is an entry within the series
Other strong options:
- Membership – similar to Club/Person (the post is a member of the series)
- Placement – focuses on the post’s position within the series
- Serialization – emphasizes the serial nature of the relationship
- Episode – works well if the series is episodic in nature
Context-dependent alternatives:
- Publication – if emphasizing the act of publishing within a series
- Arrangement – if order/sequence is important
- Composition – if viewing the series as composed of posts
My personal favorite is Inclusion for its clarity and semantic precision, though Installment works beautifully if you want to emphasize the sequential, serialized nature of blog posts within a series.
OpenAI ChatGPT (free & logged-out)
Here are good semantic-noun options for a relationship class between Blog Post and Series:
- Blog Post, Series ⇒ Installment (most common when posts are sequential parts of a series)
- Blog Post, Series ⇒ Episode (if the series is narrative or thematic)
- Blog Post, Series ⇒ Entry (neutral and widely applicable)
- Blog Post, Series ⇒ Inclusion (emphasizes membership in a collection)
- Blog Post, Series ⇒ Placement (emphasizes organization within the series)
- Blog Post, Series ⇒ Serialization (if posts collectively form a serialized work)
Best general-purpose choice: Installment or Entry, depending on whether sequence matters.
OpenAI ChatGPT 5.1 Thinking
Nice little modeling problem 😊
Here are some semantic noun options for the relationship between Blog Post and Series:
- Serialization — the act of publishing something as part of a series.
- Inclusion — a post being included in a series.
- Placement — a post being placed within a specific series.
- Affiliation — a post being affiliated/associated with a series.
- Aggregation — the series as an aggregation of posts.
- Composition — the series being composed of posts.
- SeriesMembership — mirrors Membership; a post “belongs” to a series.
- Enrollment (reused) — if you like the metaphor of a post being “enrolled” into a series.
If you want something that feels closest in style to Authorship/Membership/Enrollment, I’d probably shortlist:
- SeriesMembership
- Serialization
- Inclusion Meta Llama 4 Maverick
To determine a suitable semantic noun for the relationship class between “Blog Post” and “Series”, we can consider the nature of their relationship. The given examples illustrate various relationships, such as authorship, membership, enrollment, assignment, participation, and contribution, which all describe how two entities are connected.
For “Blog Post” and “Series”, the relationship is about a blog post being part of a series. Here are some potential semantic nouns for this relationship:
- Inclusion: This suggests that a blog post is included in a series.
- Grouping: This implies that blog posts are grouped into a series.
- Membership: Similar to a person being a member of a club, a blog post can be considered a member of a series.
- Ordering: If the series implies an order (e.g., a sequential series of posts), this could be a suitable noun.
- Classification: This suggests that a blog post is classified under or associated with a particular series.
- Association: A general term that indicates a blog post is associated with a series.
- Publication: This could work if the context emphasizes the act of publishing a blog post as part of a series.
However, the most straightforward and semantically relevant noun for this relationship could be Serialization.
Thus, some options are:
- Blog Post, Series => Serialization
- Blog Post, Series => Inclusion
- Blog Post, Series => Membership
Of these, “Serialization” directly conveys the idea of blog posts being part of a series, making it a strong candidate.
Google Gemini Thinking with 3 Pro
Here are several semantic noun options for the relationship class between a Blog Post and a Series, categorized by the nature of the relationship:
The Best Fit: Installment
This is the strongest semantic match for blog content. Just as a “Member” is the entity in a “Membership,” a post in a series is often called an “Installment.” It implies a periodic release schedule typical of blogging.
- Context: “We just published the third installment of our ‘Intro to React’ series.”
The “Ordering” Focus: Sequence
If the defining characteristic of the relationship is the specific numerical order (e.g., Part 1, Part 2, Part 3), “Sequence” or “Ordering” describes the link effectively.
- Context: “The sequence of this post within the series is critical for understanding the tutorial.”
The “Narrative” Focus: Episode
If the blog series tells a story or follows a journey (e.g., “Our Startup Journey,” “Travel Diaries”), this noun borrows the semantic weight of TV or podcasts to imply a narrative arc.
- Context: “This episode of the founder’s diary covers our funding round.”
The “Book” Metaphor: Chapter
If the series is educational, long-form, or intended to be read like a book, “Chapter” elevates the relationship to suggest a distinct, substantive section of a larger whole.
- Context: “Read the final chapter of our Deep Learning Guide.”
The “Curatorial” Focus: Compilation
If the series is a loose collection of related posts without a strict chronological necessity (e.g., “Best of 2023,” “Editor’s Picks”), this implies the act of gathering or curating.
- Context: “This post’s compilation into the ‘Summer Reading’ series increased its traffic.”
Summary Table for Data Modeling
| Link Entity Name | Implied Nuance |
| Installment | Periodic, sequential, standard for serial content. |
| Serialization | The act/process of turning content into a series. |
| Chapter | Structured, educational, part of a cohesive whole. |
| Entry | Neutral, purely structural (e.g., “Series Entry”). |
| Inclusion | Weak relationship; just “tagging” a post as part of a set. |
Because of the specific kinds of series I wanted to create, and because of its frequency across different responses, I chose installment as my candidate. I searched blog series "installment" and article series "installment" on Kagi, Marginalia, Google Scholar, and the O’Reilly Learning Platform. The resources I found verified that bloggers and journalists use this term for the same purpose. I looked through Technical Blogging, 2nd Edition by Antonio Cangiano and found its use in Chapter 4. I was satisfied with this term, so I did not verify other candidates.
▼ As a result
Consequently
Instead of using a substitute word, inventing a new term, or using a concept from the language of a different problem domain, you will know the jargon used by the people who need your software. This will help you build a better model of the problem, and your solution will be more likely to be useful, accepted, and appreciated.
This was the first installment in the Workflow Patterns for AI-Assisted Development series. To get notified when a new article is published, subscribe for free via email or RSS.
Get notified when I publish new articles
Subscribe for free emails when I publish new articles (not quick posts). Privacy policy applies.
The image in this article was generated using AI tools (Nano Banana Pro) based on creative prompts. The written word in this article is proudly human-made, except for the five designated LLM outputs in the Example.