A visitor looking for a 2006 paper on language identification should not have to guess whether it lives under “AI,” “Speech,” “Archives,” or “Projects.” If several labels seem equally likely, the site has passed the sorting problem to the reader. Give each page a clear primary home, then provide other routes where subjects overlap.
That matters on a historical-technical blog. A post about early medical-image analysis might touch on machine learning, computer vision, and clinical applications. Its place in the site structure should reflect the route readers are most likely to try, not every field mentioned in the post.
Start with reader tasks, not a menu design
Before rearranging the header, write down what visitors come to find. One reader may have a technology in mind, such as stereo reconstruction. Another may want an overview of speech research in the mid-2000s; a third may be checking the evidence behind a historical claim. They need different routes: a topic category, a period-focused collection, or source notes within an article.
Take inventory of the pages already published. A spreadsheet with each page’s title, primary subject, content type, period, and current location is enough. Look for near-duplicate categories, pages assigned to competing sections, and important material reachable only through search. This usually tells you more than brainstorming menu labels in isolation.
Group pages by distinctions readers can recognize. For a research archive, “Speech and Language,” “Computer Vision,” and “Document Analysis” are clearer primary categories than an internal label such as “Research Area B.” Think twice before creating a top-level category for one post: it suggests a collection, unless that is a subject you plan to develop.
Give the main menu a limited job
The main menu orients visitors; it does not need to list every page. It should show the major destinations with labels that make sense before anyone clicks. Depending on the archive, those might include Topics, a chronological archive, About, and Search. Conferences might merit a menu item if the collection is substantial and people often seek it out. Otherwise, it can sit within a broader section.
Choose familiar words when they accurately describe the material. “Speech Research” gives a better sense of a section covering recognition, synthesis, and language identification than “Voices.” Keep labels consistent, too. If “Archive” means a chronological listing in the menu, do not use the same word for downloadable documents in a page heading.
Give each level of navigation a separate job:
- Primary navigation: major site sections, visible across the site.
- Section navigation: subdivisions within one area, such as recognition and synthesis under Speech and Language.
- Page-level navigation: a table of contents linking to headings in a long article.
- Contextual links: connections between genuinely related pages in different sections.
Putting every subdivision in the main menu makes it harder to scan without necessarily making the pages easier to find.
Choose one primary home for overlapping topics
Historical research rarely fits into separate boxes. A post on the data limitations of 2000s speech experiments may belong under Speech and Language, though it also has implications for evaluation methods. Place it where its central question belongs, then point to related material in the article or on a relevant category page. For example, a path to [What Small Speech Corpora Could—and Couldn’t—Show in the 2000s](/speech-recognition-under-resourced-languages-2000s-trials/) should lead readers to an examination of what limited datasets could support, not promise a general guide to speech recognition.
Tags can mark ideas that recur across categories, such as “evaluation” or “low-resource languages.” They should not become a second category system by accident. A tag attached to one loosely related page may not need a public archive. Decide what each device does: categories identify a page’s main subject; tags, if used, identify specific cross-cutting concepts; dates locate work in time.
Use labels that predict what lies beyond them
Read a proposed category label on its own, then check the titles of pages assigned to it. If “Methods” holds image processing, conference history, and a biography, it is probably too vague. If “2000s” covers nearly the whole site, a chronological index may serve readers better than treating the decade as a topic.
A short description on a category page can set its boundaries. “Computer Vision,” for instance, might say that it covers tracking, stereo reconstruction, and image analysis. The description should help readers choose where to go next, rather than repeat the label.
Make location and recovery visible
Visitors often land on an individual article, not the homepage. Once there, they need a way to see where it belongs and move outward. A category link near the title, breadcrumbs where the hierarchy is stable, and consistent section headings can help. Breadcrumbs should show a real path through the site; they become confusing if an article appears to have several interchangeable parents.
Search is a useful fallback for someone who knows a term but not your categories. Make it easy to find, and show readable titles, brief descriptions, and dates where they affect interpretation. It cannot make up for unclear labels: a visitor unfamiliar with the terminology may not know what to type.
Check the menu on small screens. Are the labels legible? Can visitors open submenus without precise hovering? A collapsed menu should still show the major sections in a sensible order, with Search accessible rather than buried several levels down.
Test the paths with real questions
A navigation plan can look tidy on paper while its labels send readers the wrong way. Give a few people concrete tasks without suggesting a category: find material on object tracking around 2005, locate an explanation of early speech synthesis, or find posts about how conferences influenced research. Ask where they would click first. A pause or a wrong guess tells you more about a label than a review of the finished menu.
Record both success and effort. Did they find the target? Where did they take a wrong turn, and did they recover through the menu, a category page, or search? Finding a page after several failed clicks shows that the content is accessible, but the intended path may still need work. After changing labels or moving pages, run the tasks again.
Keep checking as the archive grows. When you add a post about stereo reconstruction, choose its primary category and then view that category page as a reader would. If the new title lands among unrelated articles with no useful grouping, fix the section page or its subdivisions before adding another main-menu item.