Map: Terms

Your vocabulary counted twice: in your own corpus and in the published record.

Terms opens with three task views: Explore vocabulary, Sources vs notes, and Google Scholar. Explore the words in your own material, compare reading with writing, or check how terms circulate in the published record. The difference from Themes is the point of the tab: Themes interprets the shape of your library, whereas Terms counts. Your own corpus is available on every plan; Google Scholar analysis needs a paid plan.

Explore vocabulary

Your project vocabulary is counted from what you already have, in the browser. Top terms ranks the vocabulary in the selected part of your project. Terms found near each other finds single terms that repeatedly share a nearby passage. The Pinned button beside Top terms opens the vocabulary you follow as the library grows, and explains how to start if you have no pins yet.

Choose what to count

Start with Count in. All project material combines the three named content streams, Sources counts source titles and abstracts, Notes counts note titles and bodies, and Author keywords counts only the keywords supplied on source records. Keywords never silently stand in for a missing abstract, so switching scopes really does separate what you are reading, what you are writing, and how authors index their own work. The task view, scope, term length and ranking method stay in the URL, which means a reload or shared link restores the same lens. If a scope has nothing in it, it stays empty rather than borrowing results from All.

Choose Sources vs notes at the top of the page for the focused comparison. It ranks vocabulary that leans toward source prose, vocabulary shared by both sides, and vocabulary that leans toward your notes under the same term-size and ranking lens. Source and note occurrence counts are labelled separately. The percentages are within-side shares of positive ranking scores, so a larger pile of source text does not win merely by being larger. Treat the result as a difference in vocabulary, not automatic evidence of novelty or contribution.

Two more controls change how the active scope gets counted. Term length offers Single words, Two-word phrases, and Three-word phrases, which matters because “identity” and “identity formation” are different claims about your vocabulary. Rank by offers Most frequent · Count, Distinctive within documents · TF-IDF, and Unusually frequent · Log-likelihood. Count tells you what occurs most; the statistical methods surface concentration within documents or unexpected frequency against a baseline. Open How these counts work below the table for the exact method. Export CSV takes the current scoped table out.

Read the ranked table

The columns answer different questions:

• # is the term’s position under the current ranking method.
• Occurrences is the exact number of occurrences in the selected scope, even when TF-IDF or log-likelihood is doing the ranking.
• Matching items is the number of matching items. An item can be a source, a note or a source’s author-keyword record.
• In count mode, the bar accompanies Occurrences. Statistical rankings show a separate TF-IDF score or Signed G² score; on narrow screens this appears under the term. Bars are relative to the strongest score magnitude in this view.

The table opens with twelve rows so that the strongest vocabulary stays readable. Show more terms reveals the rest; changing the scope or lens recomputes the order in place.

Inspect the items behind a count

Select the labelled item count, such as 2 items ↓, to list every matching source, note and author-keyword record in the active scope. Each line shows how many exact occurrences came from that item, and its title opens the underlying source or note. The item counts add back up to the row’s total Occurrences, so you can move from an aggregate to the records behind it without reconstructing the search yourself.

This is the useful check before interpreting a prominent term. Ten occurrences concentrated in one note say something different from ten occurrences spread across eight sources, even though the raw total is the same.

Pin important terms and remove noise

Open Manage terms for the judgments that no extractor can make for you. Pin keeps a term visible and follows it in the Pinned panel, including a term you enter that is not in the current results. A pin with no occurrences says Not found in the current scope and gets no invented score. Opening Manage terms reveals row checkboxes and Exclude actions. Exclude removes one result, checkboxes let you exclude several together, and Restore brings an excluded term back.

Pins and exclusions apply only to this project, survive a change of lens or a reload, and follow the project to another device. The management panel stays folded when you are only reading results, and Pinned remains available even before you have anything to follow. The ranked table, cloud, co-occurrence pairs, pinned counts and export all honor the same project curation and active scope.

Read terms found near each other

Choose View pairs to open this section. It looks for single words that share overlapping 50-token windows in the active scope. Windows advance by 25 tokens, pairs seen in only one window are omitted, and the remaining pairs are ordered by pointwise mutual information (PMI). PMI asks whether two terms meet more often than their separate frequencies would lead you to expect; it does not say that one causes the other or that their relationship is important.

Each card leads with Observed together in n windows, because support is the defensible fact. Position says where the pair sits among the other reported pairs in this corpus, while labels such as Weak support, Repeated support and Notable lead are bounded reading cues rather than claims about the field. Term length and ranking controls affect the term table; pairs always use single words and PMI. The first six pairs stay visible; Show more pairs opens the remainder.

Select View evidence to see the cautious interpretation and the individual sources, notes or author-keyword records behind the pair. Each item reports its matching-window count and opens the underlying record. Those counts add back up to the pair’s observed support. Read the passages before deciding what the proximity means for your argument, especially when several overlapping windows come from one item.

Open Method and limits below the cards for the exact PMI formula, windowing and exclusion rules.

Use the secondary views when you need them

View as term cloud stays folded beneath the table. It shows up to forty of the same ranked terms in alphabetical order, so position never implies importance. Type size follows the active ranking score (its absolute magnitude for signed G²) on a square-root scale from 14 to 26 pixels, keeping the long tail legible; when the scores sit within 10% of one another, every term uses one size rather than stretching a small difference into a false hierarchy. The exact score is available on each term.

When the active view has enough evidence, Explain these results requests one optional reading of that exact scope or comparison, term size, ranking, visible top terms, and well-supported pairs. Nothing is generated when the page loads. Repeating the same view reuses its cached reading; changing the corpus or lens creates a different request. Thin views show no action, and when AI is turned off an existing reading stays visible as previously generated, read-only text.

Read it as a check on yourself. A term central to your questions that ranks nowhere means you have been writing around the thing rather than about it, and a term at the top you never chose is worth knowing before a reader points it out.

Google Scholar

Choose Google Scholar at the top of the page to see how the field uses those words. This view queries the scholarly record.

Running an analysis

Pick two to seven terms and run it, with suggestions coming from your project keywords, set on Overview, and the tab telling you when a suggestion came from there.

An analysis usually takes a few minutes, with elapsed time and a Cancel action shown while it runs. You can switch to Explore vocabulary without cancelling the request. The elapsed timer does not claim to measure backend progress.

Choosing the terms is the skill, since the comparisons that pay off are between things that might be alternatives: your construct against its nearest rival name, your framing against the field’s standard framing, or the term your supervisor uses against the term the literature uses. Compare seven unrelated terms and you get seven unrelated numbers.

Frequencies

How present is each term in the field? Raw Scholar counts for each term on its own. The analysed terms and date stay above the results. Open View counts and percentages for the detailed table; percentages divide each count by the sum of the selected term counts, not by a count of unique publications.

This measures published attention, and it is useful in two opposite directions. A term with very few results is either genuinely new or nobody’s word for the thing, and finding out which is worth an hour of your life, whereas a term with millions is established, so building an argument on it means engaging with a great deal of prior work.

Do not read the counts as importance, since they measure how much has been written, which tracks age and field size at least as closely as significance.

Overlap

Which terms travel together? Where your terms co-occur in the record. Switch between Overlap diagram, Intersection plot, and Count tables to inspect the same analysis. Chart expansion, Scholar links, and exports remain available in their respective views.

This half is the more interesting one. A large overlap marks an established conversation, meaning these two ideas already get discussed together and there is a literature you have to engage with. A thin overlap can mark an opening, since two well-populated terms that rarely appear together may be an unexploited connection, which is often exactly what a dissertation is.

Keep two interpretation notes in mind. The bars use a square-root scale so small counts stay visible, with the exact figure printed above each. The counts are Scholar co-occurrence totals rather than exclusive intersections, so they tell you both terms appear without telling you the work is about their combination.

Combinations with no co-occurrence get omitted and the tab reports how many, while Open in Google Scholar takes you to the underlying query for any pair, which is how you check a number before citing it.

The reading

A short advisor reading sits below the charts and frames what the numbers show, reading figures that have already been computed, and you can copy it.

History

Open Previous analyses at the top of the Google Scholar view to reopen or delete a saved analysis.

Use it deliberately, because running the same term set at the start of your project and again two years later makes the change a real finding about your field’s attention. It also records what you checked, which matters when someone asks whether you considered an alternative framing.

Its honest limits

Scholar counts are a rough instrument, including preprints, theses and citations of citations, shifting between runs as the index changes, and unable to tell a term used centrally from one mentioned in passing.

They are still the right tool for the question is this word used, and used alongside that one, and unlike everything else in this Map anyone reading your work can check them. Quote them as what they are, meaning Google Scholar result counts on a stated date.

Next: Connections.