What is a frequency dictionary, and how does it tell me whether a word is worth learning?

The scarcest resource in learning a language is time. One anime series or light novel holds hundreds or thousands of unknown words; turning every one into a card buries you in reviews, and making none means you never remember them. A frequency dictionary answers "which one first": it tells you where a word ranks across a large body of real text — the higher it ranks, the more common it is and the more it's worth learning first.

What a frequency dictionary is ​

A frequency dictionary isn't a definition dictionary. It explains nothing; it just gives each word a rank (or an occurrence count): take a large corpus — novels, anime subtitles, news, web articles, Wikipedia and so on — count how often each word appears and sort from most to least. する, ある and こと sit at the very top; obscure literary words rank hundreds of thousands down.

For a learner, ranks are in some ways more useful than definitions:

• Words near the top are words you'll keep meeting. The first few thousand words cover most everyday text; get them solid and the density of unknown words drops fast.
• Words far down can wait. A word ranked past 100,000 may not show up again for a year; making a card for it just adds review load.
• Different corpora give different frequencies. Novel, anime and news corpora each lean their own way: if you mostly watch anime, look at anime / subtitle frequencies; if you read light novels, look at novel frequencies. Comparing two frequency dictionaries from different corpora is more accurate than relying on one.

Fushi uses Yomitan-format frequency dictionaries (a zip containing freq records in term_meta_bank). The Yomitan community maintains plenty of free frequency lists built from novels, anime, Wikipedia, web text and other corpora. Import them like any other dictionary — see How do I import dictionaries — and they appear in the "Frequency" group in Settings → Lookup → Dictionaries. The recommended pack already includes frequency dictionaries.

Reading frequency in the lookup popup ​

With a frequency dictionary installed, every entry in the popup gets a frequency row:

• By default it's a single "avg" tag followed by a number. That's "Aggregate word frequencies": Fushi combines the ranks from your frequency dictionaries into one number using the harmonic mean. The harmonic mean leans towards the smaller numbers — if one dictionary says the word is very common, the combined value isn't dragged far away by a large number from another.
• Tap the tag to expand it into each frequency dictionary's own value (each prefixed with the dictionary's name); tap again to collapse.
• If you don't want the combined value, turn off "Aggregate word frequencies" (on by default) under Settings → Lookup → Entry content and the popup lists each dictionary's value directly.

Some frequency dictionaries show more than a bare number — a suffix, or special formatting such as reading-specific marks. For the combined value Fushi only uses the leading digits.

How to read the number ​

The number is a rank: smaller means more common. A rough intuition (it varies with the corpus and dictionary):

These are rules of thumb, not rules. A word ranked 40,000 that shows up in every episode of the show you're watching is a high-frequency word for you — a frequency list describes the average case, and your own input has the final say.

Using frequency to decide whether to mine a word ​

A practical routine:

1. Look words up as usual while watching and reading — not every word you look up becomes a card.
2. Glance at the avg number in the popup: if it ranks high and you don't know it, tap the plus and mine it.
3. Let the low-ranked ones go for now. If you keep running into one, mine it then — repeating in your input is itself proof it matters to you.
4. To keep the number of new cards per day under control, see How many cards a day should I make?

Sentence cards stick better than bare word cards because you remember "I saw it in this scene"; Fushi's plus brings the original sentence and audio along by default — see How do I make my first card?

Frequency on Anki cards: two placeholders ​

When you mine with the plus, frequency produces two field values:

With the Lapis deck Fushi creates in one tap, the Frequency field is mapped to {frequencies} and FreqSort to {frequency-harmonic-rank}. FreqSort exists for sorting: a plain number that the "Reorder new cards by frequency" feature below can read directly, without looking anything up again. With your own note type, map fields to these placeholders in Settings → Card creation → Field mappings.

Reordering new Anki cards by frequency: common words first ​

After a mining session Anki shows new cards in the order they were added — a rare word you mined from anime today may come before a common one you mined from a book tomorrow. Fushi can re-queue them by frequency.

Look for "Reorder new cards by frequency" under Settings → Card creation:

1. Pick a deck (subdecks included; cards in filtered decks are skipped).
2. Choose the "Frequency source":
   • Frequency dictionaries: look up each card's word in your frequency dictionaries. You can tick which ones to use; hidden frequency dictionaries are not loaded, so unhide them in dictionary management first. Nothing selected means all loaded dictionaries.
   • Note field: read the field mapped to {frequency-harmonic-rank} (FreqSort in Lapis) directly — no dictionary lookup, the fastest option.
3. With several dictionaries, pick how to "Combine multiple dictionaries": "Harmonic mean" or "Lowest rank" (the best rank among them).
4. To flip it so rare words come first, tick "Rare words first".
5. Hit "Preview" first: it tells you how many new cards there are, how many have a frequency, how many don't (those stay at the end) and how many positions will change. Then hit "Apply".

A few rules:

• Only new cards are touched; review cards never are. A card that stops being new while the reorder runs is skipped.
• Mistakes can be undone: "Undo last reorder" restores the positions saved before the last reorder; cards you've studied since stay where they are.
• The deck's new card order must be sequential: if the deck options gather / sort new cards randomly, the written positions are ignored. Set the new card gather / sort order to sequential in Anki's deck options.
• Turn on "Auto-reposition after mining" and the deck's new cards are reordered by frequency about 30 seconds after the last card is mined, so you don't have to run it by hand.
• Only available with AnkiConnect, i.e. on a computer connected to desktop Anki. Cards mined on a phone with AnkiDroid / AnkiMobile can be reordered on the computer after they sync.

Recommended Anki settings (FSRS, how to use the buttons) are in What should I set up in Anki after connecting?

Common problems ​

• No frequency in the popup: no frequency dictionary is installed, or it's hidden; check the "Frequency" group in dictionary management.
• Some words have no frequency: the list doesn't include the word, or it lists a different spelling / reading than the one you found (common with kana spellings and variant kanji). Cards without frequency go to the end when reordering.
• The avg number is far from one dictionary's value: the harmonic mean leans towards the smaller (more common) rank; tap the tag to see each dictionary, or turn off "Aggregate word frequencies".
• The order in Anki didn't change after reordering: check whether the deck options use random new card order.
• "Only available with AnkiConnect.": reordering can't run on a phone; do it on the computer.