Build a vocabulary list from Chinese reading

A word that appears repeatedly in a Chinese passage is a useful candidate for review, but a frequency list needs interpretation. Chinese writing does not normally put spaces between words. Before counting, a tool must decide where each word starts and ends; a mistaken boundary can put the wrong entry near the top of your list.

Start with a short passage you actually want to read. Check its segmentation, choose which words belong in the statistics, then attach a sentence and a meaning to the vocabulary you keep. The worked example below follows one original paragraph from raw text to a small study list, with numbers you can reproduce.

Choose a passage with a purpose

A chapter, short article or class reading gives the list a clear scope. Copy the body text and preserve a clean original. Remove repeated navigation, advertisements, page headers and duplicated OCR lines before analysis: their repeated words otherwise compete with the vocabulary in the reading itself. Keep the title if you want it counted, and record that decision when comparing passages.

Use this original practice paragraph. It describes a library visit and a reading routine; it is not an extract from a published book or a report about a real person:

周末，我去图书馆看书。图书馆有中文书，也有英文书。我借了两本中文书，准备下周还书。看书以后，我在笔记本上写下新词，再用新词写句子。

The narrator goes to a library at the weekend, borrows two Chinese-language books, plans to return them next week and records new words in a notebook. This context matters later: it explains both what kind of books are meant and what the verb in 还书 means.

Paste the paragraph into the Chinese word frequency analyzer, leaving custom and excluded words empty for the first run. Read the segmented text as well as the frequency table. The table compresses repeated occurrences into one row; the segmented view keeps their original order and punctuation.

Check boundaries before counting

In the current tool, the first run returns 33 counted tokens and 25 distinct entries. It splits 中文书 into 中 and 文书. Both are possible character sequences, but that split misses the intended meaning of Chinese-language books in this paragraph. An apparently plausible dictionary entry is not enough to establish the right reading.

Add 中文书 to the custom words field, one term on its own line, and analyze again. Both occurrences now stay together. The result has 31 counted tokens and 24 distinct entries. Here we deliberately treat Chinese-language book as one study item; this is a choice for this reading exercise, not a claim that every Chinese compound has one universally correct segmentation.

Other study goals may call for keeping 中文 and 书 separately, or discussing 两 and the book classifier 本. A tokenizer's output is a starting point for that judgment. Do not fill the custom field with entire sentences simply to make a list look shorter. Add a name or expression only when its occurrences really should be grouped for your task.

Custom terms affect matching occurrences throughout the input. Check the surrounding sentences after adding a short or ambiguous term. For comparisons across chapters, use the same custom terms and retain a copy of them with your results. Changing boundaries halfway through a project changes both the vocabulary entries and the totals being compared.

Read counts and shares together

A token is one counted occurrence; a distinct entry is one different token string. With 中文书 kept together, 图书馆, 看书, 中文书 and 新词 each occur twice. 我 occurs three times. Punctuation and spaces appear in the segmented view but do not contribute to the frequency total.

The share is the word's count divided by the total number of retained tokens. Before exclusions, the calculation for 图书馆 is 2 ÷ 31 × 100, approximately 6.45%. After exclusions, it is 2 ÷ 22 × 100, approximately 9.09%. The word has not appeared an extra time; only the denominator has changed.

For a vocabulary-focused second pass, enter the six excluded words shown in the table on separate lines. They remove nine occurrences in total. Exclusions match complete token strings, so excluding 书 will not remove the custom token 中文书. The segmented text remains available even when its words are excluded from the statistics.

These six words are useful in Chinese and should not automatically be discarded from every lesson. Keep them when studying grammar or sentence patterns. Likewise, excluding numbers can help with some passages but hide useful information in a timetable or shopping text. Decide what you want to learn before choosing filters, and label saved results accordingly.

Keep meaning in its sentence

Choose words using frequency, your existing knowledge and the reading task together. A repeated word you already know needs little attention, while a word appearing once may explain the entire paragraph. In this example, 借 occurs once but tells you that the books were borrowed. A top-ten cutoff alone would not reliably produce the best lesson.

For each chosen item, keep the exact sentence or a short complete clause, its contextual meaning and a checked pronunciation. In 准备下周还书, 还 means return and is read huán. Learning it only as hái, meaning still or also in other contexts, would lead you away from the sentence. Similarly, 笔记本 here is a paper notebook used for writing, rather than a notebook computer.

A useful record can contain the word, pronunciation, meaning in this passage, original sentence and passage title. Mark uncertainty for later checking instead of treating an automatically generated reading as final. For a name, technical term or ambiguous pronunciation, consult the full sentence and an appropriate dictionary before making a card.

Export a list you can reuse

Download the CSV after you have settled the custom terms and exclusions. It contains the retained word, count and percentage, including all result pages. The screen rounds percentages to two decimal places; the CSV keeps four. Small rounding differences do not mean the counts changed, and displayed percentages need not add to exactly 100%.

The CSV is a statistical list, not a ready-made set of definitions. Keep it alongside the original passage and a note of the settings. A practical filename can include the passage identifier and whether filters were applied. When comparing texts of different lengths, examine normalized shares as well as counts, using consistent segmentation and filtering.

For review, put selected words into the Chinese vocabulary flashcards tool. Prepare the word, pinyin and your contextual meaning; inspect and edit the resulting cards before practicing or exporting them. Do not paste the frequency CSV as though its count column were a pronunciation column. Retain the source sentence in your study notes or the card's meaning field.

Return to the original reading

Reread the paragraph after reviewing a manageable group of words. Check whether you can follow the library visit without consulting the list, especially the borrowing, returning and note-taking actions. This gives the vocabulary exercise a concrete outcome beyond collecting entries.

For a longer reading, work in consistent sections and compare the results only after checking their scope and settings. Simplified and traditional spellings, names and different written forms may remain separate entries; similar meanings do not make them identical tokens. Frequency in your chosen passage describes that passage, not a universal Chinese frequency ranking, a proficiency level or a percentage of meaning understood.