Series ID: EDKS-ADV-VOC-ZH-0009 · Advanced English Vocabulary (Chinese Edition) · Lesson No.009 · Batch 03 · C1 → C2 · 简体中文辅助

How to Learn Advanced English Vocabulary (Chinese Edition) | Lesson No.009 | Use Corpus Evidence to Check Natural Word Choice | 第009课：用语料库证据验证自然选词

At advanced level, “This sentence is possible” is no longer the final question. You also need to ask: “Is this how English is normally used?”
到了高级阶段，“这句话理论上可以吗？”已经不够。你还要问：“真实英语通常这样用吗？”

You write powerful evidence. It is understandable. You write strong evidence. That is also understandable. Which is more conventional? You write discuss about the issue. Every individual word looks reasonable. Is the preposition actually needed? You want to choose between significant, substantial and considerable. A dictionary gives definitions. How can you see what nouns each adjective actually prefers?

这类问题到了 C1–C2 会越来越多。词典告诉你“这个词可以是什么意思”，语法告诉你“句子有没有违反规则”，但是高级学习还需要第三层证据：真实语言使用者实际上反复怎么说。

A language corpus is a large collection of written or spoken language stored so that patterns of usage can be examined. Corpus tools let learners look at authentic examples, recurring collocations, phrase patterns and distributions. They do not replace judgement. They give you evidence for judgement.

语料库不是“万能答案机”。它更像语言实验室：你提出一个具体问题，然后观察大量真实例子，看英语的高概率路线是什么。

This lesson owns the advanced Mandarin-supported learner workflow for using corpus evidence to verify lexical choices. It does not replace eduKateSengkang’s canonical collocation mechanism page How Collocation Turns Vocabulary Knowledge Into Natural, Precise English, nor the synonym, morphology or context-clue owners. Here the focus is evidence-based checking.

---

1. What is a corpus? | 什么是语料库？

Cambridge Dictionary defines a corpus in the language-learning sense as a collection of written or spoken material stored on a computer and used to study how language is used. The important idea is not merely size. It is systematic evidence from many real examples.

Imagine that you are unsure whether English normally says strong evidence or powerful evidence. One sentence from one website proves almost nothing. A corpus lets you inspect many examples and recurring patterns.

语料库的核心价值是：把“我觉得”变成“我可以查看真实使用证据”。

2. Corpus evidence answers a different question from grammar | 语法和语料库回答不同问题

Tool	Main question
Grammar	Is this structure permitted?
Dictionary	What does this word mean and how can it behave?
Corpus	How is this word or phrase actually distributed and combined in real language?
Your context	Which available choice best fits this meaning, audience and purpose?

Do not ask one tool to perform every job.

词典告诉你可能性；语料告诉你概率；语境决定你最终选什么。

3. “Possible” is not the same as “probable” | “可以”不等于“常见”

A sentence may be understandable and grammatical but still use an unusual lexical route.

• heavy rain — highly conventional
• strong rain — understandable, but not the normal collocation
• make progress — conventional
• do progress — non-standard in ordinary use
• strong evidence — highly conventional
• powerful evidence — possible in some contexts, but a different probability profile

Advanced naturalness depends partly on learning these probability patterns.

4. A corpus is not the whole language | 语料库也不是“整个英语”

Every corpus has a composition. It may contain web pages, newspapers, academic papers, books, conversation, subtitles or specialised texts. Therefore every result reflects the data that were collected.

如果一个语料库主要是新闻，你看到的是新闻英语的概率。如果主要是网络文本，结果会带网络语言的结构。如果是学术语料，很多表达在日常对话里可能很少。

Corpus evidence is powerful only when you ask, “Evidence from what kind of language?”

5. The first rule: ask a narrow question | 第一条规则：问题要具体

Weak corpus question:

How do people use “significant”?

Stronger questions:

• Which nouns commonly follow significant?
• Is significant improvement common?
• How does significant differ from substantial before increase?
• Is significant more common in academic than conversational contexts?
• Which adverbs modify significant?

A corpus works best when your question is small enough to test.

6. Question type 1 — Is this collocation natural? | 问题一：这个搭配自然吗？

Candidate: strong evidence.

Check how often and in what contexts the phrase occurs. Then compare alternatives such as:

• compelling evidence
• substantial evidence
• conclusive evidence
• powerful evidence

The corpus will not tell you that one adjective is universally “best”. It reveals typical combinations and contexts.

7. Question type 2 — Which verb normally goes with this noun? | 问题二：这个名词通常搭什么动词？

Target noun: conclusion.

• reach a conclusion
• draw a conclusion
• support a conclusion
• challenge a conclusion
• justify a conclusion
• lead to a conclusion

This is more useful than learning conclusion = 结论 because production usually needs the verb–noun package.

8. Question type 3 — Which adjective normally goes with this noun? | 问题三：这个名词通常搭什么形容词？

Target noun: evidence.

• strong evidence
• compelling evidence
• limited evidence
• empirical evidence
• anecdotal evidence
• conclusive evidence
• circumstantial evidence

Each adjective classifies evidence differently. Corpus examples show both frequency and discourse environment.

9. Question type 4 — Which preposition belongs here? | 问题四：这里通常用哪个介词？

• responsible for
• relevant to
• consistent with
• dependent on
• contribute to
• distinguish A from B

For advanced learners, repeated preposition errors often reflect lexical storage rather than a general failure to understand prepositions. Corpus evidence lets you inspect the whole phrase.

10. Question type 5 — Is this phrase missing a preposition? | 问题五：是不是根本不需要这个介词？

Chinese-influenced learners may produce:

• discuss about the issue
• enter into the room
• emphasise on the point

Corpus and learner-dictionary evidence can show that standard English commonly uses discuss the issue, enter the room, and emphasise the point. The error may come from translating a Chinese relational structure or from analogy with another English verb.

真正要记的是完整动词模式。

11. Question type 6 — Which complement pattern follows the word? | 问题六：这个词后面接什么结构？

• avoid doing
• consider doing
• decide to do
• prevent somebody from doing
• enable somebody to do
• insist on doing / insist that…

A corpus shows repeated patterns in context. It can also reveal whether multiple patterns exist and whether they differ in meaning or register.

12. Question type 7 — Which phrase is more conventional? | 问题七：两个都懂，哪个更常规？

Compare:

• make a decision / do a decision
• raise awareness / increase awareness
• strong evidence / powerful evidence
• heavy rain / strong rain
• reach an agreement / arrive an agreement

Some comparisons have a clear conventional winner; others depend on context. Corpus examples help you see the difference rather than treating every pair as a rule.

13. Question type 8 — What usually comes before or after this phrase? | 问题八：这个表达周围通常出现什么？

Target frame: in light of.

• in light of new evidence
• in light of recent developments
• in light of these findings
• in light of changing circumstances

Corpus evidence teaches the variable slot. You learn not one frozen phrase but a productive frame.

14. Question type 9 — Which adverb fits this adjective? | 问题九：哪个副词最自然？

• highly likely
• deeply concerned
• strongly opposed
• widely accepted
• closely related
• fundamentally different
• roughly equivalent

Learners often overuse very. Corpus exploration reveals more specific adverb–adjective partnerships.

15. Question type 10 — Is the word mostly written or spoken? | 问题十：这个词主要出现在写作还是口语？

Register is partly observable through corpus composition. Compare formal items with everyday alternatives:

• commence / start
• purchase / buy
• ascertain / find out
• approximately / about
• facilitate / help

A corpus divided by genre or register can show where each choice is concentrated.

高级语域判断最好有真实分布证据，不要只靠“这个词比较长，所以比较正式”。

16. Question type 11 — Is this academic, professional or general English? | 问题十一：这是学术、职场还是通用表达？

Words like significant, framework, criterion, implication, constitute occur widely, but their density and phrase behaviour can change across domains.

Do not label a word “academic” and stop. Ask:

• Does it occur frequently in academic texts?
• Which academic phrases contain it?
• Does it also occur in journalism or professional writing?
• Would it sound heavy in conversation?

17. Question type 12 — Which near-synonym fits this noun? | 问题十二：这个名词更喜欢哪个近义词？

Compare significant, substantial, considerable.

Search their noun partners:

• significant difference / effect / role / increase
• substantial amount / evidence / increase / investment
• considerable difficulty / variation / attention / amount

The distributions overlap, but the centres differ. Corpus evidence turns synonym learning into pattern comparison.

18. Question type 13 — Which near-synonym fits this verb frame? | 问题十三：哪个近义词适合这个动词结构？

Compare suggest, indicate, demonstrate, establish around evidence. Look at subjects and complements:

• the evidence suggests that…
• the findings indicate that…
• the experiment demonstrates that…
• the analysis establishes that…

Then inspect whether the examples reflect different evidence strength or merely stylistic variation.

19. Question type 14 — Does the word carry positive or negative company? | 问题十四：这个词通常和什么评价一起出现？

Corpus neighbourhood can reveal semantic prosody: a word may repeatedly occur in positive, negative or problem-related environments.

For learners, keep the idea practical. Ask whether a candidate tends to appear around:

• risk and failure
• benefit and success
• criticism
• formal evaluation
• neutral description

This can support connotation work, but always verify with dictionary labels and full contexts.

20. Question type 15 — Is this phrase current? | 问题十五：这个表达现在还常用吗？

Language changes. A modern corpus can help distinguish current usage from older literary or historical patterns.

However, “not frequent now” does not mean “wrong”. Historical writing, legal language and literature preserve older forms.

高级学习者要把“当前常用”“仍然正确”“特定文体使用”分开。

21. Question type 16 — Is this regional? | 问题十六：这是英式、美式还是区域偏好？

Some variants differ by region:

• at the weekend / on the weekend
• different from / different than / different to
• programme / program in different contexts

Corpus metadata can help show regional patterns. Do not convert variation into false rules.

22. Start with examples, not counts | 初学语料库时，先看例句，不要先迷上数字

A raw frequency count can be misleading. The same phrase may occur repeatedly in duplicated pages, one specialised field or one fixed expression. Before interpreting numbers, inspect the concordance lines—the examples around the word.

第一阶段先问：

• Are the examples really expressing my intended meaning?
• Are they from appropriate genres?
• Does the phrase repeat across different sources?
• Is the word being used in the same sense?

Numbers only become useful after semantic filtering.

23. One word may have several senses | 一个词多个词义，不能把所有结果混在一起

Search issue and you may see:

• policy issue
• technical issue
• latest issue of a magazine
• issue a statement
• issue a warning

A global count for issue cannot answer a sense-specific question. Filter conceptually.

高级语料库使用首先是词义工作，然后才是统计工作。

24. Search phrases, not only headwords | 不要只搜单词，要搜完整词块

If your question concerns natural production, search the phrase you might actually use:

• take into account
• raise concerns about
• be consistent with
• play a significant role in
• there is little evidence to suggest

The phrase reveals grammar, collocation and discourse environment at the same time.

25. Search variable frames | 学会搜“有空位的框架”

Corpus tools can help reveal patterns such as:

• significant + NOUN
• VERB + evidence
• responsible + PREPOSITION
• contribute + PREPOSITION
• ADVERB + likely

Tools such as SKELL simplify this through Word Sketch-style views that automatically organise common grammatical relations and collocations for learners.

26. SKELL is a good learner entry point | SKELL 很适合学习者入门

SKELL (Sketch Engine for Language Learning) is designed specifically for students and teachers. It provides example sentences, collocations through Word Sketch, and similar words using large corpora. Its simplified interface reduces the technical burden of corpus research.

A useful beginner workflow:

1. Search the target word.
2. Read examples.
3. Open Word Sketch.
4. Inspect common noun/verb/adjective partners.
5. Click a collocation and read examples.
6. Compare a near-synonym separately.

不要一上来就研究复杂统计。先学会“问问题—看例句—找模式”。

27. Sketch Engine goes deeper | Sketch Engine 可以继续往深处走

Sketch Engine offers advanced corpus tools including concordances, word sketches, word-sketch differences, frequency lists, phrase searches and specialised corpora.

For an advanced learner, one particularly useful function is comparing two similar words to see different collocational profiles.

But deeper tools increase the risk of misreading statistics. Complexity should follow a real question, not curiosity about buttons.

28. Oxford Collocations Dictionary is corpus-informed evidence packaged for learners | Oxford Collocations Dictionary 把语料证据整理成学习工具

The Oxford Collocations Dictionary provides collocational patterns and study resources organised around how words combine. It is often faster than running a full corpus query when your question is simply, “What normally goes with this noun?”

Corpus literacy also means knowing when not to use a corpus. Sometimes a well-designed learner dictionary has already done the filtering for you.

29. Use a learner dictionary first when the question is basic | 基础问题先用学习词典

If you need pronunciation, countability, transitivity, a standard preposition or a common sense, start with Oxford Advanced Learner’s Dictionary or Cambridge Dictionary.

Use corpus exploration when:

• two candidates both look possible
• you need more authentic examples
• you want collocation profiles
• you want genre/register evidence
• you want to inspect phrase variability

工具越高级，不代表每个问题都要用最复杂的工具。

30. Corpus workflow 1 — Write your sentence first | 流程一：先写你自己的句子

Do not browse language randomly. Begin from a production problem.

My draft: The study gives strong evidence that…

Your question might become:

Does English normally use “give strong evidence”, or are “provide strong evidence” and “offer strong evidence” more conventional?

Specific draft → specific question → useful search.

31. Corpus workflow 2 — Generate candidates | 流程二：先列候选表达

• give strong evidence
• provide strong evidence
• offer strong evidence
• present strong evidence

You can get candidates from your own knowledge, a learner dictionary, thesaurus or collocation dictionary. Corpus evidence then tests them.

语料库最适合“比较候选”，不一定最适合从零开始想所有表达。

32. Corpus workflow 3 — Read concordance lines | 流程三：先看上下文例句

A concordance shows the target word or phrase inside many short contexts. Read enough examples to determine whether they match your intended sense.

Ask:

• same meaning?
• same grammatical structure?
• same genre?
• same noun/verb relationship?
• same register?

Never treat a count as meaningful before checking examples.

33. Corpus workflow 4 — Look for repeated patterns | 流程四：找重复模式

If ten unrelated sources repeatedly use provide evidence for, that pattern deserves attention. If one strange phrase appears once, it may be a typo, translation, specialised usage or creative choice.

高级学习要学会区分：

• recurring convention
• possible but uncommon variation
• specialised usage
• creative deviation
• error/noise

34. Corpus workflow 5 — Check source type | 流程五：看来源属于什么文本

A phrase common in scientific papers may sound unnatural in casual speech. A phrase common in forums may be inappropriate in a legal memo.

Record source type when register matters:

• conversation
• fiction
• news
• academic
• professional
• technical
• web discussion

35. Corpus workflow 6 — Compare patterns, not just totals | 流程六：比较结构，不只比较总次数

Suppose substantial appears more often than considerable overall. That does not mean substantial is always a better choice. Compare the noun partners and senses relevant to your sentence.

总频率回答“这个词整体多不多”。搭配分布回答“在我这个结构里通常怎么用”。后者经常更重要。

36. Corpus workflow 7 — Return to your sentence | 流程七：回到你自己的句子

The corpus is not the destination. Rewrite your original sentence.

Draft: The study gives strong evidence that…
Revised: The study provides strong evidence that…

Then ask whether the revision genuinely improves meaning and naturalness.

如果只是因为“语料库里更多”就机械替换，你还没有完成判断。

37. Corpus workflow 8 — Save one useful pattern, not twenty screenshots | 流程八：保存一个可用模式，不要收藏二十张截图

Record:

• target phrase
• meaning/function
• two or three recurring partners
• register
• one authentic example
• one personal example

The goal is vocabulary learning, not corpus collecting.

38. Corpus workflow 9 — Retrieve later | 流程九：过一段时间再提取

A corpus session can create insight without durable learning. Return later:

• “What verb goes naturally with conclusion?”
• “What adjective commonly describes strong evidence?”
• “Which preposition follows consistent?”
• “How does substantial differ from significant?”

Corpus evidence should feed retrieval practice, not replace it.

39. Compare strong evidence vs powerful evidence | 示例一：strong evidence vs powerful evidence

Both can be understood. Corpus exploration typically reveals strong evidence as a highly conventional evaluative collocation across formal writing. Powerful evidence can occur, but may carry a more rhetorical feel or appear in narrower environments.

The lesson is not “powerful evidence is forbidden”. It is:

If you simply mean well-supported evidence, strong evidence is the safer conventional route.

语料库让你从“语法正确/错误”升级成“高概率/低概率 + 风格效果”的判断。

40. Compare significant increase vs substantial increase | 示例二：significant increase vs substantial increase

Both are common, but they can foreground different ideas.

• significant increase — noteworthy/important; may carry technical statistical meaning in research contexts
• substantial increase — large enough in amount or degree to matter

Corpus examples show the environments. Dictionary definitions explain the boundary. Your sentence determines which meaning you need.

41. Compare result in vs result from | 示例三：result in vs result from

• X results in Y → X causes/leads to Y
• Y results from X → Y originates/is caused by X

A corpus gives many examples of both directions. This is especially useful for Mandarin speakers because a single Chinese causal concept may not preserve the same grammatical direction.

42. Compare responsible for vs responsible to | 示例四：responsible for vs responsible to

Corpus examples can reveal different structures:

• responsible for a task / outcome
• responsible for doing something
• responsible to a person / authority in accountability relationships

The preposition changes the relationship, not merely the grammar.

43. Compare according to vs based on | 示例五：according to vs based on

Chinese “根据” may trigger both.

• according to X — often attributes information or a view to a source
• based on X — often states the foundation for a conclusion, design or decision

Corpus examples show recurring source nouns and sentence structures. This helps break one-to-one translation.

44. Compare discuss vs discuss about | 示例六：discuss vs discuss about

Standard English normally uses:

• discuss the issue
• have a discussion about the issue

Corpus evidence makes the pattern visible across many examples. The noun discussion can take about; the verb discuss usually takes the object directly.

This is a useful reminder that related word-family members can have different grammar.

45. Compare very likely vs highly likely | 示例七：very likely vs highly likely

Both occur. Corpus evidence can show genre and style distribution. Highly likely is a conventional formal combination; very likely is broad and natural across many contexts.

The answer is not “replace very with highly”. The answer is to understand which register and rhythm fit your sentence.

46. Compare deeply concerned vs highly concerned | 示例八：deeply concerned vs highly concerned

Both may occur, but deeply concerned is a strong conventional pairing expressing emotional seriousness. Corpus evidence helps learners discover adjective-specific intensifiers.

这就是为什么“very 的高级同义词”不是一个统一答案。不同形容词有不同搭配习惯。

47. Compare make an assumption vs take an assumption | 示例九：make an assumption

Corpus exploration shows recurring verbs:

• make an assumption
• challenge an assumption
• question an assumption
• test an assumption
• relax an assumption

The noun becomes a network of reasoning actions.

48. Compare pose a risk vs give a risk | 示例十：pose a risk

• pose a risk
• create a risk
• increase risk
• reduce risk
• mitigate risk
• manage risk

Corpus patterns turn a noun into an operational vocabulary system.

49. Use corpus evidence for word-family differences | 用语料库比较同词族成员

Example: economic / economical.

• economic growth
• economic policy
• economic conditions
• economical car
• economical use of resources

The collocational profiles make the semantic split visible.

Word-family form does not guarantee interchangeable meaning.

50. Use corpus evidence for polysemy | 用语料库拆开多义词

Target: address.

• address a problem
• address a concern
• address the audience
• email address
• home address

Corpus examples cluster around different senses. Learners can observe which objects signal which meaning.

51. Use corpus evidence to build phrase frames | 用语料库建立可生产的词块框架

Target:

There is little evidence to suggest that X.

Inspect what kinds of claims fill X. Then create your own examples across domains.

• There is little evidence to suggest that the policy caused the decline.
• There is little evidence to suggest that the intervention produces lasting benefits.
• There is little evidence to suggest that the effect generalises to all age groups.

Corpus evidence teaches the frame; deliberate practice makes it yours.

52. Use corpus evidence to detect noun-heavy style | 用语料库观察名词化风格

Academic corpora contain nominalisations such as implementation, evaluation, identification. That does not mean every sentence should imitate the highest noun density you can find.

Corpus examples show what exists; stylistic judgement decides what belongs in your sentence.

真实存在 ≠ 对你这个句子最好。

53. Use corpus evidence to detect phrase flexibility | 用语料库看词块哪里固定、哪里可变

Frame:

play a ___ role in

• play a significant role in
• play a major role in
• play a central role in
• play a limited role in
• play an important role in

You learn a productive pattern rather than five unrelated phrases.

54. Use corpus evidence for professional workflow vocabulary | 用语料库验证职场流程词汇

• allocate resources
• assess feasibility
• identify constraints
• mitigate risk
• resolve issues
• revise estimates
• defer a decision
• assign responsibility

Search examples to see what subjects, objects and contexts normally surround each phrase.

This helps distinguish textbook translation from operational professional English.

55. Use corpus evidence for academic reporting verbs | 用语料库验证学术报告动词

• argue
• suggest
• indicate
• demonstrate
• report
• observe
• maintain
• establish

Inspect common subjects and complements. Does evidence suggest, indicate or demonstrate? Do authors argue, maintain or observe? Distribution reveals discourse roles.

56. Use corpus evidence for hedging | 用语料库观察“谨慎表达”

• may indicate
• appears to suggest
• is likely to
• to some extent
• under certain conditions
• cannot rule out

Corpus examples show how caution clusters with evidence language. This can help Mandarin-speaking learners who otherwise translate claims too directly and too strongly.

57. Use corpus evidence for spoken academic English | 学术口语也要看真实用法

• what I mean is…
• the main point is…
• if we look at…
• the thing to notice is…
• let me put that another way…

Spoken corpora remind learners that advanced academic speech is often lexically simpler than polished journal prose but structurally rich in signposting.

58. Use corpus evidence to compare spoken and written register | 用语料库比较口语和写作

Concept: approximately.

• conversation may prefer about
• formal written contexts may use approximately

Concept: commence.

• ordinary speech usually prefers start/begin
• official/formal writing can use commence

Corpus distributions make register visible.

59. Frequency is useful, but frequency is not importance | 高频有用，但高频不等于“最重要”

Common words deserve attention because they recur. But a lower-frequency term may be crucial in your field.

An engineer may need tolerance, redundancy, degradation. A finance professional may need liquidity, leverage, exposure. A teacher may need retrieval, misconception, scaffolding.

Corpus frequency must be combined with personal relevance.

60. Raw frequency can mislead | 原始次数会误导

If one corpus has ten times more academic text than another, raw counts cannot be compared directly. If one phrase occurs 1,000 times in a huge corpus and another 500 times in a much smaller corpus, absolute totals hide relative frequency.

For ordinary learners, the practical rule is simple:

Do not treat a number as meaningful until you know what it is counting and what the corpus contains.

61. Association scores can also mislead beginners | 搭配强度分数也可能误导初学者

Corpus systems may rank collocations using statistical association measures. A rare but unusually exclusive pairing can score highly, while a very useful common phrase may not look as dramatic statistically.

You do not need to master corpus statistics to learn vocabulary. First inspect:

• frequency
• range across sources
• meaning
• register
• usefulness to you

Statistics support judgement; they do not replace it.

62. Duplicated web text can inflate counts | 网络语料可能有重复文本

Large web corpora are powerful but can contain duplicated or templated language. Modern corpus builders use cleaning and deduplication methods, but no dataset is perfect.

If a surprising phrase appears many times, check whether examples come from diverse sources.

“出现很多次”比“很多不同作者都这样用”弱。

63. Search-engine counts are not the same as corpus evidence | 搜索引擎结果数量不等于语料库证据

General web search can help locate examples, but result counts are not designed as stable linguistic frequency measures. Search ranking, duplication, personalisation, indexing and query interpretation all interfere.

Use web search for discovery. Use corpus/learner resources for systematic language evidence when the distinction matters.

64. Corpus examples can contain bad English | 语料里也会有错误英语

Real language includes learner errors, typos, informal deviations and poorly edited writing. A corpus records usage; it does not certify every sentence.

This is especially important in broad web corpora.

Use repeated patterns across credible contexts, learner dictionaries and editorial judgement together.

65. Corpus evidence cannot decide your intended meaning | 语料库不知道你真正想表达什么

Suppose significant increase is more frequent than substantial increase in a corpus. If you specifically want to emphasise magnitude rather than importance, substantial may still be the better word.

The corpus sees language. You know the communicative intention.

频率不能替你思考。

66. Corpus evidence cannot decide your audience | 语料库也不知道你的听众是谁

A phrase may be common in academic papers but inappropriate in a message to a friend. Always return to register and audience.

Naturalness is contextual, not merely statistical.

67. Corpus evidence cannot replace subject knowledge | 语料库不能替代专业知识

A phrase can be common and still be factually wrong in your sentence. Technical vocabulary may require domain expertise to select accurately.

For example, correlation, causation, association are common terms, but no corpus can decide whether your study design justifies a causal claim.

Language evidence and world evidence are different.

68. Use corpus evidence after meaning, not before meaning | 先搞清意思，再查语料

Weak workflow:

1. search random synonyms
2. choose the most frequent
3. insert into sentence

Strong workflow:

1. clarify intended meaning
2. generate plausible candidates
3. check definitions and boundaries
4. inspect corpus patterns
5. return to sentence
6. choose by meaning + usage + audience

Corpus checking belongs inside a larger lexical decision process.

69. Build a corpus-check notebook | 建立“语料验证笔记”

Field	Example
My draft	give strong evidence
Question	give vs provide evidence?
Candidates	give, provide, offer, present
Corpus pattern	provide evidence highly recurrent
Register	neutral-formal
Decision	provide strong evidence
Personal sentence	The results provide strong evidence for…
Next retrieval	3 days

One solved language problem becomes one reusable route.

70. Build a “my repeated uncertainties” list | 记录你经常不确定的类型

• prepositions
• adjective–noun collocations
• academic reporting verbs
• formal vs neutral alternatives
• Chinese-to-English verb choice
• countability
• near-synonyms
• phrasal verbs

If 60% of your corpus checks concern prepositions, your first weak link is not “vocabulary generally”. It is dependent-preposition control.

71. Use corpus checks to build personal rules | 把反复发现变成“个人规则”

• I often add unnecessary about after verbs.
• I overuse very instead of adjective-specific intensifiers.
• I overuse improve where a more specific mechanism verb is needed.
• I choose formal words in ordinary conversation.
• I know nouns but not their normal verbs.

Corpus research becomes powerful when it changes your future behaviour.

72. Use corpus evidence before high-stakes writing | 重要写作前，用语料验证高风险表达

For an application, report, paper or professional presentation, check phrases that:

• carry the main claim
• express evidence strength
• state limitations
• use specialist terminology
• contain unfamiliar collocations
• sound translated to you

You do not need to corpus-check every sentence. Check the lexical decisions with the highest consequence.

73. Use corpus evidence during editing, not only drafting | 语料库最适合放在编辑阶段

Draft for meaning first. Then flag uncertain phrases.

• ? collocation
• ? preposition
• ? register
• ? synonym
• ? academic phrase

After the paragraph exists, corpus checking answers real questions instead of interrupting every sentence.

74. Use corpus evidence for oral preparation too | 口语准备也可以用语料验证

Before a presentation or interview, verify high-value phrase frames:

• the main challenge is…
• one possible explanation is…
• the evidence suggests…
• the key trade-off is…
• from a practical perspective…

Then practise speaking them. Corpus evidence alone will not make a phrase active.

75. Use corpus evidence to create better flashcards | 用语料证据改进单词卡

Instead of:

constraint = 限制

Build:

• budget constraint
• time constraint
• capacity constraint
• operate under constraints
• face constraints

The corpus supplies authentic phrase routes; retrieval practice makes them available.

76. Use corpus evidence to create contrast cards | 用语料证据改进近义词对比卡

Target pair: effective / efficient.

• effective treatment
• effective method
• efficient process
• efficient system
• energy-efficient

Definitions give the boundary; collocation profiles make it memorable.

77. Use corpus evidence to create register cards | 用语料证据做语域卡

Target concept: “find out”.

• find out — common conversational/neutral
• determine — neutral-formal analytical
• ascertain — formal, often professional/legal/technical

Collect examples from different genres. Now register becomes evidence-based.

78. Use corpus evidence to create phrase-family cards | 用语料证据做“词块族”

Target frame:

be consistent with

• consistent with the evidence
• consistent with previous findings
• consistent with the hypothesis
• broadly consistent with
• not entirely consistent with

The phrase becomes flexible rather than frozen.

79. The FENCE corpus-check system | FENCE 语料验证系统

Fence	Question	Action
F0 Intent	What do I mean?	state meaning before searching
F1 Candidate	What forms might work?	generate 2–4 candidates
F2 Dictionary	Are candidates semantically possible?	check sense/grammar/register
F3 Examples	Do real contexts match my sense?	read concordance lines
F4 Pattern	What combinations repeat?	identify collocations/frames
F5 Distribution	Where is the pattern used?	check genre/domain/region
F6 Compare	How do candidates differ?	compare profiles, not just counts
F7 Decide	Which fits my sentence?	meaning + probability + audience
F8 Learn	Is this worth retaining?	record pattern, retrieve, transfer

The corpus sits in the middle of the decision, not at the top of it.

80. A two-minute corpus check | 两分钟快速验证

1. Write your intended phrase.
2. Search exact phrase.
3. Read 5–10 examples.
4. Search one alternative.
5. Compare source types.
6. Return to sentence.

Use this for low-risk questions such as collocation or preposition checks.

81. A ten-minute deep corpus check | 十分钟深入验证

1. Clarify intended meaning.
2. Check dictionary senses.
3. Generate candidates.
4. Search each candidate.
5. Read 15–30 relevant examples.
6. Compare collocation profiles.
7. Check register/domain.
8. Record one decision rule.
9. Rewrite your sentence.
10. Create one new sentence without notes.

Use deep checks for important writing or recurring weaknesses.

82. A 30-minute corpus-learning session | 30 分钟语料学习训练

Time	Task
0–8 min	Choose 3 real language questions from your writing/reading.
8–18 min	Search examples and collocations.
18–23 min	Write decision rules.
23–27 min	Rewrite original sentences.
27–30 min	Close tools and retrieve the patterns.

Three solved questions are enough.

83. A seven-day corpus literacy challenge | 七天语料库学习挑战

Day	Focus
1	verb–noun collocations
2	adjective–noun collocations
3	dependent prepositions
4	near-synonym comparison
5	spoken vs written register
6	academic/professional phrase frames
7	mixed checks from your own writing

Do not collect random discoveries. Every search should begin from a real language question.

84. A 30-day corpus literacy challenge | 30 天语料验证挑战

• Solve 50 real lexical questions.
• Record 30 collocation patterns.
• Compare 15 near-synonym pairs.
• Check 15 dependent-preposition or complement patterns.
• Build 10 spoken-vs-written register comparisons.
• Build 10 academic/professional phrase frames.
• Revisit 10 old pieces of writing and repair corpus-verifiable weaknesses.
• Complete 4 no-tool retrieval tests.

The result should not be “I know how to use a corpus interface.” The result should be “I ask better language questions and make better lexical decisions.”

85. A 12-week progression | 12 周语料证据路线

Weeks	Focus	Evidence of progress
1–2	examples + exact phrase checks	fewer obvious collocation errors
3–4	Word Sketch / collocation profiles	stronger phrase networks
5–6	near-synonym comparison	more precise lexical choice
7–8	register/domain variation	better audience calibration
9–10	academic/professional phrase evidence	more natural formal output
11–12	independent editing + retrieval	less dependence on tools for repeated patterns

86. The first weak link diagnostic | 第一个弱点诊断

Symptom	Likely weak link	Repair
I search random words for too long.	question design	start from a sentence-level problem
I trust counts without reading examples.	semantic filtering	read concordance first
I choose the most frequent synonym automatically.	meaning control	clarify intent before frequency
I copy phrases but misuse them.	context/register	record source type + function
I find patterns but forget them.	retrieval	save fewer patterns + spaced recall
I use formal corpus phrases in casual speech.	register calibration	compare spoken/written sources
I find contradictory results.	corpus composition/sense mixture	filter by sense, genre, domain
I spend more time researching than writing.	tool overuse	time-box checks, edit after drafting

Corpus literacy should reduce uncertainty efficiently, not become a new form of procrastination.

87. How a strong tutor teaches corpus literacy | 高水平导师怎样教语料库学习？

A weak lesson demonstrates a tool. A strong lesson starts with a learner’s real sentence and turns the tool into evidence for a decision.

• ask learner to state intended meaning
• generate candidates before searching
• read examples before counts
• compare collocation profiles
• identify source register
• ask learner to justify final choice
• return to original writing
• retrieve pattern later without tool
• track repeated uncertainty types
• reduce tool dependence over time

The tutor’s goal is not to become the learner’s permanent corpus operator. The goal is to make evidence-based checking independent.

88. The learner should become an evidence-based lexical editor | 最终，学习者要成为“有证据的选词编辑”

You should increasingly be able to think:

• I know what I mean.
• Two words seem possible.
• The dictionary confirms both senses.
• The corpus shows different collocation profiles.
• This one is more common in my register.
• This phrase better matches my intended meaning.
• I can now use it without reopening the tool.

That is a mature lexical decision process.

89. Canonical ownership and collision boundary | 本课的所有权边界

This lesson owns the Chinese-supported C1–C2 learner route for using corpus evidence to verify word choice and phrase probability. Existing pages retain mechanism ownership.

• How Collocation Turns Vocabulary Knowledge Into Natural, Precise English — collocation mechanism.
• How Synonyms, Antonyms and Category Relationships Build a More Precise Vocabulary — lexical relationship mechanism.
• How Connotation Changes Meaning Beyond Dictionary Definitions — connotation mechanism.
• How Spelling and Word Formation Build English Control — morphology/word-formation mechanism.
• How Context Clues Build Vocabulary Independence — contextual inference mechanism.

This page coordinates them around one advanced learner problem: “My sentence seems possible. How do I check whether English usually makes this lexical choice?”

90. Recommended apex resources | 推荐优质资源

• SKELL — Sketch Engine for Language Learning — learner-friendly examples, collocations and similar-word exploration using large corpora.
• Sketch Engine — advanced concordance, Word Sketch, word comparison, frequency and corpus-query tools.
• English SKELL Corpus — documentation for the learner-oriented English corpus behind SKELL.
• Oxford Collocations Dictionary — learner-focused collocation patterns and study resources.
• Oxford Advanced Learner’s Dictionary — definitions, examples, grammar, register and collocations.
• Cambridge Dictionary — learner definitions, examples and pronunciation; Cambridge dictionaries are corpus-informed.
• Cambridge Dictionary: corpus — concise explanation of a language corpus and its use in studying language.
• Cambridge English Vocabulary in Use: Advanced — C1–C2 vocabulary developed with corpus-informed selection and contextual practice.
• English Collocations in Use: Advanced — systematic advanced collocation study.

91. SEO language map | 本课关键词范围

This lesson naturally serves English corpus for learners, corpus linguistics vocabulary learning, how to check English collocations, natural English word choice, SKELL English, Sketch Engine English learning, English corpus examples, corpus collocations, corpus vocabulary, advanced English collocations, C1 vocabulary, C2 vocabulary, advanced English vocabulary for Chinese speakers, English phrase frequency, English register corpus, near synonyms corpus, academic English corpus, professional English vocabulary, Oxford Collocations Dictionary, 语料库英语学习, 英语语料库, 用语料库学英语, 英语搭配查询, 英语自然表达, 英语词频, 英语词块, 英语语域, 高级英语词汇, C1 英语词汇, C2 英语词汇 and 中文母语英语学习.

Search language belongs inside genuine teaching, not mechanical repetition.

92. Your assignment | 本课作业

Take one piece of English you wrote recently. Mark exactly five lexical uncertainties.

For each one:

1. State what you actually mean.
2. Generate two or three candidate phrases.
3. Check dictionary meaning and grammar.
4. Search examples in SKELL or another reliable corpus resource.
5. Read at least ten relevant examples.
6. Identify repeated collocations or phrase frames.
7. Check register/source type.
8. Choose the best option for your sentence.
9. Rewrite the sentence.
10. Record one decision rule.
11. Close the tool and create one new example from memory.

Do not search a sixth phrase until you can explain the first five decisions.

真正的语料库学习，不是“找到更多例句”，而是“以后少做一次不必要的猜测”。

93. The final rule | 最后一条规则

Use corpus evidence to test your language, not to outsource your judgement.

用语料库验证你的语言，不要把判断外包给语料库。

Meaning comes first. Corpus evidence shows conventional routes. Register narrows the field. Domain knowledge protects factual accuracy. Your communicative purpose makes the final choice.

At first, you will check many phrases. Later, repeated evidence becomes internalised. Strong evidence, make progress, reach a conclusion, responsible for, consistent with begin to feel available without a search.

That is the real purpose of corpus literacy: not permanent dependence on a tool, but a more accurate internal model of how English behaves.

看真实英语怎样走，最后让这些高概率路线进入你自己的英语。

---

Series: EDKS-ADV-VOC-ZH · Lesson 009 · Batch 03.