The "Google Translate" Error

You paste a sentence into a translator.

• English: "I am running late."
• Translation (Literal): "Estoy corriendo tarde." (I am physically jogging... late?)

Native speakers laugh. Why? Because "Running late" is not about locomotion. It is a Chunk. In Spanish, you don't "run" late. You "have" lateness ("Tengo retraso").

The machine translated the Words (Atoms), but missed the Molecule. Language is not made of isolated words. It is made of Lexical Chunks (Collocations and Idioms). This is the core insight of the Lexical Approach (Michael Lewis). Fluency is not "Knowing many words." Fluency is "Knowing which words go together."

Part 1: Collocations (The Magnetic Words)

Collocations are words that have a statistical probability of appearing together. They are "Magnetic."

• Strong Collocation: "Heavy rain." (Everyone says this).
• Weak Collocation: "Strong rain." (Grammatically correct, but sounds weird).
• Wrong Collocation: "Big rain." (Native speakers calculate you are a beginner).

Types of Collocations:

1. Adjective + Noun: Heavy traffic, Bright future, Deep sleep.
2. Verb + Noun: Make a decision (Not "do"), Take a shower (Not "make"), Pay attention (Not "give").
3. Adverb + Adjective: Fully aware, Highly controversial, Fast asleep.

The Learning Problem: Most dictionaries fail here. If you look up "Heavy," it says "Weighs a lot." That definition is useless for "Heavy traffic." Cars don't weigh more when there are many of them. You need a dictionary that defines "Heavy Traffic" as a single unit meaning "Congestion."

Part 2: Idioms (The Cultural Code)

Idioms are "Fixed Expressions" where the meaning is completely detached from the individual words. They are metaphors that have fossilized.

• "Kick the bucket" = Die. (Nothing to do with feet or pails).
• "Spill the beans" = Reveal secrets.
• "Piece of cake" = Easy.

The Transparency Scale:

1. Transparent: "See the light" (Understand). You can guess this.
2. Semi-Opaque: "Vote with your feet" (Leave). Makes sense metaphors.
3. Opaque: "Red Herring" (Distraction). Impossible to guess without history.

Standard translators often literalize idioms, turning poetry into nonsense. AI is uniquely good at detecting Idioms because it recognizes the vector pattern. It sees "Piece of cake" and maps it to the concept "Easy," ignoring the concept "Dessert."

Part 3: The Polysemy Problem (The Word "Set")

The English word "Set" has 430 definitions in the Oxford English Dictionary. It is the most complex word in the language.

• Set a table. (Arrange).
• Set a record. (Establish).
• Set of tennis. (Group).
• Set in concrete. (Harden).
• The sun sets. (Go down).
• Set fire. (Ignite).

How does a human know which "Set" is being used? Context. We look at the neighbor words (Collocates). If neighbor = "Sun," Set = "Down." If neighbor = "Table," Set = "Arrange."

Learners get stuck because they memorize Definition #1 and try to apply it everywhere. "The sun arranged itself." (Confusion). Tools like Text Clarifier perform "Word Sense Disambiguation" (WSD). They highlight the specific definition relevant to this specific sentence.

Part 4: Phrasal Verbs (The Nightmare)

English loves Phrasal Verbs. VERB + PREPOSITION = NEW MEANING.

• Look: Use eyes.
• Look up: Search.
• Look after: Care for.
• Look down on: Disrespect.
• Look forward to: Anticipate.

These are essentially "Mini-Idioms." They destroy comprehension for learners. "I ran into my friend." Learner thinks: "Did you hurt him? Did you collide?" No. "Ran into" = Met unexpectedly.

Part 5: Conclusion

Stop collecting words like stamps. Collect Chunks. Don't write down "Take." Write down "Take a break," "Take a chance," "Take advantage." That is how natives speak. If you ignore Collocations, you will always speak "Correctly" but sound "Foreign."

References:

• Lewis, M. (1993). The Lexical Approach.
• Nattinger, J. (1992). Lexical Phrases and Language Teaching.
• Moon, R. (1998). Fixed Expressions and Idioms in English.