Text splitters in Prompt Engineering / GenAI - Model Pipeline Trace

Model Pipeline - Text splitters

This pipeline breaks long text into smaller pieces so a model can understand and work with it better. It splits text by sentences or paragraphs before processing.

Data Flow - 4 Stages

1Raw Text Input

1 document x 1000 words→Receive full text document→1 document x 1000 words

"Today is sunny. We will go to the park. Then have lunch."

↓

2Sentence Splitter

1 document x 1000 words→Split text into sentences using punctuation→3 sentences x variable length

["Today is sunny.", "We will go to the park.", "Then have lunch."]

↓

3Paragraph Splitter

1 document x 1000 words→Split text into paragraphs by newline characters→3 paragraphs x variable length

["Today is sunny. We will go to the park.", "Then have lunch.", "After that, we will read books."]

↓

4Chunk Creation

10 sentences or 3 paragraphs→Group sentences or paragraphs into chunks of max 100 words→5 chunks x up to 100 words each

["Today is sunny. We will go to the park.", "Then have lunch. After that, we will read books."]

Training Trace - Epoch by Epoch

Loss
0.5 |****
0.4 |*** 
0.3 |**  
0.2 |*   
0.1 |    
     +----
      1 2 3 4 Epochs

Epoch	Loss ↓	Accuracy ↑	Observation
1	0.45	0.60	Initial split quality is moderate, some sentences split incorrectly.
2	0.30	0.75	Improved splitting rules reduce errors, better sentence boundaries.
3	0.20	0.85	Splitting is mostly correct, chunk sizes optimized for model input.
4	0.15	0.90	Final tuning reduces overlap and preserves context well.

Prediction Trace - 3 Layers

Layer 1: Input Raw Text

Layer 2: Sentence Splitter

Layer 3: Chunk Creation

Model Quiz - 3 Questions

What is the main purpose of the sentence splitter stage?

A To combine sentences into paragraphs

B To break text into smaller sentences for easier processing

C To remove punctuation from the text

D To translate text into another language

Key Insight

Splitting text into smaller, meaningful pieces helps models understand context better and improves processing efficiency. Training improves the splitting rules to reduce errors and optimize chunk sizes.

Practice

1. What is the main purpose of a text splitter in AI applications?

easy

A. To translate text into different languages

B. To generate new text from a prompt

C. To break long text into smaller, manageable pieces

D. To summarize text into a single sentence

Solution

1. Step 1: Understand the role of text splitters

   Text splitters are designed to divide long text into smaller parts for easier processing.

2. Step 2: Compare options to the definition

   Only To break long text into smaller, manageable pieces describes breaking text into smaller pieces, which matches the purpose of text splitters.

3. Final Answer:

   To break long text into smaller, manageable pieces -> Option C

4. Quick Check:

   Text splitter purpose = break text [OK]

Hint: Text splitters cut text into chunks for easier handling [OK]

Common Mistakes:

• Confusing splitting with translation
• Thinking splitters summarize text
• Assuming splitters generate new text

2. Which of the following is the correct way to set chunk size and overlap in a text splitter?

easy

A. chunk_size='100', overlap=20

B. chunkSize=100, overlap=20

C. chunk_size=100, overlap=twenty

D. chunk_size=100, overlap=20

Solution

1. Step 1: Identify correct parameter names and types

   Parameters should be named with underscores and numeric values for size and overlap.

2. Step 2: Check each option for syntax and type correctness

   chunk_size=100, overlap=20 uses correct parameter names and numeric values; others have wrong names or types.

3. Final Answer:

   chunk_size=100, overlap=20 -> Option D

4. Quick Check:

   Correct param names and numeric values = chunk_size=100, overlap=20 [OK]

Hint: Use underscores and numbers for chunk size and overlap [OK]

Common Mistakes:

• Using camelCase instead of snake_case
• Passing string instead of number for overlap
• Misspelling parameter names

3. Given the following Python code using a text splitter:

text = "Hello world! This is a test of text splitting."
chunk_size = 12
overlap = 4
splitter = TextSplitter(chunk_size=chunk_size, overlap=overlap)
chunks = splitter.split(text)
print(chunks)

What is the expected output?

medium

A. ["Hello world!", "world! This is", "This is a test", "a test of text", "of text splitting."]

B. ["Hello world! This", "This is a test of", "test of text splitting."]

C. ["Hello world! This is a test of text splitting."]

D. ["Hello", "world!", "This", "is", "a", "test", "of", "text", "splitting."]

Solution

1. Step 1: Understand chunk size and overlap effect

   Chunk size 12 means each piece has up to 12 characters; overlap 4 means next chunk starts 4 characters before previous ends.

2. Step 2: Apply splitting logic to the text

   Chunks are: "Hello world!" (12 chars), then start 8 chars in (12-4=8) at "world! This is", and so on, producing the listed chunks in ["Hello world!", "world! This is", "This is a test", "a test of text", "of text splitting."].

3. Final Answer:

   ["Hello world!", "world! This is", "This is a test", "a test of text", "of text splitting."] -> Option A

4. Quick Check:

   Chunk size 12 + overlap 4 = overlapping chunks [OK]

Hint: Chunk size limits length; overlap repeats last part [OK]

Common Mistakes:

• Ignoring overlap and making chunks non-overlapping
• Using wrong chunk sizes
• Returning entire text as one chunk

4. Consider this code snippet that tries to split text but raises an error:

text = "Sample text for splitting."
splitter = TextSplitter(chunk_size='10', overlap=3)
chunks = splitter.split(text)

What is the most likely cause of the error?

medium

A. chunk_size should be an integer, not a string

B. overlap cannot be less than 5

C. TextSplitter requires a minimum chunk_size of 20

D. The text variable is too short to split

Solution

1. Step 1: Check parameter types

   chunk_size is given as a string '10' instead of an integer 10, which causes a type error.

2. Step 2: Validate other options

   Overlap 3 is valid; no minimum chunk size of 20 is required; text length is sufficient.

3. Final Answer:

   chunk_size should be an integer, not a string -> Option A

4. Quick Check:

   Parameter type mismatch = chunk_size should be an integer, not a string [OK]

Hint: Use numbers, not strings, for chunk size and overlap [OK]

Common Mistakes:

• Passing chunk_size as string
• Assuming overlap minimum is 5
• Thinking text length causes error

5. You have a very long document and want to split it for an AI model that can only process 500 tokens at a time. You want some context overlap to keep meaning. Which approach best balances chunk size and overlap?

hard

A. Set chunk_size to 600 tokens and overlap to 0 tokens

B. Set chunk_size to 500 tokens and overlap to 100 tokens

C. Set chunk_size to 400 tokens and overlap to 200 tokens

D. Set chunk_size to 100 tokens and overlap to 50 tokens

Solution

1. Step 1: Understand model token limit and overlap purpose

   The model can process 500 tokens max; overlap adds repeated context to help understanding.

2. Step 2: Evaluate options for chunk size and overlap

   Set chunk_size to 500 tokens and overlap to 100 tokens uses chunk size 500 (max allowed) and overlap 100 (reasonable context). Others exceed limit or have too small chunks.

3. Final Answer:

   Set chunk_size to 500 tokens and overlap to 100 tokens -> Option B

4. Quick Check:

   Chunk size ≤ 500 with overlap for context = Set chunk_size to 500 tokens and overlap to 100 tokens [OK]

Hint: Keep chunk size at max limit, add moderate overlap [OK]

Common Mistakes:

• Exceeding model token limit
• Setting overlap too large or zero
• Using very small chunk sizes unnecessarily