[Projeto] [Desafio] Resumindo um texto longo

Para cumprir o desafio, utilizei os conhecimentos aprendidos no curso junto com as orientações do próprio ChatGPT.

Prompt:
´´´
Hi, GPT! I'm working on a paper where I have to summarize a long text. My goal is to have a short, cohesive and coherent text that highlights the most important points and is still completely true to the original, without missing any relevant information. How do you suggest we approach it?
´´´

Resposta:

A good summary is usually less about “cutting words” and more about making deliberate decisions about meaning, hierarchy, and structure.

For the kind of summary you described — short, cohesive, coherent, faithful to the original, and complete in terms of relevant information — I’d suggest a process closer to analytical condensation rather than simple shortening.

Here’s the approach I recommend:
1) First: Understand the Text Before Trying to Reduce It
    a) Central Thesis
    b) Supporting Pillars
2) Reduce by “Idea Compression,” Not Sentence Deletion
3) Preserve Logical Flow
4) Maintain Fidelity to the Original
5) Think in Layers
6) A Practical Workflow We Can Use Together
    a) Step 1 — You send the original text
    b) Step 2 — We identify:
        i) central idea, 
        ii) essential points, 
        iii) optional details, 
        iv) repeated information.
    c) Step 3 — We create:
        i) a first condensed version, 
        ii) then refine it (cohesion, coherence, conciseness, academic tone, word limit, or readability)
7) A Very Useful Technique: “Progressive Compression”
    a) Lower the number of words by each version while preserving meaning
        i) start with 300, then 150, then 80
8) Follow Technical Guidelines, depending on the context

A técnica de dividir o texto em partes pode se tornar complicada quando temos muitos pré-requisitos e contextos específicos. Então, para garantir a exatidão do resumo final ao mesmo tempo em que seguimos as sugestões oferecidas pelo ChatGPT, eu pedi para que ele criasse prompts que assegurassem a correta execução durante todo o processo.

O resultado veio em forma de um “sistema de resumo” dividido em 4 partes (cada uma com seu prompt):

Resposta:

    Layer 1 — Structural Extraction
    Layer 2 — Faithful Condensation
    Layer 3 — Cohesion & Integration
    Layer 4 — Final Compression Pass

**The Most Important Rule for Long Projects**

We should maintain a: ***“Living Context Memory”***

Meaning:
Before summarizing each new section, we briefly keep track of:
* what has already been covered, 
* recurring concepts, 
* terminology choices, 
* and the overall direction of the text. 

This dramatically improves consistency.

Para finalizar, pedi para o ChatGPT comparar a versão original com o resumo e recebi o seguinte retorno:

**Final Verdict**

The summary is substantially faithful to the original text.
It:
    • preserves the essential technical meaning, 
    • avoids factual invention, 
    • maintains conceptual relationships, 
    • and accurately represents Python’s philosophy, architecture, paradigms, implementations, and ecosystem. 

The main tradeoffs are:
    • reduced historical granularity, 
    • compressed technical explanations, 
    • lighter institutional/contextual detail, 
    • and slight shifts in emphasis toward conceptual synthesis over chronological exposition. 

**Conclusion**
The summary can be considered:
    • academically reliable, 
    • technically accurate, 
    • and conceptually faithful, 

though not fully exhaustive regarding:
    • historical evolution, 
    • institutional context, 
    • and some implementation nuances.

Minha Análise

Resumo do resultado:

• Criei 6 chats diferentes (dentro de um mesmo projeto, para manter algum contexto entre eles com a ferramenta)
• Texto original: 3737 palavras | Resumo final: 466 palavras
• Resumo manteve todos os marcadores inicialmente solicitados (coesão, coerência, clareza, fluidez e fidelidade ao original).

Criar resumos de textos longos pode ser trabalhoso, e notei que uma das partes mais importantes do processo é a contextualização.
Definir necessidades e objetivos claros, assim como incluir a IA nessa etapa prévia, faz com que as técnicas aprendidas no curso funcionem de um jeito muito mais ágil e eficiente.