C#によるAIパワーPDF処理：IronPDFでドキュメントを要約、抽出、分析する

PM > Install-Package IronPdf
PM > Install-Package IronPdf.Extensions.AI
PM > Install-Package Microsoft.SemanticKernel
PM > Install-Package Microsoft.SemanticKernel.Plugins.Memory

// Initialize IronPDF AI with Azure OpenAI credentials
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel with Azure OpenAI
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

// Create memory store for document embeddings
var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

// Initialize IronPDF AI
IronDocumentAI.Initialize(kernel, memory);

Console.WriteLine("IronPDF AI initialized successfully with Azure OpenAI");

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel with Azure OpenAI
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

' Create memory store for document embeddings
Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

' Initialize IronPDF AI
IronDocumentAI.Initialize(kernel, memory)

Console.WriteLine("IronPDF AI initialized successfully with Azure OpenAI")

// Synthesize insights across multiple related documents (e.g., quarterly reports into annual summary)
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

// Define documents to synthesize
string[] documentPaths = {
    "Q1-report.pdf",
    "Q2-report.pdf",
    "Q3-report.pdf",
    "Q4-report.pdf"
};

var documentSummaries = new List<string>();

// Summarize each document
foreach (string path in documentPaths)
{
    var pdf = PdfDocument.FromFile(path);
    string summary = await pdf.Summarize();
    documentSummaries.Add($"=== {Path.GetFileName(path)} ===\n{summary}");
    Console.WriteLine($"Processed: {path}");
}

// Combine and synthesize across all documents
string combinedSummaries = string.Join("\n\n", documentSummaries);

var synthesisDoc = PdfDocument.FromFile(documentPaths[0]);

string synthesisQuery = @"Based on the quarterly summaries below, provide an annual synthesis:
1. Overall trends across quarters
2. Key achievements and challenges
3. Year-over-year patterns

Summaries:
" + combinedSummaries;

string synthesis = await synthesisDoc.Query(synthesisQuery);

Console.WriteLine("\n=== Annual Synthesis ===");
Console.WriteLine(synthesis);

File.WriteAllText("annual-synthesis.txt", synthesis);

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.IO

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

' Define documents to synthesize
Dim documentPaths As String() = {
    "Q1-report.pdf",
    "Q2-report.pdf",
    "Q3-report.pdf",
    "Q4-report.pdf"
}

Dim documentSummaries = New List(Of String)()

' Summarize each document
For Each path As String In documentPaths
    Dim pdf = PdfDocument.FromFile(path)
    Dim summary As String = Await pdf.Summarize()
    documentSummaries.Add($"=== {Path.GetFileName(path)} ==={vbCrLf}{summary}")
    Console.WriteLine($"Processed: {path}")
Next

' Combine and synthesize across all documents
Dim combinedSummaries As String = String.Join(vbCrLf & vbCrLf, documentSummaries)

Dim synthesisDoc = PdfDocument.FromFile(documentPaths(0))

Dim synthesisQuery As String = "Based on the quarterly summaries below, provide an annual synthesis:" & vbCrLf &
    "1. Overall trends across quarters" & vbCrLf &
    "2. Key achievements and challenges" & vbCrLf &
    "3. Year-over-year patterns" & vbCrLf & vbCrLf &
    "Summaries:" & vbCrLf & combinedSummaries

Dim synthesis As String = Await synthesisDoc.Query(synthesisQuery)

Console.WriteLine(vbCrLf & "=== Annual Synthesis ===")
Console.WriteLine(synthesis)

File.WriteAllText("annual-synthesis.txt", synthesis)

// Generate executive summary from strategic documents for C-suite leadership
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("strategic-plan.pdf");

string executiveQuery = @"Create an executive summary for C-suite leadership. Include:

**Key Decisions Required:**
- List any decisions needing executive approval

**Critical Findings:**
- Top 3-5 most important findings (bullet points)

**Financial Impact:**
- Revenue/cost implications if mentioned

**Risk Assessment:**
- High-priority risks identified

**Recommended Actions:**
- Immediate next steps

Keep under 500 words. Use business language appropriate for board presentation.";

string executiveSummary = await pdf.Query(executiveQuery);

File.WriteAllText("executive-summary.txt", executiveSummary);
Console.WriteLine("Executive summary saved to executive-summary.txt");

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("strategic-plan.pdf")

Dim executiveQuery As String = "Create an executive summary for C-suite leadership. Include:" & vbCrLf & vbCrLf & 
"**Key Decisions Required:**" & vbCrLf & 
"- List any decisions needing executive approval" & vbCrLf & vbCrLf & 
"**Critical Findings:**" & vbCrLf & 
"- Top 3-5 most important findings (bullet points)" & vbCrLf & vbCrLf & 
"**Financial Impact:**" & vbCrLf & 
"- Revenue/cost implications if mentioned" & vbCrLf & vbCrLf & 
"**Risk Assessment:**" & vbCrLf & 
"- High-priority risks identified" & vbCrLf & vbCrLf & 
"**Recommended Actions:**" & vbCrLf & 
"- Immediate next steps" & vbCrLf & vbCrLf & 
"Keep under 500 words. Use business language appropriate for board presentation."

Dim executiveSummary As String = Await pdf.Query(executiveQuery)

File.WriteAllText("executive-summary.txt", executiveSummary)
Console.WriteLine("Executive summary saved to executive-summary.txt")

// Analyze contract clauses and identify key terms, risks, and critical dates
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("contract.pdf");

// Define JSON schema for contract analysis
string clauseQuery = @"Analyze this contract and identify key clauses. Return JSON:
{
    ""contractType"": ""string"",
    ""parties"": [""string""],
    ""effectiveDate"": ""string"",
    ""clauses"": [
        {
            ""type"": ""Termination|Liability|Indemnification|Confidentiality|IP|Payment|Warranty|Other"",
            ""title"": ""string"",
            ""summary"": ""string"",
            ""riskLevel"": ""Low|Medium|High"",
            ""keyTerms"": [""string""]
        }
    ],
    ""criticalDates"": [
        {
            ""description"": ""string"",
            ""date"": ""string""
        }
    ],
    ""overallRiskAssessment"": ""Low|Medium|High"",
    ""recommendations"": [""string""]
}

Focus on: termination rights, liability caps, indemnification, IP ownership, confidentiality, payment terms.
Return ONLY valid JSON.";

string analysisJson = await pdf.Query(clauseQuery);

try
{
    var analysis = JsonSerializer.Deserialize<JsonElement>(analysisJson);
    string formatted = JsonSerializer.Serialize(analysis, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Contract Clause Analysis:");
    Console.WriteLine(formatted);

    File.WriteAllText("contract-analysis.json", formatted);

    // Display high-risk clauses
    Console.WriteLine("\n=== High Risk Clauses ===");
    foreach (var clause in analysis.GetProperty("clauses").EnumerateArray())
    {
        if (clause.GetProperty("riskLevel").GetString() == "High")
        {
            Console.WriteLine($"- {clause.GetProperty("type")}: {clause.GetProperty("summary")}");
        }
    }
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse contract analysis");
    File.WriteAllText("contract-analysis-raw.txt", analysisJson);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("contract.pdf")

' Define JSON schema for contract analysis
Dim clauseQuery As String = "Analyze this contract and identify key clauses. Return JSON:
{
    ""contractType"": ""string"",
    ""parties"": [""string""],
    ""effectiveDate"": ""string"",
    ""clauses"": [
        {
            ""type"": ""Termination|Liability|Indemnification|Confidentiality|IP|Payment|Warranty|Other"",
            ""title"": ""string"",
            ""summary"": ""string"",
            ""riskLevel"": ""Low|Medium|High"",
            ""keyTerms"": [""string""]
        }
    ],
    ""criticalDates"": [
        {
            ""description"": ""string"",
            ""date"": ""string""
        }
    ],
    ""overallRiskAssessment"": ""Low|Medium|High"",
    ""recommendations"": [""string""]
}

Focus on: termination rights, liability caps, indemnification, IP ownership, confidentiality, payment terms.
Return ONLY valid JSON."

Dim analysisJson As String = Await pdf.Query(clauseQuery)

Try
    Dim analysis = JsonSerializer.Deserialize(Of JsonElement)(analysisJson)
    Dim formatted As String = JsonSerializer.Serialize(analysis, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Contract Clause Analysis:")
    Console.WriteLine(formatted)

    File.WriteAllText("contract-analysis.json", formatted)

    ' Display high-risk clauses
    Console.WriteLine(vbCrLf & "=== High Risk Clauses ===")
    For Each clause In analysis.GetProperty("clauses").EnumerateArray()
        If clause.GetProperty("riskLevel").GetString() = "High" Then
            Console.WriteLine($"- {clause.GetProperty("type")}: {clause.GetProperty("summary")}")
        End If
    Next
Catch ex As JsonException
    Console.WriteLine("Unable to parse contract analysis")
    File.WriteAllText("contract-analysis-raw.txt", analysisJson)
End Try

// Extract financial metrics from annual reports and earnings documents
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("annual-report.pdf");

// Define JSON schema for financial extraction (numbers in millions)
string financialQuery = @"Extract financial metrics from this document. Return JSON:
{
    ""reportPeriod"": ""string"",
    ""company"": ""string"",
    ""currency"": ""string"",
    ""incomeStatement"": {
        ""revenue"": number,
        ""costOfRevenue"": number,
        ""grossProfit"": number,
        ""operatingExpenses"": number,
        ""operatingIncome"": number,
        ""netIncome"": number,
        ""eps"": number
    },
    ""balanceSheet"": {
        ""totalAssets"": number,
        ""totalLiabilities"": number,
        ""shareholdersEquity"": number,
        ""cash"": number,
        ""totalDebt"": number
    },
    ""keyMetrics"": {
        ""revenueGrowthYoY"": ""string"",
        ""grossMargin"": ""string"",
        ""operatingMargin"": ""string"",
        ""netMargin"": ""string"",
        ""debtToEquity"": number
    },
    ""guidance"": {
        ""nextQuarterRevenue"": ""string"",
        ""fullYearRevenue"": ""string"",
        ""notes"": ""string""
    }
}

Use null for unavailable data. Numbers in millions unless stated.
Return ONLY valid JSON.";

string financialJson = await pdf.Query(financialQuery);

try
{
    var financials = JsonSerializer.Deserialize<JsonElement>(financialJson);
    string formatted = JsonSerializer.Serialize(financials, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Extracted Financial Data:");
    Console.WriteLine(formatted);

    File.WriteAllText("financial-data.json", formatted);
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse financial data");
    File.WriteAllText("financial-raw.txt", financialJson);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("annual-report.pdf")

' Define JSON schema for financial extraction (numbers in millions)
Dim financialQuery As String = "Extract financial metrics from this document. Return JSON:
{
    ""reportPeriod"": ""string"",
    ""company"": ""string"",
    ""currency"": ""string"",
    ""incomeStatement"": {
        ""revenue"": number,
        ""costOfRevenue"": number,
        ""grossProfit"": number,
        ""operatingExpenses"": number,
        ""operatingIncome"": number,
        ""netIncome"": number,
        ""eps"": number
    },
    ""balanceSheet"": {
        ""totalAssets"": number,
        ""totalLiabilities"": number,
        ""shareholdersEquity"": number,
        ""cash"": number,
        ""totalDebt"": number
    },
    ""keyMetrics"": {
        ""revenueGrowthYoY"": ""string"",
        ""grossMargin"": ""string"",
        ""operatingMargin"": ""string"",
        ""netMargin"": ""string"",
        ""debtToEquity"": number
    },
    ""guidance"": {
        ""nextQuarterRevenue"": ""string"",
        ""fullYearRevenue"": ""string"",
        ""notes"": ""string""
    }
}

Use null for unavailable data. Numbers in millions unless stated.
Return ONLY valid JSON."

Dim financialJson As String = Await pdf.Query(financialQuery)

Try
    Dim financials = JsonSerializer.Deserialize(Of JsonElement)(financialJson)
    Dim formatted As String = JsonSerializer.Serialize(financials, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Extracted Financial Data:")
    Console.WriteLine(formatted)

    File.WriteAllText("financial-data.json", formatted)
Catch ex As JsonException
    Console.WriteLine("Unable to parse financial data")
    File.WriteAllText("financial-raw.txt", financialJson)
End Try

// Extract structured research metadata from academic papers
using IronPdf;
using IronPdf.AI;
using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Memory;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using System.Text.Json;

// Azure OpenAI configuration
string azureEndpoint = "https://your-resource.openai.azure.com/";
string apiKey = "your-azure-api-key";
string chatDeployment = "gpt-4o";
string embeddingDeployment = "text-embedding-ada-002";

// Initialize Semantic Kernel
var kernel = Kernel.CreateBuilder()
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey)
    .Build();

var memory = new MemoryBuilder()
    .WithMemoryStore(new VolatileMemoryStore())
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey)
    .Build();

IronDocumentAI.Initialize(kernel, memory);

var pdf = PdfDocument.FromFile("research-paper.pdf");

// Define JSON schema for research paper extraction
string researchQuery = @"Extract structured information from this research paper. Return JSON:
{
    ""title"": ""string"",
    ""authors"": [""string""],
    ""institution"": ""string"",
    ""publicationDate"": ""string"",
    ""abstract"": ""string"",
    ""researchQuestion"": ""string"",
    ""methodology"": {
        ""type"": ""Quantitative|Qualitative|Mixed Methods"",
        ""approach"": ""string"",
        ""sampleSize"": ""string"",
        ""dataCollection"": ""string""
    },
    ""keyFindings"": [
        {
            ""finding"": ""string"",
            ""significance"": ""string"",
            ""confidence"": ""High|Medium|Low""
        }
    ],
    ""limitations"": [""string""],
    ""futureWork"": [""string""],
    ""keywords"": [""string""]
}

Focus on extracting verifiable claims and noting uncertainty.
Return ONLY valid JSON.";

string extractionResult = await pdf.Query(researchQuery);

try
{
    var research = JsonSerializer.Deserialize<JsonElement>(extractionResult);
    string formatted = JsonSerializer.Serialize(research, new JsonSerializerOptions { WriteIndented = true });

    Console.WriteLine("Research Paper Extraction:");
    Console.WriteLine(formatted);

    File.WriteAllText("research-extraction.json", formatted);

    // Display key findings with confidence levels
    Console.WriteLine("\n=== Key Findings ===");
    foreach (var finding in research.GetProperty("keyFindings").EnumerateArray())
    {
        string confidence = finding.GetProperty("confidence").GetString() ?? "Unknown";
        Console.WriteLine($"[{confidence}] {finding.GetProperty("finding")}");
    }
}
catch (JsonException)
{
    Console.WriteLine("Unable to parse research extraction");
    File.WriteAllText("research-raw.txt", extractionResult);
}

Imports IronPdf
Imports IronPdf.AI
Imports Microsoft.SemanticKernel
Imports Microsoft.SemanticKernel.Memory
Imports Microsoft.SemanticKernel.Connectors.OpenAI
Imports System.Text.Json

' Azure OpenAI configuration
Dim azureEndpoint As String = "https://your-resource.openai.azure.com/"
Dim apiKey As String = "your-azure-api-key"
Dim chatDeployment As String = "gpt-4o"
Dim embeddingDeployment As String = "text-embedding-ada-002"

' Initialize Semantic Kernel
Dim kernel = Kernel.CreateBuilder() _
    .AddAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .AddAzureOpenAIChatCompletion(chatDeployment, azureEndpoint, apiKey) _
    .Build()

Dim memory = New MemoryBuilder() _
    .WithMemoryStore(New VolatileMemoryStore()) _
    .WithAzureOpenAITextEmbeddingGeneration(embeddingDeployment, azureEndpoint, apiKey) _
    .Build()

IronDocumentAI.Initialize(kernel, memory)

Dim pdf = PdfDocument.FromFile("research-paper.pdf")

' Define JSON schema for research paper extraction
Dim researchQuery As String = "Extract structured information from this research paper. Return JSON:
{
    ""title"": ""string"",
    ""authors"": [""string""],
    ""institution"": ""string"",
    ""publicationDate"": ""string"",
    ""abstract"": ""string"",
    ""researchQuestion"": ""string"",
    ""methodology"": {
        ""type"": ""Quantitative|Qualitative|Mixed Methods"",
        ""approach"": ""string"",
        ""sampleSize"": ""string"",
        ""dataCollection"": ""string""
    },
    ""keyFindings"": [
        {
            ""finding"": ""string"",
            ""significance"": ""string"",
            ""confidence"": ""High|Medium|Low""
        }
    ],
    ""limitations"": [""string""],
    ""futureWork"": [""string""],
    ""keywords"": [""string""]
}

Focus on extracting verifiable claims and noting uncertainty.
Return ONLY valid JSON."

Dim extractionResult As String = Await pdf.Query(researchQuery)

Try
    Dim research = JsonSerializer.Deserialize(Of JsonElement)(extractionResult)
    Dim formatted As String = JsonSerializer.Serialize(research, New JsonSerializerOptions With {.WriteIndented = True})

    Console.WriteLine("Research Paper Extraction:")
    Console.WriteLine(formatted)

    File.WriteAllText("research-extraction.json", formatted)

    ' Display key findings with confidence levels
    Console.WriteLine(vbCrLf & "=== Key Findings ===")
    For Each finding In research.GetProperty("keyFindings").EnumerateArray()
        Dim confidence As String = finding.GetProperty("confidence").GetString() OrElse "Unknown"
        Console.WriteLine($"[{confidence}] {finding.GetProperty("finding")}")
    Next
Catch ex As JsonException
    Console.WriteLine("Unable to parse research extraction")
    File.WriteAllText("research-raw.txt", extractionResult)
End Try