Class Categorization

Namespace: LMKit.TextAnalysis

Assembly: LM-Kit.NET.dll

Provides functionality to classify content into predefined categories using a language model.

public sealed class Categorization

Inheritance: object

Categorization

Inherited Members: object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.ReferenceEquals(object, object)

object.ToString()

Examples

Example: Basic text classification

using LMKit.Model;
using LMKit.TextAnalysis;
using System;
using System.Collections.Generic;

// Load the language model
LM model = LM.LoadFromModelID("lmkit-tasks:4b-preview");

// Create the categorization engine
Categorization categorizer = new Categorization(model);

// Define categories
var categories = new List<string> { "sports", "technology", "politics", "entertainment" };

// Classify text
string article = "The team won the championship in overtime.";
int bestIndex = categorizer.GetBestCategory(categories, article);

Console.WriteLine($"Category: {categories[bestIndex]}");
Console.WriteLine($"Confidence: {categorizer.Confidence:P1}");

Example: Classification with category descriptions

using LMKit.Model;
using LMKit.TextAnalysis;
using System;
using System.Collections.Generic;

LM model = LM.LoadFromModelID("lmkit-tasks:4b-preview");
Categorization categorizer = new Categorization(model);

// Categories with descriptions for better accuracy
var categories = new List<string> { "urgent", "normal", "low" };
var descriptions = new List<string>
{
    "Critical issues requiring immediate attention",
    "Standard requests with normal priority",
    "Minor issues that can wait"
};

string ticket = "Server is down, customers cannot access the website!";
int priority = categorizer.GetBestCategory(categories, descriptions, ticket);

Console.WriteLine($"Priority: {categories[priority]}");

Remarks

This engine supports categorization tasks for various content types, including plain text, images, PDF documents, HTML files, and Microsoft Office formats (DOCX, XLSX, PPTX).

For a complete list of supported file formats, see the Attachment class documentation.

Key Features

Single-category classification with GetBestCategory(IList<string>, string, bool, CancellationToken)
Multi-category classification with GetTopCategories(IList<string>, string, int, bool, CancellationToken)
Support for category descriptions to improve accuracy
Confidence scoring via Confidence property
Optional embedding-based classification for performance
Multimodal support for text, images, and documents

Constructors

Categorization(LM): Initializes a new instance of the Categorization class using the specified language model.

Properties

AllowUnknownCategory: Gets or sets a value indicating whether the categorization engine can identify when content does not match any of the predefined categories. If set to true, the engine returns -1 (for single-category methods) or an empty list (for multi-category methods) when no suitable category is identified. If set to false, the engine always assigns the content to the nearest matching category, even if confidence is low.

Confidence: Gets the confidence level of the most recent categorization operation.

Guidance: Gets or sets optional guidance text that can influence the categorization process. This can be used to steer the model toward certain themes or constraints.

MaxInputTokens: Gets or sets the maximum number of input tokens that the categorization engine will analyze.

Model: Gets the underlying language model instance associated with this categorization object.

PreferredInferenceModality: Gets or sets the preferred modality for inference. This determines whether text, image, or both modalities are used when processing input. Defaults to Multimodal.

UseEmbeddingClassifier: Gets or sets a value indicating whether the classifier should use an embedding-based classification strategy. When set to true, the engine is forced to classify using embeddings instead of a completion-based approach.

Methods

GetBestCategory(IList<string>, Attachment, bool, CancellationToken): Classifies the content provided as an attachment into one of the predefined categories.

GetBestCategory(IList<string>, ImageBuffer, bool, CancellationToken): Classifies the content provided in an image into one of the predefined categories. The underlying language model must support vision input.

GetBestCategory(IList<string>, IList<string>, Attachment, bool, CancellationToken): Classifies the content provided as an attachment into one of the predefined categories.

GetBestCategory(IList<string>, IList<string>, string, bool, CancellationToken): Classifies the specified plain text into one of the predefined categories.

GetBestCategory(IList<string>, string, bool, CancellationToken): Classifies the specified plain text into one of the predefined categories.

GetBestCategoryAsync(IList<string>, Attachment, bool, CancellationToken): Asynchronously classifies the content provided as an attachment into one of the predefined categories.

GetBestCategoryAsync(IList<string>, IList<string>, Attachment, bool, CancellationToken): Asynchronously classifies the content provided as an attachment into one of the predefined categories.

GetBestCategoryAsync(IList<string>, IList<string>, ImageBuffer, bool, CancellationToken): Asynchronously classifies the given image into one of the predefined categories. The underlying language model must support vision input.

GetBestCategoryAsync(IList<string>, IList<string>, string, bool, CancellationToken): Asynchronously classifies the specified plain text into one of the predefined categories.

GetBestCategoryAsync(IList<string>, string, bool, CancellationToken): Asynchronously classifies the specified plain text into one of the predefined categories.

GetTopCategories(IList<string>, Attachment, int, bool, CancellationToken): Classifies the content provided as an attachment into up to a specified maximum number of categories.

GetTopCategories(IList<string>, ImageBuffer, int, bool, CancellationToken): Classifies the given image into up to a specified maximum number of categories. The underlying language model must support vision input.

GetTopCategories(IList<string>, IList<string>, Attachment, int, bool, CancellationToken): Classifies the content provided as an attachment into up to a specified maximum number of categories.

GetTopCategories(IList<string>, IList<string>, string, int, bool, CancellationToken): Classifies the specified plain text into up to a specified maximum number of categories.

GetTopCategories(IList<string>, string, int, bool, CancellationToken): Classifies the specified plain text into up to a specified maximum number of categories.

GetTopCategoriesAsync(IList<string>, Attachment, int, bool, CancellationToken): Asynchronously classifies the content provided as an attachment into up to a specified maximum number of categories.

GetTopCategoriesAsync(IList<string>, ImageBuffer, int, bool, CancellationToken): Asynchronously classifies the given image into up to a specified maximum number of categories. The underlying language model must support vision input.

GetTopCategoriesAsync(IList<string>, IList<string>, Attachment, int, bool, CancellationToken): Asynchronously classifies the content provided as an attachment into up to a specified maximum number of categories.

GetTopCategoriesAsync(IList<string>, IList<string>, string, int, bool, CancellationToken): Asynchronously classifies the specified plain text into up to a specified maximum number of categories.