Tekstsummering

Tekstsummering, or text summarization, is the process of producing a concise version of a text that preserves its central ideas and key information. Summaries help readers understand the gist without reading the full document, and they support information retrieval, decision making, and accessibility across domains. The field covers summaries written in the same language as the source and multilingual cases.

Summaries are typically categorized as extractive or abstractive. Extractive methods select existing sentences or passages from the source, while abstractive methods generate new wording that paraphrases the content. Approaches range from rule-based and statistical techniques to modern neural models, including sequence-to-sequence architectures and transformer-based systems. Evaluation commonly uses ROUGE scores that compare generated summaries to human references.

Applications span journalism, research, business intelligence, legal and governmental work, education, and digital assistants. In practice, abstractive systems are valued for brevity and readability, while extractive approaches are praised for factual reliability, though both face challenges in preserving nuance.

Historically, early summarizers relied on hand-crafted rules; progress progressed with machine learning and, in the last decade, large pretrained language models. Current research focuses on multilingual and domain-specific summarization, improving factual accuracy, and assessing summaries with standardized benchmarks. Ethical and bias considerations are ongoing concerns.