The knowledge-intensive industry has evolved significantly in recent years, owing to technological advancements and a growing demand for data-driven insights. As the volume of media data expands, companies must adopt innovative methods to analyse and interpret media intelligence data in a timely and accurate manner.

Natural Language Generation (NLG) is one such advancement.

What is NLG and why does it matter in media intelligence?

A subset of artificial intelligence (AI), NLG uses machine learning algorithms to process information and transform data into a more human-like response. This process allows analysts and businesses to generate insights and summaries from massive datasets within minutes.

NLG is often viewed as the natural progression from Natural Language Processing (NLP). While NLP enables machines to understand and interpret language, NLG allows them to generate text from data sources. This can support applications such as chatbots, automated reporting, email generation, and more.

Potential applications of NLG in the industry

Media intelligence companies can harness NLG in several impactful ways:

Automated Abstractive SummarisationTurn long articles into concise, high-quality summaries, helping users quickly grasp key points without sacrificing meaning.
Sentiment Analysis & Topic ClassificationNLP tools help categorise content and assess tone, which NLG can then present in clear, digestible formats.
Entity ExtractionIdentify and contextualise key names, organisations, locations, and terms to enrich reporting.
Insights GenerationAutomate the production of daily, weekly, or campaign-specific reports.

These technologies allow companies to create consistent, scalable, and data-rich narratives that offer real value to clients.

The evolution of NLG: From rule-based to GPT models

NLG technology has been around since the early 2000s when it was used in basic rule-based and conversational systems. Those were the times when search engines like Google started indexing news sites on the web. Since then, the technology has evolved to advanced systems powered by neural networks and deep learning.

This shift enabled the rise of large language models (LLMs) such as GPT-4o and ChatGPT, which can generate increasingly accurate and coherent language at scale. The integration of transfer learning has made it easier to fine-tune these large language models for domain-specific applications, such as media intelligence.

NLG solutions’ applications span journalism, marketing, customer support, and content creation, and its use in media intelligence results in increased efficiency, cost savings, and real-time insights.

However, this evolution has also brought in new challenges – including the need for high-quality training data, mitigation of biases, and privacy safeguards.

Why custom LLMs matter more than ever

Generic pre-trained language models, while powerful, may fall short in accuracy, relevance, and compliance with industry-specific demands. Media intelligence organisations often deal with sensitive data, nuanced topics, and complex language.

Custom language models tailored to specific tasks and incorporating domain knowledge can provide more accurate insights into media intelligence. These models can also address data privacy challenges and mitigate biases in pre-trained models.

More specifically, custom LLMs can:

  • Improve accuracy by fine-tuning LLMs on relevant, high-quality datasets
  • Reduce bias and enhance fairness
  • Meet stringent data privacy requirements
  • Reflect domain-specific language and terminology

The need for transparency and trust in AI-driven media analysis makes custom solutions a critical differentiator.

Training and fine-tuning of large language models can be achieved using a large corpus of text documents (such as news articles), where each document contains multiple words.

Not a long ago, at FIBEP Tech Day 2023, held in Rome on April 21st, Nesin Veli (CEO at Identrics) presented on the future of NLG in the context of ChatGPT.

His talk explored why everybody talks about LLMs today, as he explored the transformation of NLG from rule-based logic to deep learning systems and the implications for media intelligence. He also emphasised how advanced LLM models like ChatGPT are already enhancing the media intelligence industry.

However, Nesin also examined the risks associated with generic LLMs – including that pre-built models may not be able to deliver the required level of precision in the absence of such curated data, particularly in the case of industry-specific terminologies and jargons.

Furthermore, as media intelligence companies handle confidential and sensitive data, Nesin proposed solutions involving the use of domain-specific datasets with high-quality, diverse, and well-labelled data in building custom language models.

Knowledge bases play a crucial role in media intelligence. They serve as a vital resource for tracking, analysing, structuring and contextualising massive amounts of data, enabling businesses to make smarter data-driven decisions and effectively adapt their strategies to stay ahead of the competition.

A comprehensive and well-maintained knowledge base not only improves productivity by saving time and resources, but also enhances decision-making, problem-solving, and collaboration.

When integrated with NLG and LLMs, knowledge bases:

  • Support more accurate, contextually-aware language generation
  • Improve collaboration and efficiency across teams
  • Enhance user experience with clearer and more relevant content

By providing easy access to accurate and up-to-date information, knowledge bases empower employees, customers, and stakeholders alike to make informed choices, leading to improved customer satisfaction, reduced support costs, and increased innovation.

To sum it up, in the age of AI, a robust knowledge base acts as the foundation for building intelligent systems that understand both language and context.

How Identrics supports this NLG transformation

Identrics offers a range of NLG and NLP services designed to help organisations automate their processes and extract more value from text data. 

Highlights include:

Custom-made automated abstractive summarisation. Automated abstractive summarisation enables you to read and comprehend lengthy texts efficiently by producing condensed, high-quality summaries. With custom-made solutions, you can enjoy the benefits of automated summarisation without sacrificing the accuracy and relevance of the output.

Our technologies can offer a range of content enrichment tools such as Sentiment Analysis, Entity Extraction, Organic Document Grouping, Topic Modelling, and more.

These solutions can be combined and used as components of various automated reports, allowing businesses to gain valuable insights from their text data.

Let’s talk about the future of media intelligence

The media intelligence landscape is changing fast, and AI-powered technologies like NLG are leading the way. With the right strategies and tools – including custom LLMs, robust NLP systems, and structured knowledge – companies can stay competitive, compliant, and customer-focused.


Contact Identrics today to explore how our AI and language technologies can transform your media workflows.