In this NLP training you learn how text data is prepared for models, how embeddings capture meaning as vectors and how to measure the quality of a text search. It is aimed at developers and data scientists with Python skills. You work with tokenisation, Sentence Transformers and cosine similarity and rate results with metrics. The course is designed for intermediate learners. We agree data examples and scope with you when you enquire.
What you’ll take away.
- Clean, normalise and tokenise text and estimate the effect on model quality
- Explain embeddings as vector representations and compute similarity with cosine distance
- Build a semantic search over a document collection with Sentence Transformers
- Measure search quality with recall at k, MRR and nDCG against a reference set
- Compare and combine classic methods such as TF-IDF and BM25 with embedding search
Your training in detail
- Who is it for?
- Developers, data scientists and search owners who build text analysis, search or classification into products.
- Prerequisites
- Python skills and a basic understanding of machine learning. Examples in German and English are used.
- Format & planning
- Suggested format: 3 days live online or on site, split into modules. Scope depends on prior knowledge.
Agenda: from understanding to application
Our independently developed agenda proposal. Focus areas, lab environment and scope are agreed before booking.
Overview of text-processing tasks: classification, entity recognition, summarisation, search.
Text preparation with cleaning, normalisation, tokenisation, subwords and handling German compounds.
Classic representations with bag of words, TF-IDF and BM25 as a baseline.
Embeddings with Sentence Transformers: vectors, cosine similarity, multilingual models and limits.
Vector search with indexes, chunking of long texts, hybrid search and reranking.
Evaluation with reference queries, recall at k, MRR and nDCG, plus error analysis of individual results.
Suggested exercise: You build a semantic search over 200 help articles, write 30 test queries with expected hits and compare BM25, embedding search and their combination by recall at 5.
Exams, licences & credentials
There is no single authoritative external exam for NLP. Attending alone awards no vendor certificate. Exam fees and licences are only included if the offer confirms them.
Vendor references & review date
Sources checked on 2026-11-03. These sources describe the technology or external exam requirements. They do not imply a partnership with NextNowa.
Your learning content
These objectives describe the training topic. The specific syllabus, duration, dates, trainer and price are agreed before booking.
Frequently asked questions
What is an embedding and what do I need it for?
An embedding is a vector of numbers that represents the meaning of a text. Similar texts sit close together in vector space. This lets you find hits by meaning rather than exact words, for search, duplicate detection or clustering. Quality depends on the model and your data and should be measured.
Does embedding search replace classic keyword search?
Usually not entirely. Keyword methods such as BM25 are strong on exact terms, product numbers and names. Embeddings find paraphrases and synonyms. A hybrid search often delivers the best results. In the course you measure this on your own test set instead of relying on assumptions.
Does NLP also work well with German text?
Yes, if you choose suitable models. Multilingual models cover German well but have quirks with compounds and technical terms. That is why you test quality on German examples from your own field. In the course we use German and English texts and test exactly that.
Does Natural language processing include a certificate?
There is no single authoritative external exam for NLP. Attending alone awards no vendor certificate. Exam fees and licences are only included if the offer confirms them.