+49 (0)211 22 06 77 0

 

Glossary

  • Alignment is a text order process. Sentences in a source text are assigned to the corresponding sentences in the translation of the text. An alignment must be performed whenever text is in an unsegmented body text. The result is then transferred to a translation memory or used as the basis(...)
  • CAT tool (computer-assisted/computer-aided translation) refers to all computer-aided auxiliary systems of the translator, i.e. dictionaries, reference texts, translation memories, etc. Today, CAT tools are often used synonymously with translation memory system, a system for managing and(...)
  • A compound is a word, usually consisting of two or more nouns. The most common group in German is the determinative compound (e.g German Erd-ball, Erd-apfel, Erd-umrundung). In addition to the pure noun compounds, there are also adjective compounds (e.g. German "schwarz-weiß"), and(...)
  • A corpus is a digitized collection of text data that is used for computer-aided processing of natural language. For example, machine translation uses aligned, bilingual parallel corpora as the basis for statistical analysis and MT engine training.
  • Error metrics measure in the quality evaluation (QE) whether the previously agreed language quality has been achieved. There are various metrics (MQM, DQF, SAEJ2450, LISA). All metrics provide error criteria, error points, and severity levels that should be calibrated in advance. In addition,(...)
  • Internationalization deals with the formal localization of a software or website to the conditions and specifications of the target culture/country. This is primarily about numbers, data, formats, units, to prepare and present calendars in line with the target culture.
  • Language Quality Assurance is a preventive measure to avoid problems and errors in text creation and translation. There are a large number of tools that are used for language quality assurance, so-called authoring tools or quality assurance tools (QA tools). Language quality assurance is(...)
  • During localization, an editor adapts a text and its components to the language and cultural specifics of a target country. This customization is often done as part of a translation. The result of localization is a product (such as text, software, website), tailored to the target market and(...)
  • A MT engine (MT = machine translation) is a system-bound model for automatically translating a language into another language in a one-way manner. The methods are either rule-based (RBMT), statistical (SMT) or neural (NMT). In order to train statistical and neural MT engines, large amounts of(...)
  • Named entity recognition (NER) refers to methods for recognizing language units in natural language processing. NER methods are used to identify and mark information such as names, locations and products in text corpora. Recent developments in the field of artificial intelligence (AI) make(...)
  • A naming is a word or several words used to designate a term (DIN ISO 26162:2016-12). In a term-based terminology database, for example, several namings are stored per term if required. As a rule, a preferred naming should always be specified, which is used uniformly (e.g. "passenger car").(...)
  • Natural language processing (NLP) is the computer-aided processing of natural language. For example, data cleansing for digital language processing or machine translation using segmentation, tokenization, text statistics, anonymization and named entity recognition (NER). NLP is an important(...)
  • Neural machine translation (also known as NMT) is a method of machine translation performed with recurrent neural networks (RNN). An encoder encrypts sequences of words in the source text in vectors. A decoder then decrypts these vectors and decodes the target text word by word. By taking(...)
  • The acronym NMT stands for neural machine translation and is a method of machine translation performed with recurrent neural networks (RNN). An encoder encrypts sequences of words in the source text in vectors. A decoder then decrypts these vectors and decodes the target text word by(...)
  • An ontology is a recognized, formal modeling of knowledge about a domain (see also DIN EN 62656-5 and VDE 0040-8-5:2018-05). It is represented in the form of ontology languages that map classes, objects, relations, and attributes. The essence of an ontology is to make the hierarchical and(...)
  • Post editing (PE) is the activity of editing and correcting a machine-generated translation (ISO standard PE 18587). A distinction is made between light and full post editing. The aim of light post editing is to obtain a comprehensible text. The aim of full post editing is to obtain a product(...)
  • Quality assurance tools ensure language quality when creating text (e.g. Congree), translation (e.g. QA Distiller) and the maintenance of language resources, such as terminology (e.g. quickTerm). This will detect and correct errors before publishing text or using resources. The specific use(...)
  • A quality evaluation (QE) is used to assess the achievement of a previously defined translation quality. It is possible to determine whether texts and translations achieve these quality objectives. Various error metrics are used. In machine translation, human quality evaluations (also known(...)
  • A quality measurement (QM) is a measure for the ongoing assessment of the language quality. This is particularly important when using machine translation. Here, the speeds of post editing are measured using productivity measurement methods (e.g. time-to-edit, see also TAUS, DQF). So you know(...)
  • The acronym SMT stands for statistical machine translation and is a machine translation (MT) method. SMT is based on the calculation of probabilities for phrasal translations (translation model) and word sequences (language model). To achieve good results, SMT is trained with large bilingual(...)