The Translation Memory (TM) actually allows storing and retrieving the aligned multilingual text segments against a number of search conditions. The TM systems differ with the information stored with the raw text and the retrieval methods.
The functions of Translation Memory can be mainly bifurcated into two main categories, off-line and online functions. The offline functions comprise of analysis, import and export while the online function includes the in-translation functionality. It is just like functioning of the database into its management and the database.
The offline functions mainly include import, analysis and export detailed as under.
- Import: the import function transfers a text and its translation from the text file to the Translation Memory. The Raw Format data where the external source is available along with its translation is imported in TM. This sort of importing of data requires preprocessing of the texts by the users that are outside the system and also interactive editing of text inside the system. The Checklists involved in this case is the input formats, preprocessing and post-editing. The Native Format is also used to import in the TM and the memories of translation are saved in files. This format retains the segmentation, alignment and the control information. Its checklist includes the kind of native format, the format that is used in the TM Combinatorics.
- Analysis: It is the processing of the multilingual text before the import to the TM or the monolingual source text before its submission to the translation, while defining the input and output in each case. It involves the parsing of the source and target texts, the steps involved in the process are noted below.
- Textual Parsing: the recognition of the punctuation is very important to distinguish between the punctuation marks. The markup makes up some sort of pre-editing, which is contained in the aid programs of the translator. The differing formatting conventions and mark-up languages are the symbols of a good analysis. The other special text elements, like codes, names, numbers, dates, currencies, need not to be translated.
- Linguistic Parsing: the list of words and text is prepared by base form reduction also known as lemmatization for retrieval of the terms from the term bank. The phraseology recognition is used for extraction of multi-word terms from the source text. The linguistic parsing is required for normalization of word order variation of phraseology.
- Segmentation: the purpose of segmentation is to choose the useful translation unit. It is a type of parsing, which is done monolingually by using the superficial parsing. The checklist involved in the segmentation process is the input format, the segment definition, output format and possibility of user intervention. The translation of one sentence depends upon the surrounded translations; however the translator proceeds sentence by sentence in the translation process.
- Alignment: The translation correspondence between the source and the target text is defined in the Alignment. It provides feedback to segmentation and ideally it corrects the initial segmentation.
- Term Extraction: it is the automatic stage of analysis, also known as vocabulary. It also estimates the amount of work involved in the job of translation, commonly referred as the text statistics.
- Export: It involves the transfer of text from the memory to the external text file.
The actual purpose of the TM is to provide the best possible choice of translation to the user. The online functioning of TM includes the below-mentioned pre-requisites;
- The source and the target text of the response must be shown.
- The differences and the identities of the query and the response must be indicated.
- The alternative of responses must be evaluated against the preferences.
- The response must be easily inserted into the translation.
However, the functionality of the on-line Translation memory is;
- Retrieval of different types of matches. Exact Match is appeared when the source and stored segment is matched, character by character. Fuzzy Match ranging from 0-100% appears where the exact match is not found.
- Updating of the database of TM is done when the translator has accepted the new translation. It can be modified by deleting and changing the entries available in the TM. In some cases, the multiple translations are saved for one segment.
- Automatic Translation is the automatic retrieval and substitution in TM tools. The results are automatically displayed when a translator is moving through the document. While the automatic substitution is done if the exact match is found.
- Networking allows the group of translators to come together for faster translation of the text by making available the phrases and sentences of one translator to the other. However, the sharing of the TM opens up more opportunities and options.