SLTES Open cheat sheet

Register has three common axes: formality (intimate to frozen), field (specialized domain), and mode (spoken vs. written); a shift on any axis requires a lexical shift in the target text.

Cohesion is built by reference, substitution, ellipsis, conjunction, and lexical reiteration; cohesion is surface text-linking while coherence is the reader's logical reconstruction of meaning.

Grice's maxims - quantity, quality, relation, manner - describe cooperative communication; implicature is meaning conveyed by flouting a maxim rather than by literal content.

Speech acts have locutionary (utterance), illocutionary (intended force), and perlocutionary (effect on hearer) dimensions; directives and commissives carry the strongest cross-language force mismatches.

Theme is the clause-initial element (what the message is about); rheme is the remainder; English and Chinese differ in tolerance for marked themes, which affects word order in translation.

Given/new information ordering: end-weight and end-focus place new or heavy information late in the English clause, a pattern many SOV and topic-prominent languages do not share.

Politeness strategies split into negative (avoiding imposition) and positive (claiming closeness); languages encode these with different grammatical devices such as honorifics, modals, or particles.

False fluency traps: cognate deception, collocational mismatch, and idiom opacity; a bilingual who understands a text may still fail to produce acceptable collocations under time pressure.

Nida's formal equivalence preserves surface structure and content; dynamic equivalence targets equivalent effect on the receptor, prioritizing naturalness over structural mirroring.

Vinay and Darbelnet's seven procedures: borrowing, calque, literal translation (direct); transposition, modulation, equivalence, adaptation (oblique); apply direct procedures first, oblique when literal fails.

Skopos theory (Vermeer): the purpose of the target text governs strategy; adequacy to the brief outranks fidelity to the source, and the brief should specify function, audience, and medium.

House's model distinguishes overt translation (source-culture-bound, reader knows it is a translation) from covert translation (functions as an original in the target culture).

Chesterman's translation strategies: syntactic (literal, transposition, unit shift), semantic (synonymy, hyponymy, converse, emphasis change), and pragmatic (cultural filtering, explicitation, compensation).

Explicitation (adding implied information) is a documented universal tendency in translation; implicitation is its inverse and is riskier because it assumes shared knowledge.

Translation brief elements to confirm before starting: function, target audience, register, terminology source, deliverable format, deadline, and revision policy.

Back-translation is a validity check for questionnaires and legal wording, not a production method; it exposes ambiguity but does not guarantee equivalence.

Unit-of-translation choice (word, phrase, clause, sentence, paragraph) should be driven by the smallest segment that preserves meaning, not by sentence boundaries alone.

Consecutive interpreting with note-taking typically runs in segments of 1-5 minutes; the interpreter speaks after the source speaker pauses, so total delivery time roughly doubles.

Note-taking conventions: diagonal layout, one idea per line, symbols and abbreviations over words, arrows for direction and verb tense, and vertical brackets to mark opening and closing of ideas.

Simultaneous interpreting has an ear-voice span (décalage) of roughly 2-6 seconds; the interpreter must hold source input, render output, and monitor their own delivery concurrently.

Simultaneous interpreting is performed in soundproof booths with relay for languages not covered directly; whispered (chuchotage) and sight translation are distinct modes with different cognitive loads.

Working-memory tactics: chunking, syntactic anticipation, and segmentation at clause boundaries; interpreters should avoid holding more than one clause of unfinished source structure.

Error and repair: self-corrections should be brief and lexical, not full restarts; omitting a redundant list item is preferable to inventing content under load.

Stress and fatigue management: professional simultaneous teams rotate roughly every 20-30 minutes; fatigue degrades lexical retrieval before it degrades grammar.

Numbers, names, and units are the highest-risk elements; interpreters write them down at the moment of hearing rather than relying on memory.

High-context cultures (much meaning in shared situation, indirect style) versus low-context cultures (explicit, text-bound style) predicts how much must be added or removed in translation.

Power distance, individualism-collectivism, uncertainty avoidance, and long-term orientation are Hofstede's dimensions; they inform register and politeness choices, not individual judgments.

Taboo and sensitivity screening: religious references, political entities, gendered forms, body and death terms, and alcohol or dietary references vary by market and must be flagged before delivery.

Culture-specific items (realia) have four common renderings: transference (loan), cultural substitution, descriptive paraphrase, and omission; each has a different reader cost.

Date, number, and name conventions differ: day-month-year versus month-day-year, decimal comma versus point, and family-name order in East Asian languages must be normalized to the target locale.

Idioms and metaphors rarely transfer literally; map to a target idiom with the same function, or paraphrase the sense, and reserve literal rendering for cases where the image is the point.

Nonverbal and paralinguistic cues (gesture, silence, turn-taking norms) carry meaning in dialogue interpreting and may require an explicit verbal gloss for the other party.

Localization versus translation: localization adapts content, format, and function to a locale and may require re-designing layout, currency, and legal notices, not just words.

Core interpreter and translator ethics: accuracy and completeness, confidentiality, impartiality, conflict-of-interest disclosure, and competence boundaries (decline work outside expertise).

Confidentiality survives the end of the assignment; it covers notes, recordings, glossaries, and drafts, and should be governed by a written agreement or NDA.

Impartiality does not mean neutrality about accuracy: an interpreter must correct errors that change meaning, and must not soften, strengthen, or summarize a speaker's position.

Role boundaries in community interpreting: the interpreter is not an advocate, cultural broker beyond explanation, or advisor; any cultural clarification should be brief and labeled as such.

Conflict-of-interest situations include prior personal relationships with parties, financial interest in the outcome, and prior involvement as investigator or witness; disclose and withdraw if material.

Competence and continuing development: standards bodies and associations expect ongoing training, peer review, and adherence to a published code of conduct; certification renewal often requires documented hours.

Errors and complaints: preserve contemporaneous records, report material errors to the client promptly, and follow the contract's remedy clause rather than informal settlement.

Refusal and withdrawal grounds: safety risk, unlawful instructions, inability to interpret competently, or a client demanding a biased rendering; document the reason and hand over cleanly.

Translation memory (TM) matches are graded: exact (100%), fuzzy (typically 75-99%), and context (in-context exact); fuzzy matches require editing, so leverage rates overstate real savings.

Terminology databases store concept-oriented entries with term, definition, part of speech, subject field, source, status, and context; one concept may have several terms per language.

CAT tool segmentation splits on sentence boundaries and paragraph marks; segmentation errors are a leading cause of poor matches and should be fixed before translation begins.

Machine translation post-editing has two levels: light (make understandable, fix critical errors) and full (publishable quality, style alignment); the brief must specify which.

Quality metrics such as MQM classify errors by type (accuracy, fluency, terminology, style) and severity (critical, major, minor), enabling comparable scoring across vendors.

Standards to know: ISO 17100 for translation services, ISO 18587 for machine translation post-editing, and ISO 13611 for community interpreting; each defines process requirements.

Term extraction: candidate terms are identified by frequency, collocation strength, and domain specificity; human validation is required because frequency alone produces noise.

File and encoding hygiene: UTF-8 as default, consistent tag handling in XLIFF/HTML, and locked segments for placeholders prevent broken markup in delivered files.