| { |
| "paper_id": "Y14-1004", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T13:44:08.863676Z" |
| }, |
| "title": "Discourse for Machine Translation", |
| "authors": [ |
| { |
| "first": "Bonnie", |
| "middle": [], |
| "last": "Webber", |
| "suffix": "", |
| "affiliation": { |
| "laboratory": "", |
| "institution": "University of Edinburgh", |
| "location": {} |
| }, |
| "email": "" |
| } |
| ], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "Statistical Machine Translation is a modern success: Given a source language sentence, SMT finds the most probable target language sentence, based on (1) properties of the source; (2) probabilistic source-target mappings at the level of words, phrases and/or sub-structures; and (3) properties of the target language.", |
| "pdf_parse": { |
| "paper_id": "Y14-1004", |
| "_pdf_hash": "", |
| "abstract": [ |
| { |
| "text": "Statistical Machine Translation is a modern success: Given a source language sentence, SMT finds the most probable target language sentence, based on (1) properties of the source; (2) probabilistic source-target mappings at the level of words, phrases and/or sub-structures; and (3) properties of the target language.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "Abstract", |
| "sec_num": null |
| } |
| ], |
| "body_text": [ |
| { |
| "text": "SMT translates individual sentences because the search space even for a single sentence can be vast. But sentences are parts of texts, and texts have properties beyond those of their individual sentences, including:", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "\u2022 document-wide properties, such as style, register, reading level and genre, that are visible in the frequency and distribution of words, word senses, referential forms and syntactic structures;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "\u2022 patterns of topical or functional sub-structures that mean that frequencies and distributions of words, word senses, referential forms and syntactic structures will vary across a text;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "\u2022 relations between clauses or between referring expressions that can be signaled explicitly or implicitly, that reflect a text's coherence;", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "\u2022 frequent appeal to reduced expressions that rely on context to \u2022 efficiently convey their message.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Recognizing and deploying these properties promises to improve both fluency and accuracy in SMT --i.e., whether the sequence of sentences in the target text conveys the same information as those in its source, in as readable a manner. This presentation describes how researchers are attempting to do this, without bringing translation to a halt.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": {}, |
| "ref_entries": {} |
| } |
| } |
