So if a user expects that certain source patterns (e.g., complex verbal constructions involving auxiliaries and clitics such as “ne les auraient pas vues”) are better translated by a rule-based system than by the given phrase-based SMT system, then a rule-based translation “would not have seen them” could be inserted in the bi-phrase graph as the dynamic bi-phrase (ne les auraient pas vues, would not have seen them) along with some features indicating that the translation came from this specific system. Dynamic: having active strength of body or mind. In one embodiment, this problem is addressed by artificially constructing the development corpus 64 so as to contain several instances of the dynamic bi-phrase types whose λ weights are to be estimated. After providing the static bi-phrase table, for a sentence of the source text, the method includes applying at least rule which, when fired, generates at least one respective dynamic bi-phrase based on the source text in the sentence, associating a respective value of at least one dynamic feature with the at least one dynamic bi-phrase, retrieving static bi-phrases from the static bi-phrase table which each cover at least one word of the source text, combining the retrieved static bi-phrases and any generated dynamic bi-phrases to form a set of active bi-phrases, generating translation hypotheses for at least a part of the source sentence using active bi-phrases from the set, scoring the translation hypotheses with a translation scoring model which takes into account the static feature values of static bi-phrases in the hypothesis and dynamic feature values of dynamic-bi-phrases in the hypothesis, and outputting a translation of the source text sentence based on the scoring of the hypotheses. Copyright © National Institute of Information and Communications Technology. At S104, each input description is converted into a separate (pattern-based) rule for later generation of dynamic bi-phrases based on input text. Also called an action verb or an event verb. ThoughtCo uses cookies to provide you with a great user experience. The purpose of the training is to assign weights which will provide the best quality of translation which can be obtained, using the available bi-phrases. Instead the decoding utilizes dynamic bi-phrases which are defined after the generation of the static bi-phrase table, e.g., at run-time, on the basis of the given source sentence. Features are associated with the dynamic bi-phrase which allow it to compete with other (e.g., static) bi-phrases during scoring, without requiring that it will be favored over static bi-phrases. It will determine whether to use or not use it on the basis of the log-linear score that is computed for the different translation candidates. No.
The same static bi-phrase table is then accessed for all translations of text between the source and target languages. One or more dynamic bi-phrases are generated, based on the source text and the application of one or more rules, which may be based on user descriptions. Data memory 16, integral with or linked to the computer 12, stores a static bi-phrase table 32, which has been generated automatically from a bilingual corpus 34 of sentence pairs, each pair including a sentence in the source language and a sentence in the target language which have been determined, through automated means, to be likely translations of each other. When words/phrases are input from a PC (S31: YES), candidate words and phrases corresponding to the input words/phrases and the feeling level are retrieved from a word/phrase candidate database 131 (S40), and transmitted Moreover, even if the bi-phrase (degrés Celsius, degrees Fahrenheit) were to have a frequency in the training corpus, use of this bi-phrase would lead to an incorrect translation since 40.4 degrees Fahrenheit is not a correct translation of 40.4 degrés Celsius. Input 20 or 22 may also receive dynamic bi-phrase descriptions 44 input by a user, which may be stored in data memory 16 and used in the generation of rules from which dynamic bi-phrases can be generated at translation time to be stored, at least temporarily, in a dynamic bi-phrase table 46. For example, for static bi-phrases, the values of the dynamic features are all 0, while conversely, for all dynamic bi-phrases, the values of the static features are all 0. A dynamic verb (such as run, ride, grow, throw) is primarily used to indicate an action, process, or sensation. Although the translation rule/pattern may be available to the system prior to the input of the sentence to be translated, the actual bi-phrase, which is the result of applying the rule to the pattern (a specific temperature value followed by degrés Celsius) used in the sentence, cannot be determined until the sentence to be translated is input, since the value 40.4 is not known. . In other aspects, a methodology for associating features with the dynamic bi-phrases and for training them is provided.