Patent Application
20130297284
This application claims the benefit of Korean Patent Application No. 10-2012-0046138, filed on May 2, 2012, which is hereby incorporated by reference in its entirety into this application.
1. Technical Field
The present invention relates generally to an apparatus and method for generating polite expressions for automatic translation and, more particularly, to an apparatus and method for generating polite expressions for automatic translation that are capable of overcoming the linguistic and cultural differences between polite expressions.
2. Description of the Related Art
Currently, with the popularization of the Internet, an information channel has been established across the world, yet the communication channel is still blocked because of language barriers between countries. For this reason, in reality, the Internet is segmented into intranets by the borders of respective countries.
Conventional English-to-Korean translation technology is dependent on automatic English-to-Korean translation technology. Automatic translation technology is technology that translates text in a foreign language into text in another language such as the native language using, for example, the processing capability of a computer system. In automatic translation technology, a database that models the knowledge of languages is established, and a translation engine performs translation while referring to this database. However, it is difficult to achieve an automatic translation above a predetermined quality while reliably reflecting the intention of the original text because of the polysemy of natural language. Furthermore, the problems and limitations of the conventional automatic translation technology are as follows:
First, the conventional rule-based automatic translation approach cannot achieve complete translation because natural languages have many exceptions such as the polite expressions of Korean. This defect always generates a single type of polite expression regardless of the interlocutors, thereby making automatic translation results unnatural and confusing the context of a conversation. For example, even if automatic translation were to generate a different type of polite expression for a different interlocutor, the conventional automatic translation technology has the problem of always generating a single type of polite expression.
Second, the conventional automatic translation technology has the problem of having no countermeasures against input that does not comply with rules, such as typographical errors and ungrammatical sentences.
In order to overcome the problems of the conventional simple automatic translation technology, Korean Patent No. 265548 discloses an automatic translation method and apparatus that provide a plurality of translation environments and enable a translation environment related to a genre, field, and purpose to which the original text pertains to be automatically selected. However, this invention simply attains the level of determining the field to which a sentence belongs, selecting a translation environment suitable for the field and translating the sentences in the selected environment rather than improving a translation apparatus and method themselves.
Furthermore, Korean Patent Application No. 1997-37040 discloses an idiom translation method for an automatic English-to-Korean translation system. This invention relates to a method of dividing each sentence into words, performing translation using morpheme interpretation, and processing normal idiomatic expressions while referring to an idiom dictionary. However, this invention extends the translation capability from word translation to idiom translation, but does not extent the capability to overall sentence translation.
As described above, the conventional automatic English-to-Korean translation technologies have the advantage of improving the general translation level, but are limited in their ability to translate Korean polite expressions.
Accordingly, the present invention has been made keeping in mind the above problems occurring in the prior art, and an object of the present invention is to provide an apparatus and method for generating polite expressions for automatic translation, which decide the social relationship with an interlocutor based on non-linguistic information such as personal information provided by a social network, and linguistic information such as conversational expressions, when translating a conversation between first and second language users into a target language, thereby overcoming linguistic and cultural differences between polite expressions and then providing automatic translation results.
Another object of the present invention is to provide an apparatus and method for generating polite expressions for automatic translation, which decide the social relationship with an interlocutor and select a polite level for a conversation between first and second language users, thereby providing automatic translation results suitable for the interlocutor.
In order to accomplish the above object, the present invention provides an apparatus for generating polite expressions for automatic translation, including a relationship recognition unit for extracting relationship information from a conversation between first and second language users and personal information of the first and second language users and then recognizing the social relationship between the language users; a polite level selection unit for selecting a polite level for the conversation between the first and second language users based on the extracted relationship information; and a translation unit for generating polite expressions corresponding to the selected polite level and translating the conversation between the first and second language users into a target language based on the generated polite expressions.
The relationship information may be information including at least one of lexical expression information, age information, relationship intimacy information and social association information extracted from the conversation and the personal information.
The relationship recognition unit may include a first information collection unit for collecting lexical expression information from the conversation between the first and second language users; a part-of-speech classification unit for classifying the collected lexical expression information into parts of speech; and a first information extraction unit for extracting the information about the relationship between the first and second language users by analyzing results of the classification into the parts of speech.
The part-of-speech classification unit may classify the lexical expression information into parts of speech such as a lexical honor, an affix, a particle, a prefinal ending, and a final ending.
The relationship recognition unit may include a second information collection unit for collecting personal information from social networks that are used by the first and second language users; and a second information extraction unit for extracting the information about the relationship between the first and second language users by analyzing the collected personal information.
The relationship recognition unit may include an information provision unit for receiving the relationship information from the first and second language users and providing the relationship information.
The polite level selection unit may include a level classification unit for classifying a polite level into a plurality of preset polite levels; and a level mapping unit for selecting and assigning a polite level for the conversation between the first and second language users by mapping the extracted relationship information into one of the classified polite levels.
The translation unit may include a polite expression generation unit for generating the polite expressions corresponding to the selected polite level; a morpheme analysis unit for performing morpheme analysis on sentences of the conversation between the first and second language users;
a structure analysis unit for analyzing structures of the sentences of the conversation between the first and second language users while referring to a translation memory DB and a translation pattern DB; a translation generation unit for generating a translation in which the sentences of the conversation between the first and second language users have been converted into the target language while referring to morphemes and structures, obtained by the analysis based on the generated polite expressions, and the translation pattern DB and a translation dictionary DB; and a translation output unit for outputting the generated translation.
In order to accomplish the above object, the present invention provides a method of generating polite expressions for automatic translation, including, by a relationship recognition unit, extracting relationship information from a conversation between first and second language users and personal information of the first and second language users, and then recognizing the social relationship between the language users; by a polite level selection unit, selecting a polite level for the conversation between the first and second language users based on the extracted relationship information; and by a translation unit, generating polite expressions corresponding to the selected polite level, and translating the conversation between the first and second language users into a target language based on the generated polite expressions.
The recognizing the social relationship between the language users may include extracting the information about the relationship between the first and second language users by analyzing results obtained by classifying lexical expression information, collected from the conversation between the first and second language users, into parts of speech.
The recognizing the social relationship between the language users may include extracting the information about the relationship between the first and second language users by analyzing personal information collected from social networks that are used by the first and second language users.
The recognizing the social relationship between the language users may include receiving the relationship information from the first and second language users.
The selecting a polite level for the conversation between the first and second language users may include selecting the polite level for the conversation between the first and second language users by performing mapping to one of a plurality of preset levels.
The translating the conversation between the first and second language users into a target language may include generating and outputting a translation in which sentences of the conversation between the first and second language users have been converted into the target language while referring to morphemes and structures obtained by analysis based on the generated polite expressions.
The above and other objects, features and advantages of the present invention will be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings, in which:
Preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings in order to enable those having ordinary knowledge in the field of art to which the present invention pertains to easily practice the technical spirit of the present invention. It should be noted that the same or similar reference numerals are used to designate similar and identical elements throughout the accompanying drawings. Furthermore, in the following description of the present invention, detailed descriptions of well-known functions or configurations which would unnecessarily obscure the gist of the present invention will be omitted.
An apparatus and method for generating polite expressions for automatic translation according to an embodiment of the present invention will be described in detail below with reference to the accompanying drawings.
Referring to
In the present invention, the first language user is a user who uses Korean, and the second language user is a user who uses any language other than Korean. However, the present invention will be described with the second language user being a user who uses English.
The relationship recognition unit 110 extracts relationship information from a conversation between the first and second language users and the personal information of the first and second language users, and recognizes the social relationship between the language users.
Here, the relationship information is information including at least one of lexical expression information, age information, relationship intimacy information, and social association information extracted from the conversation and the personal information.
That is, the relationship recognition unit 110 obtains age information, relationship intimacy information, and social association information from previously stored information about the relationship between the first and second language users, lexical expression information collected from the conversation between the first and second language users, the personal information of the first and second language users collected from a social network, and relationship information input directly by the first and second language users, and recognizes the social relationship between the first and second language users.
In greater detail, if the relationship between the first and second language users has already been stored, the relationship recognition unit 110 immediately extracts information about the relationship between the two language users, and transfers the information to the polite level selection unit 120. In contrast, if the relationship between the first and second language users has not been stored, the relationship recognition unit 110 immediately extracts relationship information from the conversation between the two language users and personal information, and transfers the information to the polite level selection unit 120. Furthermore, if relationship information cannot be extracted even from the conversation between the first and second language users or personal information, the relationship recognition unit 110 may receive relationship information directly from the two language users, and transfer it to the polite level selection unit 120.
For this purpose, the relationship recognition unit 110, as shown in
The first information collection unit 111 may collect lexical expression information from the conversation between the first and second language users.
The part-of-speech classification unit 112 may classify the collected lexical expression information into parts of speech. Here, the part-of-speech classification unit 112 may classify lexical expression information into parts of speech, such as a lexical honor, an affix, a particle, a prefinal ending, and a final ending. For example, the lexical honor is classified as an ordinary lexicon or a polite lexicon, the affix is “ (Transliteration: nim; affix of polite lexicon),” the particle is “
(Transliteration: kkaeseo; particle of polite lexicon),” the prefinal ending is “
(Transliteration: si; prefinal ending of polite lexicon),” and the final ending is classified as “
(Transliteration: haerache; final ending of ordinary lexicon),” “
(Transliteration: haeche; final ending of ordinary lexicon),” “
(Transliteration: haeyoche; final ending of polite lexicon),” “
(Transliteration: hapsyoche, final ending of polite lexicon),” or the like. Here, “
(Transliteration: haeche)” may have a polite level higher than “
(Transliteration: haerache)”. “
(Transliteration: haeyoche)” may have polite level higher than “
(Transliteration: haerache)”. “
(Transliteration: hapsyoche)” may have polite level higher than “
(Transliteration: haeyoche)”.
The first information extraction unit 113 may extract information about the relationship between the first and second language users by analyzing the results of the part-of-speech classification.
The second information collection unit 114 may collect personal information from social networks that are used by the first and second language users. That is, the second information collection unit 114 may estimate the social relationship between the two language users using the disclosed personal profiles and digital personal connection information of the two language users obtained from a social network such as Facebook or Twitter. In this case, the reason why the personal information disclosed over the social network is used is that the illegitimate use of personal information violates the personal information usage-related prescriptions of Information Communications Laws.
The second information extraction unit 115 may extract information about the relationship between the first and second language users by analyzing the collected personal information.
The information provision unit 116 may receive relationship information directly from the first and second language users. That is, when relationship information cannot be extracted from the conversation between the first and second language users, the information provision unit 116 receives relationship information by means of the selection of options from a previously stored relationship menu by the first and second language users. Here, examples of a variety of relationships, such as a friend, an intimate junior colleague, an intimate senior colleague, an intimate professor, an intimate boss, and a stranger, have been stored in the previously stored relationship menu.
The polite level selection unit 120 selects a polite level for the conversation between the first and second language users based on the extracted relationship information.
For this purpose, the polite level selection unit 120, as shown in
The level classification unit 121 classifies polite expressions into a plurality of levels. In this case, the level classification unit 121 according to the present invention determines a polite level based on the relationship information extracted by the relationship recognition unit 110, and classifies the polite level as one in a range from the lowest level 0 to the highest level 3. The levels are not limited thereto.
The level mapping unit 122 selects and allocates a polite level for the conversation between the first and second language users by mapping the extracted relationship information to one of preset polite levels. For example, when the relationship information has been decided based on lexical expression information, the level mapping unit 122 may map the relationship information to polite level 0 in the case in which the final ending corresponds to “ (Transliteration: haeche),” as in “
(Transliteration: i-geo-jom-hae, Translation: Do some of this)” or “
(Transliteration i-geo-jom-doe-wa-jwer, Translation: Help me do some of this)” to polite level 1 in the case in which the final ending corresponds to “
(Transliteration: haeyoche)” as in “
(Transliteration: i-geo-jom-haeyo!, Translation: Please do some of this!)” or “
(Transliteration: i-geo-jom-doe-wa-jwer-yo!, Translation: Please help me do some of this!),” to polite level 2 in the case in which the lexical honor corresponds to a polite lexicon, the affix is “
(Transliteration: nim),” the prefinal ending is “
(Transliteration: si)” and the final ending corresponds to “
(Transliteration: haeyoche),” as in “
,
(Transliteration: kwa-jang-nim,-sik-sa-ha-shut-a-yo?Translation: Did you have a meal, supervisor?)” and to polite level 3 in the case in which the lexical honor corresponds to a polite lexicon, the affix is “
(Transliteration: nim),” the particle is “
(Transliteration: kkeseo),” the affix is “
(Transliteration: nim),” the prefinal ending is “
(Transliteration: si)” and the final ending corresponds to “
((Transliteration: hapsyoche)”, as in “
(Transliteration: a-beo-nim.-Jin-ji-jap-sue-shut-seup-ni-kka?, Translation: Did you have a meal, father?).” An example of the polite level mapping method of the level mapping unit 122 for the cases in which relationship information is decided from personal information other than the above-described lexical expression information or by the direct input of the relationship information will be described in detail below.
The translation unit 130 generates polite expressions corresponding to the selected polite level, and translates the conversation between the first and second language users into a target language based on the generated polite expressions. The translation unit 130 generates and outputs a polite expression in accordance with the selected polite level. As a simple example, when the second language user utters “I can tell you the web site,” it is translated into the expression “ (Transliteration: website-leul-mal-hae-jul-sue-it-seup-ni-da)” in the case in which polite level 3 has been selected for the conversation, and it is translated into the expression “
(Transliteration: website-leul-mal-hae-jul-sue-it-a)” in the case in which polite level 0 has been selected for the conversation.
That is, the translation unit 130 applies the polite level selected by the polite level selection unit 120 to the conversation between the first and second language users, and each sentence of the conversation is output after undergoing morpheme analysis, structure analysis, and translation generation. In this case, it is determined whether a translation target sentence is present in a translation memory DB, and, if the translation target sentence is present, the translated sentence of the translation memory DB corresponding to the transferred polite level is output as translation results. In contrast, if the translation target sentence is not present in the translation memory DB, it is determined whether the translation target sentence may be applied to the translation pattern DB. If the translation target sentence may be applied to the translation pattern DB, the translated pattern of the translation pattern DB corresponding to the transferred polite level is output as translation results. If the translation target sentence is not present in the translation pattern DB, it is determined whether the translation target sentence may be applied to a translation dictionary DB. If the translation target sentence may be applied to a translation dictionary DB, the translated words of the translation dictionary DB corresponding to the transferred polite level are output as translation results.
For this purpose, the translation unit 130, as shown in
The polite expression generation unit 131 generates polite expressions corresponding to the selected polite level.
The morpheme analysis unit 132 performs morpheme analysis on each sentence of the conversation between the first and second language users.
The structure analysis unit 133 analyzes the structure of each sentence of the conversation between the first and second language users. Here, the structure analysis unit 133 causes a polite expression to be selected while referring to a translation memory DB and a translation pattern DB.
The translation generation unit 134 generates a translation in which the sentences of the conversation between the first and second language users have been converted into a target language based on morphemes and structures obtained by analysis based on the generated polite expressions. In this case, the translation generation unit 134 causes honorific translated words to be selected while referring to the translation pattern DB and the translation dictionary DB.
The translation output unit 135 outputs the generated translation.
Referring to
The relationship recognition unit 110 recognizes the relationship between two users based on the social network personal information of the first language user and the social network personal information of the second language user, as shown in
Thereafter, a polite level for a conversation between the first and second language users is selected based on the extracted relationship information using the polite level selection unit 120 at step S200. That is, the polite level selection unit 120 selects a polite level for the conversation between the first and second language users by performing mapping to one of a plurality of polite levels, as shown in
Thereafter, using the translation unit 130, polite expressions corresponding to the selected polite level are generated and then a conversation between the first and second language users is translated into the target language based on the generated polite expressions at S300. That is, the translation unit 130 may generate a translation in which the sentences of the conversation between the first and second language users have been translated into the target language while referring to morphemes and structures obtained by the analysis of the generated polite expressions, and then output it.
The present invention has the advantage of recognizing the social relationship with an interlocutor based on non-linguistic information such as personal information provided by a social network, and linguistic information such as conversational expressions, when translating a conversation between first and second language users into a target language, thereby overcoming linguistic and cultural differences between polite expressions and then providing automatic translation results.
The present invention has the advantage of recognizing the social relationship with an interlocutor and selecting a polite level for a conversation between first and second language users, thereby providing automatic translation results suitable for the interlocutor.
Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, without departing from the scope and spirit of the invention as disclosed in the accompanying claims.
| Number | Date | Country | Kind |
|---|---|---|---|
| 10-2012-0046138 | May 2012 | KR | national |
