KR20110116790A - Apparatus and method for providing translation service in portable terminal - Google Patents

Abstract

The present invention is to provide a translation service in a portable terminal, the translation service providing method, if an input sentence of the first language is input, extracting at least one concept word from the input sentence, and the candidate sentence stored in the construction data Selecting and presenting at least one similar sentence among the at least one similar sentence; and displaying a screen for inquiring whether to substitute the at least one similar sentence if the at least one similar sentence includes a substitution target conceptual word. And a step of replacing the substitute target conceptual word with a conceptual word included in the input sentence, and displaying a translated sentence in which a sentence including the substituted conceptual word is translated into the second language.

Description

Apparatus and method for providing a translation service in a mobile terminal {APPARATUS AND METHOD FOR PROVIDING TRANSLATION SERVICE IN PORTABLE TERMINAL}

The present invention relates to a portable terminal, and more particularly, to an apparatus and method for providing a translation service in a portable terminal.

In the situation where foreign language is needed, such as when traveling abroad, the demand for foreign language conversation assistance is increased and roaming function makes mobile phone mandatory for foreign travel. A conversation sentence translation service is being developed.

Conventional approaches to foreign language translation are based on natural language analysis of the native language and natural language generation of the foreign language. The position of emphasizing the practical aspects of translation also approaches translation as a kind of search problem. The position interpreted as a problem of search is composed of a pair of mother language sentences and foreign language sentences to form a document set, and then searches for the most similar native language sentences for the user's input, and searches for foreign language sentences corresponding to the sentences selected by the user in the found sentence sets. Use the method provided. Since the above-described method always provides the correct foreign language sentence, there is an advantage that even a person who is ignorant of the foreign language can speak a foreign language appropriate to the situation. In searching for similar sentences, there are a method of measuring similarity for each sentence and a method of extracting an inverse file and an index word used in the search.

In the case of mobile terminals with limited system resources such as mobile phones and personal data assistants (PDAs), due to capacity limitations and keyword search limitations, several steps are suggested rather than directly entering the phrase the user wishes to use. If you select a sentence or enter a phrase directly through the keyword search method is applied to provide an incomplete service. In addition, the speech service has a limitation in supporting the speech service in the case of directly inputting the speech file that is composed of the pre-stored voice file.

 Among the conventional techniques, the technique of translating through language analysis and generation is universal, but it is often difficult to cope with individual situations, and it is difficult to use practically because foreign sentences that are actually made include many errors. Therefore, in practical terms, it is possible to collect sufficient translation pairs and update user information, and a similarity-based translation approach of searching for similar sentences in the information can be efficient. The methods for determining the similarity between direct sentences are not effective in the way that the collection of sufficient pairs of speech influences performance because the amount of decision increases proportionally as the number of registered sentences increases. As with the regular search, the construction of the reverse file and index terms is effective because it can perform relatively independently on the amount of candidate sentence pairs. However, the method of using such a search depends on whether or not the words consisting of the index items actually appear. However, the performance may be degraded due to the characteristics of the sentence, which may have completely different index terms even though they have the same meaning. have.

Accordingly, an object of the present invention is to provide an apparatus and method for providing a foreign language translation service in a portable terminal.

Another object of the present invention is to provide an apparatus and method for constructing a native language sentence to be translated by a user by substituting a specific word in a portable terminal.

Another object of the present invention is to provide an apparatus and method for extracting a conceptual word from an input sentence in a portable terminal.

According to a first aspect of the present invention for achieving the above object, a translation service providing method for translating a first language into a second language in a portable terminal, when an input sentence of the first language is input, at least in the input sentence Extracting one conceptual word, selecting and presenting at least one similar sentence among candidate sentences stored in the construction data, and including a target word for substitution in the similar sentence selected by the user among the at least one similar sentence In this case, a process of displaying a screen for inquiring whether to substitute the target conceptual word, replacing the conceptual target word with a conceptual word included in the input sentence, and a sentence including the substituted conceptual word in the second language. And displaying the translated translation sentence, wherein the conceptual object to be substituted is included in the selected similar sentence and It is a conceptual word belonging to the same concept as the conceptual word included in the power sentence but representing a different object.

According to a second aspect of the present invention for achieving the above object, a portable terminal device for providing a translation service for translating a first language into a second language, when the input sentence of the first language is input, A control unit for extracting at least one conceptual word, selecting and presenting at least one similar sentence among candidate sentences stored in the construction data, and a substitution target conceptual word included in the similar sentence selected by the user among the at least one similar sentence A display unit for displaying a screen for inquiring whether to substitute the conceptual word for substitution, and the control unit replaces the replacement target conceptual word with a conceptual word included in the input sentence, and the display unit replaces a sentence including the substituted conceptual word. A translation sentence translated into a second language is displayed, and the conceptual object to be substituted is included in the selected similar sentence. Characterized in the same concept as the conceptual word included in the input sentence, but represents a different object.

In the portable terminal, an effective translation service may be provided by mapping a foreign language sentence and a native language sentence according to a meaning and providing a function of correcting a sentence through a conceptual word.

1 is a diagram illustrating a conceptual structure of a translation system according to an embodiment of the present invention;
2 is a diagram illustrating an example of a sentence type determination operation according to an embodiment of the present invention;
3 is a diagram illustrating a conceptual word extraction process according to an embodiment of the present invention;
4 is a diagram illustrating clustering according to an embodiment of the present invention;
5 is a diagram illustrating conceptual word substitution using a conceptual word band dictionary according to an embodiment of the present invention;
6 is a block diagram of a portable terminal according to an embodiment of the present invention;
7 is a diagram illustrating an operation procedure of a portable terminal according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to the preferred embodiments of the present invention, examples of which are illustrated in the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

Hereinafter, a technology for providing a foreign language translation service in a portable terminal will be described. In the following description, the portable terminal is a cellular phone, a personal communication system (PCS), a personal digital assistant (PDA), an international mobile telecommunication-2000 (IMT2000) terminal, a laptop (lap-top) computer. , Desktop (desk-top) computer, etc.

1 illustrates a conceptual structure of a translation system according to an embodiment of the present invention.

As shown in FIG. 1, the translation system includes a data management module 110, a sentence analysis module 120, and a translation application 130.

The data management module 110 constructs and stores data necessary for translation, and the data management module 100 includes a configuration for data construction and expansion operation and a configuration for construction data storage operation.

The data construction and expansion is a process of constructing a sentence pair provided as a result of the actual search. The data construction and expansion may include selecting a category suitable for a situation and determining a target sentence using information that can be inquired in each category (111), expanding a sentence pair according to the meaning of the sentence (113), and a sentence. The additional information attaching step (115) attaching the conceptual information and sentence type information included in the information is performed. Looking at each step in detail as follows.

The category cleanup and target sentence determination step 111 are as follows. One method of translation is the use of sentence pairs. The technique of using the sentence pair is to predefine a foreign language sentence corresponding to a specific native language sentence, and to provide the corresponding foreign language sentence when a translation request of the specific native language sentence is requested. The technique of using the sentence pair is more effective in an environment in which sentences are restricted. For example, in the case of translation for travel conversation, as can be seen in various travel booklets, sentences frequently used in each situation may be composed of a certain set. Table 1 below shows examples of collected categories and sentences belonging to each category.

category
Sentence count
category
Sentence count
category

Sentence count PC room 30 Department store 7 Confirm / Cancel / Change Reservation 25 Dining 37 Bus 123 post office 50 Pass / Transfer 27 Hospital / Pharmacy 113 Youth hostel 15 Airport Tourist Information Center 28 Boarding Pass 44 Bank 58 tourism 209 Lost / stolen 38 Immigration Examination 37 Preview 97 accident 64 telephone 132 Return procedure 53 Social / Entertainment 130 gas station 26 In-flight service 95 Ship 29 Subway / Railway 108 Meal order 35 Customs inspection 26 check out 37 Cabin seating 38 Find Shopping Center 9 taxi 97 Souvenir shop 7 Find Baggage 29 fastfood 16 Get Directions 40 supermarket 4 front 52 Car rental 61 Restaurant reservation / use 93 flight booking 38 Room service 43 Meals and Decisions 69 Hotel Restaurant 12 Duty Free Shop 4 Meal order 89
total
2697 Pick / Choose 121 Report Form 16 Article exchange / return / refund 28 Reservation / Check-in 158

The sentence pair expansion step 120 is as follows. In the translation system according to an embodiment of the present invention, a search target is a sentence. In sentence retrieval, sentence retrieval is difficult to solve by a general retrieval method because there are sentences of similar meanings in which keywords do not match. To this end, the present invention collects sentences of the same meaning as much as possible from various sentence sources. In this case, the present invention binds the collected sentences with the same meaning, and constructs sentence data by adding sentences having different expressions having the same meaning when a sufficient number of sentences are not collected.

The additional information attaching step 115 is as follows. In general, sentences can be expressed in short and various ways. Because of this, it is not possible to guarantee that users speak in a form similar to the candidate sentence, so that constructing band data in one sentence may make access to the candidate sentence data difficult. Thus, the present invention utilizes native language sentences with the same meaning. To this end, a process of extending each collected sentence to sentences having the same meaning is required. Since finding and managing sentences having the same meaning is a difficult problem, it is preferable that foreign sentences have the same sentences as semantically identical sentences. For example, it is preferable that at least three native language sentences having the same meaning correspond to one foreign language sentence. In addition, as a situation that may occur in the process of searching for a sentence by sentence comparison, there may be cases in which sentences having similar meanings but having different purposes are present. For example, a sentence requesting a ticket to a particular country may be identical except for the country name. Since the information collected from the various resources has no consideration of the items that can be changed according to this situation, the present invention defines items that change as the country name. For example, the types of the changing items are shown in Table 2 below.

Main Category

Subclass
Explanation
tag Person Person Also includes titles such as Hong Gil-dong and Uncle PER

location


Location Place name (country, city, etc.) LOC Restaurant McDonald's, VIIPS, 'Restaurant' RST Hotel Hilton hotel, Seoul hotel HTL Airport Incheon International Airport, Gimpo International Airport AIR Station Seoul Station, Sinchon Station STA Store Hyundai Department Store, Seven Eleven STO Position Buildings, offices and places of a kind not defined above POS
Object
Produce Air conditioner, bag PRD Brand Levis, Louis Vuitton BRD Food Food name FOD Bevarage Drink name BEV

Num

Money $ 10, $ 1000 MNY Time 10 pm, 2 pm TIM Day July 1, next month 5 DAY Duration Duration, 1 hour DUR Telephone Phone number TEL Number Numbers of a kind not defined above NUM Tool Transportation Transportation such as subway, bus TRP

Traditional approaches to sentence comparison use keywords such as verbs, verbs, etc. to construct a vector and compare the vectors. However, the traditional approaches have limitations in that they cannot consider grammatical elements in finding the actual sentence type. Moreover, taking into account all morphemes and various surface information requires a lot of computation. Accordingly, the present invention defines a sentence type and performs sentence comparison using the sentence type. For example, the sentence type to be defined is shown in Table 3 below.

Main Category

Subclass
Explanation
example Declarative Inform Notification, Information I am vegetarian. I don't like oily food. One more person will come. Imperative Request Request, request Please refill your coffee. cut please. Would you please wrap the leftovers? Propositive Offer suggestion Should we go for lunch? Let's go shopping. How about ribs for dinner? Exclamatory Admire Admiration Interrogative Ask_what Question-target What's the special dish of the day? What's the material? Ask_when Question-time When do you come when you order now? How long should I wait? What time do you close? Ask_where Question-place Where is the bathroom? Where is the department store? Ask_who Question-person Who should I ask? Ask_which Question-choice What transport will you use? Ask_why Question-why Why is food not coming? Ask_how Question-how, quantity, status How does it taste? How many are you? How much? Ask_yn Question-Y / N Will it be delivered? Is it now on sale? Other Other Etc Cheers! / One shot!

The above-described data building and expanding operation may be performed by the system designer offline or by the system for an input sentence. In addition, a sentence input as a translation target while providing a translation service may be a target of the data construction and expansion operation.

Next, through the constructing data storing operation, candidate sentence data is converted into a form of a search through the constructing and expanding the data, and then stored as the search information by the sentence storing system 117. In general, in order to determine the similarity of sentences, an approach through a search method is used to solve the speed problem of individual sentence comparison methods. The sentence storage system 117 constructs an index structure to sufficiently reflect the added information. For example, the contents stored in the index structure are as shown in Table 4 and Table 5 below. Table 4 below shows the structure of the document information file, and Table 5 below shows the structure of the posting file.

Contents
key (number)
Data
Sentence information sid stf, stype, GEN_num (sid1; sid2; ...) Native language sentences; English sentence CS information in sentences Last sid + sid (cs_type, start_index, cs_len) +

In Table 4, 'sid' is a sentence identifier, 'stf' is the number of index words appearing in the sentence, 'stype' is the sentence type, and 'GEN_num' is the index word except the conceptual word in the sentence. The number, 'cs_type' is the conceptual word type, 'start_index' is the occurrence position in the sentence, 'cs_len' means the sentence length of cs appeared in the sentence.

Contents
key (string)
Data
Frequency of the entire term #term frequency Frequency of overall cs @cs frequency Frequency of full st $ st frequency reverse file for term Term (sf, cf, ttf) (sid, cid, tf) + reverse file for cs <cs> (sf, cf, ttf) (sid, cid, tf) + reverse file for st <st> (sf, cf, ttf) (sid, cid, tf) + The total number of terms TOTAL_TF_KEY frequency Frequency of that term within that cluster cid + term frequency Frequency of that cs in that cluster cid + cs frequency Frequency of that cluster GEN cid + GEN frequency The total number of terms in the cluster "T" + cid frequency

In Table 5, 'term' is a keyword list, 'cs' is a conceptual word,' st 'is a sentence type including a corresponding key,' cf 'is the number of clusters including a corresponding key, and' ttf 'is the total frequency of the corresponding key,' sid 'is the sentence identifier,' cid 'is the cluster identifier including the key,' tf 'is the frequency of the key in the sentence, and' stf 'is the corresponding frequency. The number of index words that appear in a sentence.

The sentence analysis module 120 extracts the necessary information from the data constructed for the search to construct a keyword. As shown in FIG. 1, the sentence analysis module 120 determines whether the sentence type automatically extracts the information described through the morphological analysis step 121 used in general natural language processing and the data construction and expansion step. A step 123 and a conceptual word extraction step 125 are included.

The morphological analysis step 121 is as follows. A morphological analysis step is essential to make natural sentence lists a key keyword. The present invention constructs a keyword centering on a statement and a verb including important information in a sentence through morphological analysis. In sentence translation, not only key words such as the suffix, the verb, etc., but also the additional morphemes that act as grammatical elements play an important role. However, taking into account all the above morphemes, the computation time increases due to the increase in the capacity of data to be processed. In order to solve this problem, the present invention configures additional morphemes excluding key words as sentence type information. This keeps only a small number of keywords as in a general search, thereby reducing storage resources and computation time.

The sentence type determination step 123 is as follows. In sentence translation, words representing various grammars and situations should be translated. However, it is difficult to use all the information due to the limitation of storage resources. To overcome this problem, the present invention defines a predetermined number of sentence types that determine the grammatical elements of the input content. For example, the sentence types are shown in Table 3 above. Defined sentence types are classified grammatically and assigned to input sentences in consideration of processing speed and storage space. In this case, it is preferable to use a lexical semantic pattern for general use of the constructed grammar so that it can be easily added or managed in the future. For example, a specific operation example of the sentence type determination step 123 to which the lexical semantic pattern is applied is shown in FIG. 2.

2 illustrates an example of a sentence type determination operation according to an embodiment of the present invention. Referring to FIG. 2, the input sentence 210 is, "Would you like a single room or a double room?" In this case, the constructed regular expression 220 is "+ (@ and | or \ s | and \ s)?. + (@ Among | or). * (@ Which | @what | optional). * (\? | \.) $-> Request ". When the input sentence 210 is analyzed in morphological units, ". + (And \ s | and \ s | or \ s | and \ s)? + (\ S of \ s | middle | or). * ( Which | any | where | what | optional). * (\? | \.) $-> Request " Therefore, the sentence type of the input sentence 210 is determined as a 'request'.

The conceptual word extraction step 125 is as follows. In order to correspond to various expressions having the same meaning, a unit capable of collectively processing the various expressions is required. In the present invention, the unit is referred to as a concept (eg, place), and a concrete expression belonging to the same concept is a concept word (eg, Seoul, New York). Accordingly, the present invention defines effective concepts for the collected corpus. For example, the definitions of the concepts are shown in Table 2 above. Constructors can express necessary concepts in a sentence when collecting corpus, but the user inputs a conceptual word rather than a concept. Therefore, the translation system should automatically extract the concept word, and perform the process of identifying the concept. 3 shows a process for automatically extracting conceptual words and specific examples of each step.

3 illustrates a process for automatically extracting conceptual words according to an embodiment of the present invention. Referring to FIG. 3, an input sentence is input in step 301. For example, the input sentence is "I would like to book a flight to New York." Then, in step 303, the morphological analysis of the input sentence is performed. The analysis result of the morphological composition is "New York / NNF + / JKB / VV + / ETM Planes / NNG Flight / NNB + / JKO Booking / NNG + / JKB Wanted / VX + / EF +. / SF". In operation 305, a conceptual word dictionary search is performed to extract conceptual words from the analyzed morphemes. In this case, it is assumed that 'New York / NNP' is not registered in the conceptual word dictionary and 'Airplane / NNG' is registered. In step 307, first tagging is performed. As a result of the primary tagging, 'plane / NNG' is tagged as '#TRP'. In operation 309, unregistered conceptual word extraction rule matching is performed. The unregistered conceptual word extraction rule uses a search and a verb combined with a noun. In the case of FIG. 3, since the unregistered conceptual word 'New York / NNP' is combined with the search '/ JKB' and the verb 'ga / VV' indicating a destination, the 'New York / NNP' is assumed to be a noun representing a place. . Therefore, the 'New York / NNP' is not registered in the conceptual word dictionary, but is determined as a conceptual word having a concept of 'place' by the unregistered conceptual word extraction rule as described above. That is, the concept of the noun is determined using at least one of a search and a verb combined with a noun. Thereafter, in step 311, the second tagging is performed, so that 'New York / NNP' is tagged as '#LOC'. Thereafter, in step 313, it is determined that 'I would like to book a flight {{Airplane; TRP}} to {{New York; LOC}}' as a result of the conceptual word extraction.

The translation application 130 searches for a sentence most similar to the input sentence among the registered sentences, and outputs the translated sentence by processing the found sentence appropriately. As shown in FIG. 1, the operation of the translation application 130 includes a query analysis step 131, a language model sentence comparison step 133, a suitable candidate selection step 135, and a result generation step 137. do.

The query analyzing step 131 is a process of extracting information necessary for searching from an input sentence.

The language model sentence comparison step 133 is as follows. The purpose of sentence comparison is to determine the constructed native language sentence most similar to the input native language sentence. For sentence comparison, the present invention uses a probabilistic language model method. The language model method determines a probability value for generating another sentence from one sentence and evaluates similarity using the probability value. For example, the probability value is determined as in Equation 1 below.

는 유사도, 상기

는 후보 문장, 상기

는 입력 문장, 상기

는 입력 문장을 구성하는 i번째 항목, 상기

은 입력 문장을 구성하는 항목들의 개수를 의미한다.In Equation 1,

Is similarity, said

Is a candidate sentence, said

Is the input sentence, said

Is the i th item constituting the input sentence,

Means the number of items constituting the input sentence.

에서 발견되지 않음으로 인해 확률 값이 0이 될 수 있다. 따라서, 위의 <수학식 1>을 구축된 전체 문장들의 집합에서

가 나타날 확률에 따라 스무싱(Smoothing)하면 하기 <수학식 2>와 같이 정리된다.In accordance with Equation 1, a large number of items

The probability value can be zero because it is not found in. Therefore, in the set of entire sentences constructed with Equation 1 above,

Smoothing according to the probability of appearing is summarized as in Equation 2 below.

는 유사도, 상기

는 후보 문장, 상기

는 입력 문장, 상기

는 입력 문장을 구성하는 i번째 항목, 상기

은 입력 문장을 구성하는 항목들의 개수, 상기

는 가중치, 상기

은 구축된 전체 문장들의 집합을 의미한다.In Equation 2,

Is similarity, said

Is a candidate sentence, said

Is the input sentence, said

Is the i th item constituting the input sentence,

Is the number of items constituting the input sentence,

Is a weight,

Means the complete set of sentences.

Similarity with an input sentence determined according to Equation 2 is determined, and a sentence having a maximum probability value is determined as a similar sentence.

There is a case where a concept is expressed as various words in a sentence. For example, words such as Seoul, Busan, and Incheon have a common concept of city name. Therefore, in addition to using a word as a keyword in a search, it is required to use the concept of a word as a keyword. The present invention defines a set of keywords used with concepts, along with words, to represent sentences in indexes and searches. For example, the keyword set is shown in Equation 3 below.

는 키워드 집합, 상기

는 문장

에 포함된 단어 집합, 상기

는 문장

의 개념 집합, 상기

는 문장이 가지는 문장 유형 타입을 의미한다.In Equation 3,

Is a set of keywords, said

Sentence

A set of words contained in the above

Sentence

Set of concepts, reminding

Means a sentence type type that a sentence has.

Here, the concept set may be expressed as Equation 4 below.

는 문장

의 개념 집합, 상기

는 개념, 상기

는 문장

에 포함된 단어 집합, 상기

는 i번째 개념, 상기

는 특정 개념을 가지지 아니하는 단어들의 개념 분류, 상기

는 상기 를 포함하는 단어 집합을 의미한다.In Equation 4,

Sentence

Set of concepts, reminding

Remind, concept

Sentence

A set of words contained in the above

Is the i th concept, said

Is a conceptual classification of words that do not have a particular concept,

Above Means a word set that includes.

이라는 개념 분류에 포함시킨다. If the keyword is added only to the keyword, the sentence containing the conceptual word is lengthened, so that the conceptual word has a lower probability than the general word. To address this problem, the present invention uses words that do not have a concept.

In the concept category

Since the translation application 130 is a comparison of sentences and sentences, relatively few words match. Thus, since sufficient probability information is not guaranteed for sentence comparison, performance may be degraded. In addition, even if the same meaning, the words constituted may be completely different. In order to prevent the similarity between sentences having the same meaning due to the diversity of words and the lack of information, the present invention uses clusters that group sentences having the same meaning rather than searching with only one sentence. The cluster is determined by clustering, and an example of the clustering is shown in FIG. 4.

4 illustrates clustering according to an embodiment of the present invention. Referring to FIG. 4, the input sentence 410 is, "How late does the car run?" Since there is no overlapping information or only one candidate sentence for the input sentence 410, an appropriate sentence is difficult to determine. However, considering the information of sentences having the same meaning, since the sentences belonging to the cluster B 402 are supported by the sentences belonging to the same cluster than the cluster A 401 and the cluster C 403, the input sentences ( Similar to 410). The information of the cluster may be applied to Equation 2 through a smoothing technique. Applying the concept of the cluster, Equation 2 is summarized as Equation 5 below.

는 유사도, 상기

는 후보 문장, 상기

는 입력 문장, 상기

는 입력 문장을 구성하는 i번째 항목, 상기

은 입력 문장을 구성하는 항목의 개수, 상기

및 상기

는 가중치, 상기

는 상기

와 동일한 의미를 가지는 문장들의 집합, 상기

은 구축된 전체 문장들의 집합을 의미한다.In Equation 5,

Is similarity, said

Is a candidate sentence, said

Is the input sentence, said

Is the i th item constituting the input sentence,

Is the number of items constituting the input sentence,

And

Is a weight,

Above

A set of sentences having the same meaning as

Means the complete set of sentences.

In Equation 5, a probability value for each item is determined as shown in Equation 6 by the Maximum-Likelihood Estimation (MLE) technique.

는 후보 문장, 상기

는 입력 문장을 구성하는 i번째 항목, 상기

은 입력 문장을 구성하는 항목의 개수를 의미한다.In Equation 6,

Is a candidate sentence, said

Is the i th item constituting the input sentence,

Means the number of items constituting the input sentence.

및

는 디리끌레 프라이어 스무싱(Dirichlet prior smoothing)을 적용하여 하기 <수학식 7>과 같이 결정된다.In addition, the weight in Equation 5

And

Is determined by Equation 7 by applying Dirichlet prior smoothing.

및 상기

확률 값 결정을 위한 가중치들, 상기

는 후보 문장, 상기

는 상기

와 동일한 의미를 가지는 문장들의 집합을 의미한다.In Equation 7,

And

Weights for determining a probability value, said

Is a candidate sentence, said

Above

Means a set of sentences having the same meaning as.

The result generating step 137 is as follows. When the sentence similarity is determined through the language model, the user selects a sentence having the closest meaning among the presented native language sentences. At this time, among the presented sentences there may not be a sentence that completely matches the sentence desired by the user. In particular, the presented sentences have the expression desired by the user, but the contents may be different because only the concepts match and the specific words do not match. The present invention performs a translation sentence correction process using a conceptual band dictionary to prevent this situation. The sentence correction is performed through substitution between words having the same concept. The conceptual band dictionary is used to connect words corresponding to a native language and a foreign language with respect to a word or phrase that may be a candidate of the conceptual word, and to substitute the corresponding conceptual word of the determined sentence. For example, the result generation using the conceptual band dictionary is shown in FIG. 5.

5 illustrates conceptual word substitution using a conceptual word band dictionary according to an embodiment of the present invention. Referring to FIG. 5, the input sentence 510 is "Please give two tickets to New York." Sentences 520 presented through a similarity comparison for the input sentence 510 is "I want to book a return ticket for a flight from Seoul to New York", "Please give me a ticket to San Francisco,""To New York. How much is the return ticket? "," Please give me a student ticket. "," Please give me a return ticket for adults. " At this point, the user selects the second sentence, "Please give me a ticket to San Francisco." Accordingly, the translation application 130 displays a phrase 530 for notifying the substitution query and the conceptual words that can be substituted. As shown in FIG. 5, the concept words that can be substituted in the second sentence are 'san francisco' and 'one chapter'. The user selects 'substitute all', and accordingly, "Two tickets to New York, please." Is output as the translation result 540.

The present invention will be described below the configuration and operation of a portable terminal providing a translation service as described above.

6 is a block diagram of a portable terminal according to an embodiment of the present invention.

As shown in FIG. 6, the display unit 602 includes an input unit 602, an input unit 604, a storage unit 606, and a control unit 608.

The display unit 602 displays status information generated during operation of the terminal and numbers, letters, and images according to execution of an application program. That is, the display unit 602 displays the image data provided from the control unit 608 on a visual screen. For example, the display unit 602 may include a liquid crystal display (LCD), an organic light-emitting diode (OLED), or the like.

The input unit 604 recognizes an input generated by a user and provides information corresponding to the input to the controller 608. That is, the input unit 604 processes a user's input through a keyboard, a keypad, a touch screen, a touch pad, a mouse, special function buttons, and the like.

The storage unit 606 stores an operating system, an application program, and user contents necessary for the operation of the portable terminal. In particular, the storage unit 606 stores translation band data, conceptual band dictionary, sentence information, index information, and translation application for a translation service according to an embodiment of the present invention. The storage unit 606 provides the stored data to the control unit 608 at the request of the control unit 608.

The controller 608 controls the overall functions of the portable terminal. For example, the control unit 608 provides image data to the display unit 602 and processes input information provided from the input unit 604. In particular, according to an embodiment of the present invention, the controller 608 controls the functions for the translation service. Functions for the translation service include a data management function for the translation service, a sentence analysis function, and an application execution function for the translation service. Looking at the function for the translation service in detail as follows.

In order to manage data for the translation service, the controller 608 stores construction data for a translation service input from the outside in the storage unit 606. The construction data includes sentence information, mapping information between a native language sentence and a foreign language sentence, category classification information of sentences, conceptual word information for sentence comparison, sentence type information, regular expression information for each sentence type, cluster information, and sentence retrieval. Index information and the like. For example, the category classification is as in Table 2, the sentence type is defined as in Table 3, the sentence information is as shown in Table 4, and the index information is as shown in Table 5. It can be configured together. In addition, the control unit 608 reinforces the construction data by storing, in the storage unit 606, an analysis result of a native language sentence that is input as a translation target when providing the translation service.

For the sentence analysis function, the controller 608 analyzes the morpheme structure of the input sentence input by the user as a target of translation, determines the sentence type of the input sentence using the morpheme structure, and then applies the input sentence. Extract included conceptual words. In this case, the conceptual word may not exist. In the morpheme analysis, the controller 608 processes key words such as a verb and a verb as keywords, and additional morphemes that are grammatical elements as information indicating a sentence type. In determining the sentence type, the controller 608 determines the sentence type of the sentence based on the result of the morpheme analysis. For example, candidates of sentence type are as shown in Table 3 above. In concept word extraction, the controller 608 extracts a concept word included in a sentence by referring to a pre-built concept word band dictionary. In the sentence type determination, the controller 608 compares the analysis result of the morpheme structure with a morpheme structure of a predefined regular expression, so that the sentence type belongs to a plurality of predefined sentence types of sentence types. To judge. For example, the controller 608 determines the sentence type as shown in FIG. 3. In the conceptual word extraction, the controller 608 first extracts a conceptual word registered in the conceptual word band dictionary, extracts an unregistered conceptual word according to an unregistered conceptual word extraction rule, and then tags the extracted word. However, since the input sentence does not always include a concept word, when the concept word is not extracted, the tagging may be omitted.

In order to execute an application for a translation service, the controller 608 executes a translation application stored in the storage unit 606. The translation application includes a user interface for a translation service, sentence comparison and search algorithm. In executing the translation application, the storage unit 606 displays user interface screens defined through the display unit 502. The user interface screens include a screen for inputting a sentence, a screen for presenting a similar sentence, a screen for replacing a conceptual word, a screen for displaying a translation result, and the like. After executing the translation application, when a native language sentence to be translated is input, the controller 608 analyzes the input sentence according to the sentence analysis function, selects and presents at least one similar sentence according to the sentence comparison algorithm. After performing the substitution process and whether or not the conceptual word, the translation sentence is displayed. To select the at least one similar sentence, the controller 608 compares the input sentence with candidate sentences stored in the storage 606. In this case, the controller 608 uses clustering of candidate sentences. In detail, the controller 608 searches for sentences including some or all of keywords of the input sentence among candidate sentences belonging to the sentence type of the input sentence determined through the sentence analysis function. In addition, the controller 608 identifies the cluster distribution of sentences including some or all of the keywords, and presents at least one similar sentence in the cluster including the largest number of sentences. For example, the controller 608 calculates probabilities that the input sentence is generated in each candidate sentence, and calculates at least one candidate sentence having the highest probability with the at least one similar sentence. It is presented as. For example, the probability may be calculated as shown in Equation 5 above. Here, the number of the at least one similar sentence may vary according to a specific embodiment. If a specific sentence is selected by the user among the at least one similar sentence, the controller 608 inquires whether to replace at least one conceptual word included in the selected sentence, and inquires the at least one conceptual word according to a user's manipulation. Substitute or retain That is, the controller 608 displays a screen indicating whether to replace or replaces all or part of the at least one conceptual word or maintains the at least one conceptual word according to the user's manipulation.

The present invention has been described with reference to FIG. 6, in which a native language sentence is input and a foreign language sentence is provided. However, according to another embodiment of the present disclosure, the controller 608 may receive a keyword other than the native language sentence. In this case, the controller 608 may allow the user to select a sentence to be translated into a foreign language by presenting at least one candidate sentence including the keyword.

7 illustrates an operation procedure of a portable terminal according to an embodiment of the present invention.

As illustrated in FIG. 7, the portable terminal stores construction data in step 701. The construction data includes sentence information and index information for a translation service. The construction data may include sentence information, mapping relation information between a native language sentence and a foreign language sentence, category classification information of sentences, conceptual word information for sentence comparison, sentence type information, regular expression information for each sentence type, cluster information, index information, and the like. For example, the sentence information and the index information may have structures such as <Table 4> and <Table 5>. Additionally, the portable terminal may store translation band data and conceptual band dictionary. After the data stored in operation 701 is configured by the system designer, the data is provided from the outside prior to providing the first translation service.

In step 703, the portable terminal checks whether a native language sentence to be translated is input. That is, after the translation application is executed, the portable terminal determines whether a native language sentence to be translated is input.

When the native language sentence is input, the portable terminal proceeds to step 705 to analyze the native language sentence, that is, the input sentence to be translated. Analysis of the input sentence includes morphological analysis, sentence type determination, and conceptual word extraction. That is, the portable terminal analyzes the morpheme structure of the input sentence, determines the sentence type of the input sentence using the morpheme structure, and extracts the conceptual word included in the input sentence. In this case, the conceptual word may not exist. In the morpheme analysis, the portable terminal processes key words such as verbs, verbs, etc. as keywords, and additional morphemes, which are grammatical elements, as information representing sentence types. In determining the sentence type, the portable terminal compares the analysis result of the morpheme configuration with a morpheme configuration of a predefined regular expression to determine which sentence type among a plurality of predefined sentence types. For example, the portable terminal determines the sentence type as shown in FIG. 3. When the conceptual word is extracted, the portable terminal first extracts a conceptual word registered in the conceptual word band dictionary, extracts an unregistered conceptual word according to an unregistered conceptual word extraction rule, and then tags the extracted word. However, since the input sentence does not always include a concept word, when the concept word is not extracted, the tagging may be omitted.

In step 707, the portable terminal stores information of the input sentence. That is, the portable terminal reinforces the construction data by storing analysis information of the input sentence obtained in step 705. According to another embodiment of the present disclosure, step 707 may be omitted or performed at another point in time.

In operation 709, the portable terminal presents at least one similar sentence. The at least one similar sentence is at least one of candidate sentences stored in the construction data, and is selected through comparison with the input sentence and the candidate sentences. In this case, the portable terminal uses clustering of candidate sentences. That is, each of the candidate sentences belongs to a specific cluster. In detail, the portable terminal searches for sentences including some or all of keywords of the input sentence among candidate sentences belonging to the sentence type of the input sentence determined in step 705. The portable terminal identifies a cluster distribution of sentences including some or all of the keywords and presents at least one similar sentence in a cluster including the largest number of sentences. For example, the portable terminal calculates probabilities of generating the input sentence in each candidate sentence, and presents at least one candidate sentence having the highest probability as the at least one similar sentence. For example, the probabilities may be calculated as shown in Equation 5 above. Here, the number of the at least one similar sentence may vary according to a specific embodiment.

After presenting the similar sentence, the portable terminal proceeds to step 711 and determines whether one of the at least one similar sentence is selected by the user. That is, the portable terminal determines whether the user's manipulation for selecting a sentence occurs.

When the sentence is selected by the user, the portable terminal proceeds to step 713 to determine whether at least one concept word to be substituted is included in the selected sentence. Here, the at least one conceptual word to be substituted means a conceptual word included in the selected sentence and belonging to the same concept as the conceptual word of the input sentence, but representing a different object. For example, when 'New York' is included in the input sentence and 'Seoul' is included in the selected sentence, 'Seoul' belongs to the concept word to be replaced. If at least one concept word to be substituted is not included, the portable terminal proceeds to step 717.

On the other hand, if the at least one concept word to be substituted is included, the portable terminal proceeds to step 715 and inquires whether the at least one concept word is replaced, and replaces or maintains the at least one concept word according to a user's operation. . That is, the portable terminal displays a screen indicating whether to replace or replaces all or part of the at least one conceptual word or maintains the at least one conceptual word according to the user's manipulation.

In operation 717, the portable terminal provides a foreign language sentence corresponding to the determined native language sentence as a translation result. In other words, the portable terminal displays a foreign language sentence stored as a translation sentence of the determined native language sentence on the screen.

The present invention has been described with reference to FIG. 7 in which a native language sentence is input and a foreign language sentence is provided. However, according to another embodiment of the present invention, the portable terminal may receive a keyword other than the native sentence. In this case, the portable terminal may allow the user to select a sentence to be translated into a foreign language by presenting at least one candidate sentence including the keyword.

Meanwhile, in the detailed description of the present invention, specific embodiments have been described, but various modifications are possible without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined not only by the scope of the following claims, but also by the equivalents of the claims.

Claims (20)

A translation service providing method for translating a first language into a second language in a portable terminal,
Extracting at least one concept word from the input sentence when an input sentence of the first language is input;
Selecting and presenting at least one similar sentence among candidate sentences stored in the construction data;
If the similar sentence selected by the user among the at least one similar sentence includes a substitution target concept word, displaying a screen for inquiring whether to substitute the conceptual word for substitution;
Substituting the concept word to be substituted into the concept word included in the input sentence;
Displaying a translation sentence in which a sentence including a substituted concept word is translated into the second language,
The substitute target conceptual word is a conceptual word included in the selected similar sentence and belonging to the same concept as the conceptual word included in the input sentence, but representing a different object.
The method of claim 1,
Extracting the at least one concept word,
Extracting the conceptual words registered in the conceptual band dictionary;
Extracting the unregistered conceptual word according to the unregistered conceptual word extraction rule;
Tagging the extracted at least one conceptual word.
The method of claim 2,
The process of extracting the unregistered conceptual word according to the unregistered conceptual word extraction rule may include:
Determining a concept of the noun using at least one of a verb and a verb combined with a noun.
The method of claim 1,
When the input sentence is input, analyzing a morpheme structure of the input sentence;
And determining a sentence type of the input sentence by using the analysis result of the morpheme composition.
The method of claim 4, wherein
The process of determining the sentence type of the input sentence,
And comparing the analysis result of the morpheme construct with a morpheme construct of a predefined regular expression.
The method of claim 4, wherein
The process of selecting the at least one similar sentence,
Searching for candidate sentences including some or all of keywords of the input sentences among candidate sentences belonging to a sentence type of the input sentences;
Identifying a cluster distribution of candidate sentences including some or all of the keywords, and selecting the at least one similar sentence within a cluster including the largest number of sentences.
The method of claim 6,
Wherein the at least one similar sentence is selected using the probabilities of generating the input sentence in each candidate sentence.
The method of claim 7, wherein
The probabilities of generating the input sentence in each candidate sentence are determined by the following equation,

Where

Is similarity, said

Is a candidate sentence, said

Is the input sentence, said

Is the i th item constituting the input sentence,

Is the number of items constituting the input sentence,

And

Is a weight,

Above

A set of sentences having the same meaning as

Means the complete set of sentences,

Where

Is a candidate sentence, said

Is the i th item constituting the input sentence,

Means the number of items that make up the input sentence.

Where

And

Weights for determining a probability value, said

Is a candidate sentence, said

Above

Means a set of sentences with the same meaning as
The method of claim 1,
And reinforcing the construction data by storing the input sentence.
The method of claim 1,
The construction data may include sentence information, mapping relationship information between a first language sentence and a second language sentence, category classification information of sentences, conceptual word information for sentence comparison, sentence type information, regular expression information for each sentence type, and sentence search. And at least one of index information for the cluster information and cluster information.
A portable terminal apparatus for providing a translation service for translating a first language into a second language,
A controller for extracting at least one concept word from the input sentence and selecting and presenting at least one similar sentence among candidate sentences stored in construction data when an input sentence of the first language is input;
A display unit configured to display a screen for inquiring whether to substitute the conceptual target word when the similar target sentence selected by the user is included in the similar sentence selected by the user among the at least one similar sentence;
The control unit substitutes a conceptual word included in the input sentence with the substitution target concept word,
The display unit displays a translation sentence in which a sentence including a substituted concept word is translated into the second language,
And the conceptual object to be substituted is a conceptual word belonging to the same concept as the conceptual word included in the selected similar sentence and representing a different object.
The method of claim 11,
The controller extracts a conceptual word registered in a conceptual word band dictionary, extracts an unregistered conceptual word according to an unregistered conceptual word extraction rule, and then tags the extracted at least one conceptual word.
The method of claim 12,
The controller may determine the concept of the noun using at least one of a verb and a verb combined with a noun.
The method of claim 11,
The controller may be configured to, when the input sentence is input, analyze a morpheme structure of the input sentence and then determine a sentence type of the input sentence using an analysis result of the morpheme structure.
The method of claim 14,
And the controller determines the sentence type by comparing the analysis result of the morpheme structure with a morpheme structure of a predefined regular expression.
The method of claim 14,
The controller may search for candidate sentences including some or all of keywords of the input sentence among candidate sentences belonging to a sentence type of the input sentence, identify a cluster distribution of candidate sentences including some or all of the keywords, and And selecting the at least one similar sentence within a cluster comprising the largest number of sentences.
The method of claim 16,
And the at least one similar sentence is selected using probabilities of generating the input sentence in each candidate sentence.
The method of claim 17,
The probability that the input sentence is generated in each candidate sentence, characterized in that determined by the following equation,

Where

Is similarity, said

Is a candidate sentence, said

Is the input sentence, said

Is the i th item constituting the input sentence,

Is the number of items constituting the input sentence,

And

Is a weight,

Above

A set of sentences having the same meaning as

Means the complete set of sentences,

Where

Is a candidate sentence, said

Is the i th item constituting the input sentence,

Means the number of items that make up the input sentence.

Where

And

Weights for determining a probability value, said

Is a candidate sentence, said

Above

Means a set of sentences with the same meaning as
The method of claim 11,
And the control unit reinforces the construction data by storing the input sentence.
The method of claim 11,
The construction data may include sentence information, mapping relationship information between a first language sentence and a second language sentence, category classification information of sentences, conceptual word information for sentence comparison, sentence type information, regular expression information for each sentence type, and sentence search. And at least one of index information for the cluster information and cluster information.

Images