Device, method and computer program for voice synthesis service

Abstract

A device providing a speech synthesis service may include an analysis unit that analyzes morphemes for an input sentence containing multiple words, a prosodic boundary information derivation unit that compares morphemes included in at least two adjacent words within the input sentence and derives prosodic boundary information between adjacent words based on the comparison result of the morphemes and a preset prosodic boundary information rule, an assignment unit that assigns the derived prosodic boundary information between adjacent words, and a speech synthesis unit that performs speech synthesis for the input sentence to which the prosodic boundary information has been assigned.

Description

Technology Field

[0001] The present invention relates to an apparatus, a method, and a computer program for providing a speech synthesis service. Background Technology

[0002] Speech synthesis technology, also commonly known as TTS (Text-To-Speech), is a technology used in applications that require human voice (e.g., public address systems, navigation, AI assistants, etc.) to play back the necessary voice without having actual human voices pre-recorded.

[0003] Conventional deep learning speech synthesis methods perform speech synthesis processing that breaks down an input sentence into pre-set word units (i.e., word units separated by spaces) and reads them.

[0004] For example, referring to FIG. 1, when the input sentence (100) “I will do my best to be your lifelong friend” is processed using a conventional deep learning speech synthesis method, a synthesized speech (102) is generated in which the lengths of the pauses between words are equal, such as “I, lifelong, friend, d, s

[0005] These conventional deep learning speech synthesis methods break down sentences into pre-set word units without analyzing the sentence structure, resulting in unnatural rhythm. Furthermore, if the pauses are incorrect, it is difficult to accurately convey the meaning of the sentence. Prior art literature

[0006] Korean Patent Publication No. 2004-0055288 (Published June 26, 2004) The problem to be solved

[0007] The present invention aims to solve the problems of the aforementioned prior art by assigning prosodic phrase boundary information between at least two adjacent words based on morphemes included in at least two adjacent words within an input sentence, and to perform speech synthesis for the input sentence to which the prosodic phrase boundary information has been assigned.

[0008] However, the technical problems that this embodiment aims to solve are not limited to the technical problems described above, and other technical problems may exist. means of solving the problem

[0009] As a technical means for achieving the technical problem described above, an apparatus for providing a speech synthesis service according to the first aspect of the present invention may include: an analysis unit that analyzes morphemes for an input sentence comprising a plurality of words; a prosodic boundary information derivation unit that compares morphemes included in at least two adjacent words within the input sentence and derives prosodic boundary information between the adjacent words based on the comparison result of the morphemes and a preset prosodic boundary information rule; an assignment unit that assigns the derived prosodic boundary information between the adjacent words; and a speech synthesis unit that performs speech synthesis for the input sentence to which the prosodic boundary information is assigned.

[0010] A method for providing a speech synthesis service performed by a speech synthesis service providing device according to a second aspect of the present invention may include: a step of analyzing morphemes for an input sentence comprising a plurality of words; a step of comparing morphemes included in at least two adjacent words within the input sentence and deriving prosodic boundary information between the adjacent words based on the comparison result of the morphemes and a preset prosodic boundary information rule; a step of assigning the derived prosodic boundary information between the adjacent words; and a step of performing speech synthesis for the input sentence to which the prosodic boundary information is assigned.

[0011] A computer program stored in a computer-readable recording medium comprising a sequence of instructions for providing a speech synthesis service according to a third aspect of the present invention may include a sequence of instructions that, when executed by a computing device, analyze morphemes for an input sentence comprising a plurality of words, compare morphemes included in at least two adjacent words within the input sentence, derive prosodic boundary information between the adjacent words based on the comparison result of the morphemes and a preset prosodic boundary information rule, assign the derived prosodic boundary information between the adjacent words, and perform speech synthesis for the input sentence to which the prosodic boundary information is assigned.

[0012] The above-described means for solving the problem are merely exemplary and should not be interpreted as intended to limit the invention. In addition to the exemplary embodiments described above, additional embodiments described in the drawings and the detailed description of the invention may exist. Effects of the invention

[0013] According to any one of the means for solving the problem of the present invention described above, the present invention can assign prosodic phrase boundary information between at least two adjacent words based on morphemes included in at least two adjacent words within an input sentence, and perform speech synthesis for the input sentence to which the prosodic phrase boundary information has been assigned.

[0014] Through this, the present invention can perform speech synthesis by adjusting the length of the pause between at least two adjacent words.

[0015] In addition, the present invention performs speech synthesis by adding prosodic phrase boundary information between at least two adjacent words as a prosodic phrase unit corresponding to a semantic unit of a sentence, thereby providing a more natural synthesized sound to the listener. Brief explanation of the drawing

[0016] Figure 1 is a diagram illustrating a conventional deep learning speech synthesis method. FIG. 2 is a block diagram of a speech synthesis service providing device according to one embodiment of the present invention. FIG. 3 is a diagram showing a rhythm phrase boundary information rule according to one embodiment of the present invention. FIGS. 4a to 4c are drawings illustrating a method for assigning prosodic phrase boundary information within an input sentence according to an embodiment of the present invention. FIG. 5 is a flowchart illustrating a method for assigning prosodic phrase boundary information within an input sentence according to an embodiment of the present invention. FIG. 6 is a flowchart illustrating a method for providing a speech synthesis service according to one embodiment of the present invention. Specific details for implementing the invention

[0017] Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0018] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected," but also cases where they are "electrically connected" with other components interposed between them. Furthermore, when a part is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0019] In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more hardware, and two or more units may be realized by one hardware.

[0020] Some of the operations or functions described herein as being performed by a terminal or device may instead be performed by a device connected to said terminal or device. Likewise, some of the operations or functions described as being performed by a device may also be performed by a terminal or device connected to said device.

[0021] Hereinafter, specific details for implementing the present invention will be described with reference to the attached configuration diagram or process flowchart.

[0022] FIG. 2 is a block diagram of a voice synthesis service providing device (20) according to one embodiment of the present invention.

[0023] The speech synthesis service providing device (20) can receive an input sentence that is the subject of speech synthesis. Such an input sentence may mean a sentence composed of at least two words. For example, the speech synthesis service providing device (20) can receive an input sentence in text form from a user or user terminal, or receive an input sentence in voice data form and convert it into text form.

[0024] The speech synthesis service providing device (20) can perform speech synthesis by adjusting the length of the gap between at least two adjacent words in the input sentence for which speech synthesis is requested.

[0025] Referring to FIG. 2, the speech synthesis service providing device (20) may include a prosodic rule generating unit (200), an analysis unit (210), a prosodic phrase boundary information derivation unit (220), an assignment unit (230), and a speech synthesis unit (240). However, the speech synthesis service providing device (20) illustrated in FIG. 2 is merely one embodiment of the present invention, and various modifications are possible based on the components illustrated in FIG. 2.

[0026] In the following, Figure 2 will be explained with reference to Figures 3 to 4c.

[0027] Generally, parts of speech for morphemes are classified into five categories (substantives, predicates, modifiers, relational words, and independent words), and more specifically into nine categories (nouns, pronouns, numerals, verbs, adjectives, determiners, adverbs, and particles). Here, nouns, pronouns, and numerals are included in substantives; verbs and adjectives are included in predicates; determiners and adverbs are included in modifiers; and particles are included in relational words.

[0028] In this invention, to express the role of morphemes more accurately, the five-word system is expanded to six-word systems by adding 'other words,' and the nine-word system is expanded to ten-word systems by adding the part of speech 'ending' to 'other words.' Here, the part of speech 'ending' refers to a morpheme that is attached after a verb.

[0029] In addition, the present invention allows for the detailed classification of morphemes by classifying 'nouns' into 'common nouns, proper nouns, and dependent nouns,' classifying 'adjectives' into 'demonstrative adjectives, adjective adjectives, and modifying adjectives,' and classifying 'adverbs' into 'adverbs modifying substantives, adverbs modifying verbs, and adverbs modifying modifiers.'

[0030] The prosodic rule generation unit (200) can set prosodic phrase boundary information rules between words based on part-of-speech information for multiple morphemes. Here, the part-of-speech information for multiple morphemes may include six words including nouns, predicates, modifiers, relational words, independent words, and other words. Here, the six words may include ten parts of speech including nouns, pronouns, numerals, verbs, adjectives, determiners, adverbs, particles, and endings. More specifically, the part of speech 'noun' may be subdivided into 'common nouns', 'proper nouns', and 'dependent nouns', the part of speech 'determiner' may be subdivided into 'demonstrative determiners', 'adjective determiners', and 'modifying determiners', and the part of speech 'adverb' may be subdivided into 'noun-modifying adverbs', 'predicate-modifying adverbs', and 'modifier-modifying adverbs'.

[0031] Referring to FIG. 3, the prosodic rule generation unit (200) can set a prosodic phrase boundary information rule (300) by selecting one of a first symbol (NP, Non Pause) indicating that no pausing is required, a second symbol (SP, Short Pause) indicating that basic pausing is required, and a third symbol (LP, Long Pause) indicating that long pausing is required, based on part-of-speech information for morphemes included in adjacent words.

[0032] The prosodic rule generation unit (200) can set a prosodic boundary information rule (300) by selecting one of the first symbol (NP), the second symbol (SP), and the third symbol (LP) as prosodic boundary information based on part-of-speech information for the left morpheme of the Nth word and part-of-speech information for the right morpheme of the N-1th word.

[0033] For example, the prosodic rule generation unit (200) can set a first symbol (NP) indicating that no pause is required between the Nth word and the N-1st word as prosodic boundary information, if the part of speech information for the left morpheme of the Nth word is 'common noun' or 'proper noun' and the part of speech information for the right morpheme of the N-1st word is 'adjective determiner'.

[0034] For example, the prosodic rule generation unit (200) can set a third symbol (LP) indicating that long pauses are required as prosodic boundary information between the Nth word and the N-1st word if the part of speech information for the left morpheme of the Nth word is 'common noun' or 'proper noun' and the part of speech information for the right morpheme of the N-1st word is 'adverb modifying a substantive'.

[0035] For example, the prosodic rule generation unit (200) can set a second symbol (SP) indicating that basic pauses are required as prosodic boundary information between the Nth word and the N-1st word if the part of speech information for the left morpheme of the Nth word is 'common noun' or 'proper noun' and the part of speech information for the right morpheme of the N-1st word is 'particle'.

[0036] When the analysis unit (210) receives an input sentence containing multiple words from a user or user terminal, it can analyze the morphemes of the input sentence.

[0037] The analysis unit (210) can analyze part-of-speech information for each word constituting the input sentence.

[0038] The prosodic phrase boundary information derivation unit (220) can compare morphemes included in at least two adjacent words within an input sentence. For example, if a previous word exists before the current word to be checked among multiple words of the input sentence, one of a substantive, a predicate, a modifier, or an independent word may appear in the left morpheme of the current word.

[0039] The prosodic phrase boundary information derivation unit (220) can compare the right morpheme of the N-1st word of the input sentence with the left morpheme of the Nth word of the input sentence.

[0040] The prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information between adjacent words based on the comparison result of at least two adjacent word morphemes in the input sentence and a preset prosodic phrase boundary information rule.

[0041] The prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information corresponding to the right morpheme of the N-1st word of the input sentence and the left morpheme of the Nth word of the input sentence from a pre-set prosodic phrase boundary information rule.

[0042] The prosodic phrase boundary information derivation unit (220) can set the prosodic phrase boundary information between the N-1st word and the Nth word in the pre-set prosodic phrase boundary information rule as a second symbol (SP) indicating that basic pause reading is required when there is no prosodic phrase boundary information corresponding to the right morpheme of the N-1st word of the input sentence and the left morpheme of the Nth word of the input sentence.

[0043] The prosodic phrase boundary information derivation unit (220) can change the prosodic phrase boundary information between the Nth word of the input sentence and the N-1st word of the input sentence based on the prosodic phrase boundary information between the N-2nd word of the input sentence and the N-1st word of the input sentence.

[0044] For example, even if the prosodic boundary information between the Nth word of the input sentence and the N-1st word of the input sentence is derived as a third symbol (LP) indicating that long pauses are required based on a pre-set prosodic boundary information rule, if the prosodic boundary information between the N-2nd word of the input sentence and the N-1st word of the input sentence has already been set as a third symbol (LP) indicating that long pauses are required, the prosodic boundary information derivation unit (220) can change the prosodic boundary information between the Nth word of the input sentence and the N-1st word of the input sentence to a second symbol (SP) indicating that basic pauses are required.

[0045] For example, referring to FIG. 3 and FIG. 4a together, when a first input sentence, “I will do my best so that I can become a lifelong friend,” is entered, the analysis unit (210) can perform morphological analysis on each of the first word of the first input sentence: “I” (401), corresponding to the first word; “lifelong” (403), corresponding to the second word; “friend” (405), corresponding to the third word; “will” (407), corresponding to the fourth word; “can” (409), corresponding to the fifth word; “to be” (411), corresponding to the sixth word; “to do my best” (413), corresponding to the seventh word; “all” (415), corresponding to the eighth word; and “I will do” (417), corresponding to the ninth word.

[0046] The prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('common noun') of 'lifelong' (403) corresponding to the second word of the first input sentence and the right morpheme ('particle') of 'I' (401) corresponding to the first word from a pre-set prosodic phrase boundary information rule (300).

[0047] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('common noun') of 'friend' (405) corresponding to the third word and the right morpheme ('particle') of 'lifelong' (403) corresponding to the second word from the pre-set prosodic phrase boundary information rule (300).

[0048] Next, the prosodic phrase boundary information derivation unit (220) can set the prosodic phrase boundary information corresponding to the left morpheme ('verb') of 'will' (407) corresponding to the 4th word in the pre-set prosodic phrase boundary information rule (300) and the right morpheme ('particle') of 'friend' (405) corresponding to the 3rd word as a second symbol (SP) indicating that basic pause reading is required.

[0049] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('dependent noun') of 'su' (409) corresponding to the 5th word and the right morpheme ('ending') of 'deul' (407) corresponding to the 4th word from the pre-set prosodic phrase boundary information rule (300).

[0050] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (first symbol (NP)) corresponding to the left morpheme ('verb') of 'is-do-ro' (411) corresponding to the 6th word and the right morpheme ('dependent noun') of 'su' (409) corresponding to the 5th word from the pre-set prosodic phrase boundary information rule (300).

[0051] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('common noun') of 'best' (413) corresponding to the 7th word and the right morpheme ('ending') of 'to be' (411) corresponding to the 6th word from the pre-set prosodic phrase boundary information rule (300).

[0052] Next, the prosodic phrase boundary information derivation unit (220) can change the prosodic phrase boundary information to be assigned between the 8th word 'da' (415) and the 7th word 'choeseon-eul' (413) to the 2nd symbol (SP), even if the prosodic phrase boundary information corresponding to the left morpheme ('adverb') of the 8th word 'da' (415) and the right morpheme ('particle') of the 7th word 'choeseon-eul' (413) is derived as the 3rd symbol (LP) based on the pre-set prosodic phrase boundary information rule (300), since the prosodic phrase boundary information assigned between the 7th word 'choeseon-eul' (413) and the 6th word 'itdorok' (411) is the 3rd symbol (LP).

[0053] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('verb') of 'I will do it' (417) corresponding to the ninth word and the right morpheme ('adverb') of 'da' (415) corresponding to the eighth word from the pre-set prosodic phrase boundary information rule (300).

[0054] For example, referring to FIG. 3 and FIG. 4b together, when a second input sentence is entered, such as ‘Please say this, channel up or channel down,’ the analysis unit (210) can perform morphological analysis on each of the first word of the second input sentence, ‘channel’ (419), the second word, ‘up’ (421), the third word, ‘or’ (423), the fourth word, ‘channel’ (425), the fifth word, ‘down’ (427), the sixth word, ‘like this’ (429), and the seventh word, ‘please say this’ (431).

[0055] The prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('common noun') of 'upward' (421) corresponding to the second word of the second input sentence and the right morpheme ('common noun') of 'channel' (419) corresponding to the first word from a pre-set prosodic phrase boundary information rule (300).

[0056] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('adverb') of 'or' (423) corresponding to the third word and the right morpheme ('particle') of 'upward' (421) corresponding to the second word from the pre-set prosodic phrase boundary information rule (300).

[0057] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('common noun') of 'channel' (425) corresponding to the fourth word and the right morpheme ('adverb') of 'or' (423) corresponding to the third word from the pre-set prosodic phrase boundary information rule (300).

[0058] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('common noun') of 'down' (427) corresponding to the 5th word and the right morpheme ('common noun') of 'channel' (425) corresponding to the 4th word from the pre-set prosodic phrase boundary information rule (300).

[0059] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('adverb') of 'like this' (429) corresponding to the 6th word and the right morpheme ('particle') of 'downward' (427) corresponding to the 5th word from the pre-set prosodic phrase boundary information rule (300).

[0060] Next, the prosodic phrase boundary information derivation unit (220) can change the prosodic phrase boundary information to be assigned between the 7th word 'please speak' (431) and the 6th word 'like this' (429) to the 2nd symbol (SP), even if the prosodic phrase boundary information corresponding to the left morpheme ('common noun') of the 7th word 'please speak' (431) and the right morpheme ('adverb') of the 6th word 'like this' (429) is derived as the 3rd symbol (LP) based on the pre-set prosodic phrase boundary information rule (300), since the prosodic phrase boundary information assigned between the 6th word 'like this' (429) and the 5th word 'downward' (427) is the 3rd symbol (LP).

[0061] For example, referring to FIG. 3 and FIG. 4c together, when a third input sentence, “It seems like the night is short and the morning comes quickly,” is input, the analysis unit (210) can perform morphological analysis on each of the first word of the third input sentence, “night is” (433), the second word, “short” (435), the third word, “morning is” (437), the fourth word, “quickly” (439), the fifth word, “coming” (441), the sixth word, “thing” (443), and the seventh word, “seems like” (445).

[0062] The prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('adjective') of 'short' (435) corresponding to the second word of the third input sentence and the right morpheme ('particle') of 'night' (433) corresponding to the first word from a pre-set prosodic phrase boundary information rule (300).

[0063] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (third symbol (LP)) corresponding to the left morpheme ('common noun') of 'morning' (437) corresponding to the third word and the right morpheme ('ending') of 'short' (435) corresponding to the second word from the pre-set prosodic phrase boundary information rule (300).

[0064] Next, the prosodic phrase boundary information derivation unit (220) can change the prosodic phrase boundary information to be assigned between the fourth word 'geumbang' (439) and the third word 'achim-eun' (437) to the second symbol (SP), even though the prosodic phrase boundary information is derived as the third symbol (LP) for the left morpheme ('adverb') of the fourth word 'geumbang' (439) and the right morpheme ('particle') of the third word 'achim-eun' (437) based on the pre-set prosodic phrase boundary information rule (300), since the prosodic phrase boundary information assigned to the third word 'achim-eun' (437) and the second word 'jjalgo' (435) is the third symbol (LP).

[0065] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (second symbol (SP)) corresponding to the left morpheme ('verb') of 'coming' (441) corresponding to the 5th word and the right morpheme ('adverb') of 'immediately' (439) corresponding to the 4th word from the pre-set prosodic phrase boundary information rule (300).

[0066] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (first symbol (NP)) corresponding to the left morpheme ('dependent noun') of 'thing' (443) corresponding to the 6th word and the right morpheme ('ending') of 'coming' (441) corresponding to the 5th word from the pre-set prosodic phrase boundary information rule (300).

[0067] Next, the prosodic phrase boundary information derivation unit (220) can derive prosodic phrase boundary information (first symbol (NP)) corresponding to the left morpheme ('adjective') of 'it' (445) corresponding to the 7th word and the right morpheme ('dependent noun') of 'thing' (443) corresponding to the 6th word from the pre-set prosodic phrase boundary information rule (300).

[0068] The assignment unit (230) can assign the derived prosodic phrase boundary information between adjacent words. For example, referring to FIG. 4a, the assignment unit (230) assigns a second symbol (SP) between 'jega' (401), corresponding to the first word, and 'pyeongsaengui' (403), corresponding to the second word, assigns a second symbol (SP) between 'pyeongsaengui' (403), corresponding to the second word, and 'chinguga' (405), corresponding to the third word, assigns a second symbol (SP) between 'chinguga' (405), corresponding to the third word, and 'deul' (407), corresponding to the fourth word, assigns a third symbol (LP) between 'deul' (407), corresponding to the fourth word, and 'su' (409), corresponding to the fifth word, and assigns a third symbol between 'su' (409), corresponding to the fifth word, and 'itdorok' (411), corresponding to the sixth word. A symbol (LP) can be assigned, a third symbol (LP) can be assigned between 'to be' (411) corresponding to the 6th word and 'to do one' (413) corresponding to the 7th word, a second symbol (SP) can be assigned between 'to do one' (413) corresponding to the 7th word and 'all' (415) corresponding to the 8th word, and a third symbol (LP) can be assigned between 'all' (415) corresponding to the 8th word and 'I will do it' (417) corresponding to the 9th word.

[0069] The speech synthesis unit (240) can perform speech synthesis for an input sentence to which prosodic phrase boundary information is assigned.

[0070] Meanwhile, a person skilled in the art will fully understand that the prosody rule generation unit (200), the analysis unit (210), the prosody phrase boundary information derivation unit (220), the assignment unit (230), and the speech synthesis unit (240) may each be implemented separately, or that one or more of these may be implemented in an integrated manner.

[0071] FIG. 5 is a flowchart illustrating a method for assigning prosodic phrase boundary information within an input sentence according to an embodiment of the present invention.

[0072] Referring to FIG. 5, in step S501, the speech synthesis service providing device (20) can derive the number of words in the input sentence.

[0073] In step S503, if the number of words in the input sentence is greater than or equal to a preset threshold (e.g., 4), the speech synthesis service providing device (20) can check the part of speech information for the left morpheme of the Nth word of the input sentence in step S505.

[0074] In step S507, the speech synthesis service providing device (20) can check part-of-speech information for the right morpheme of the N-1st word of the input sentence.

[0075] In step S509, the speech synthesis service providing device (20) can derive prosodic boundary information between the Nth and N-1st words of the input sentence according to the prosodic boundary information rule. Specifically, the speech synthesis service providing device (20) can derive prosodic boundary information corresponding to the part of speech information for the left morpheme of the Nth word of the input sentence and the part of speech information for the right morpheme of the N-1st word from the prosodic boundary information rule.

[0076] In step S511, the speech synthesis service providing device (20) can assign the derived prosodic phrase boundary information between the Nth and N-1st words of the input sentence.

[0077] In the first embodiment, if the prosodic phrase boundary information corresponding to the part of speech information for the left morpheme of the Nth word of the input sentence and the part of speech information for the right morpheme of the N-1st word does not exist in the prosodic phrase boundary information rule, the speech synthesis service providing device (20) may assign a second symbol (SP) indicating that basic pause reading is required between the Nth and N-1st words of the input sentence.

[0078] In the second embodiment, even if the prosodic boundary information to be assigned between the Nth and N-1st words of an input sentence is a third symbol (LP) indicating that long pauses are required, the speech synthesis service providing device (20) can change the prosodic boundary information to be assigned between the Nth and N-1st words from the third symbol (LP) to the second symbol (SP) if the prosodic boundary information assigned between the N-1st and N-2nd words of an input sentence has already been assigned as the third symbol (LP).

[0079] In the description above, steps S501 to S511 may be further divided into additional steps or combined into fewer steps, depending on an embodiment of the present invention. Additionally, some steps may be omitted as necessary, and the order between steps may be changed.

[0080] FIG. 6 is a flowchart illustrating a method for providing a speech synthesis service according to one embodiment of the present invention.

[0081] Referring to FIG. 6, in step S601, the speech synthesis service providing device (20) can analyze morphemes for an input sentence containing multiple words.

[0082] In step S603, the speech synthesis service providing device (20) can compare morphemes included in at least two adjacent words within an input sentence and derive prosodic phrase boundary information between adjacent words based on the comparison result for the morphemes and a preset prosodic phrase boundary information rule.

[0083] In step S605, the speech synthesis service providing device (20) can assign the derived prosodic phrase boundary information between adjacent words.

[0084] In step S607, the speech synthesis service providing device (20) can perform speech synthesis for an input sentence to which prosodic phrase boundary information is assigned.

[0085] In the description above, steps S601 to S607 may be further divided into additional steps or combined into fewer steps, depending on an embodiment of the present invention. Additionally, some steps may be omitted as necessary, and the order between steps may be changed.

[0086] One embodiment of the present invention may also be implemented in the form of a recording medium comprising computer-executable instructions, such as program modules executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, and both removable and non-removable media. Additionally, a computer-readable medium may include all computer storage media. A computer storage medium includes both volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0087] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0088] The scope of the present invention is defined by the claims set forth below rather than by the detailed description, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0089] 20: Speech synthesis service providing device 200: Prosody Rule Generator 210: Analysis Department 220: Prosodic phrase boundary information derivation section 230: Allocation Department 240: Speech Synthesizer

Claims

Claim 1 A device for providing a speech synthesis service comprises: an analysis unit for analyzing morphemes of an input sentence including a plurality of words; a prosodic phrase boundary information derivation unit for comparing morphemes included in at least two adjacent words within the input sentence and deriving prosodic phrase boundary information between the adjacent words based on the comparison result of the morphemes and a preset prosodic phrase boundary information rule; and an assignment unit for assigning the derived prosodic phrase boundary information between the adjacent words. A speech synthesis service providing device comprising a speech synthesis unit that performs speech synthesis for an input sentence to which the above-mentioned prosodic phrase boundary information is assigned, wherein the prosodic phrase boundary information derivation unit determines whether prosodic phrase boundary information corresponding to the right morpheme of the N-1st word of the input sentence and the left morpheme of the Nth word of the input sentence exists in the prosodic phrase boundary information rule, and if the above-mentioned prosodic phrase boundary information does not exist in the prosodic phrase boundary information rule, the above-mentioned prosodic phrase boundary information is set as a second symbol (SP, Short Pause) indicating that basic pause reading is required, and if the above-mentioned prosodic phrase boundary information exists in the prosodic phrase boundary information rule, the above-mentioned prosodic phrase boundary information between the N-1st word and the Nth word is determined according to the above-mentioned prosodic phrase boundary information rule. Claim 2 A speech synthesis service providing device according to claim 1, further comprising a prosodic rule generating unit that sets prosodic phrase boundary information rules between words based on part-of-speech information for a plurality of morphemes. Claim 3 A speech synthesis service providing device according to claim 2, wherein the prosodic rule generating unit sets the prosodic phrase boundary information rule by selecting one of a first symbol (NP, Non Pause) indicating that pausing is not required, a second symbol (SP, Short Pause) indicating that pausing is required, and a third symbol (LP, Long Pause) indicating that long pausing is required, based on part-of-speech information for morphemes included in adjacent words, as prosodic phrase boundary information. Claim 4 delete Claim 5 delete Claim 6 A speech synthesis service providing device according to claim 1, wherein the prosodic phrase boundary information derivation unit changes the prosodic phrase boundary information between the Nth word of the input sentence and the N-1st word of the input sentence based on the prosodic phrase boundary information between the N-2nd word of the input sentence and the N-1st word of the input sentence. Claim 7 A method for providing a speech synthesis service performed by a speech synthesis service providing device comprises: a step of analyzing morphemes for an input sentence including a plurality of words; a step of comparing morphemes included in at least two adjacent words within the input sentence and deriving prosodic boundary information between the adjacent words based on the comparison result for the morphemes and a preset prosodic boundary information rule; a step of assigning the derived prosodic boundary information between the adjacent words; and a step of performing speech synthesis for the input sentence to which the prosodic boundary information is assigned, wherein the method comprises: a step of determining whether prosodic boundary information corresponding to the right morpheme of the N-1st word of the input sentence and the left morpheme of the Nth word of the input sentence exists in the prosodic boundary information rule; and, if the prosodic boundary information does not exist in the prosodic boundary information rule, a step of setting the prosodic boundary information to a second symbol (SP, Short Pause) indicating that basic pause reading is required. A method for providing a speech synthesis service, further comprising the step of determining prosodic boundary information between the N-1th word and the Nth word according to the prosodic boundary information rule when the prosodic boundary information exists in the prosodic boundary information rule. Claim 8 A method for providing a speech synthesis service according to claim 7, further comprising the step of setting rules for prosodic phrase boundary information between words based on part-of-speech information for a plurality of morphemes. Claim 9 A method for providing a speech synthesis service according to claim 8, further comprising the step of setting a prosodic phrase boundary information rule by selecting one of a first symbol (NP, Non Pause) indicating that pausing is not necessary, a second symbol (SP, Short Pause) indicating that pausing is necessary, and a third symbol (LP, Long Pause) indicating that long pausing is necessary, based on part-of-speech information for morphemes included in adjacent words, as prosodic phrase boundary information. Claim 10 delete Claim 11 delete Claim 12 A method for providing a speech synthesis service, wherein, in claim 7, the step of deriving prosodic boundary information between adjacent words includes the step of changing prosodic boundary information between the Nth word of the input sentence and the N-1st word of the input sentence based on prosodic boundary information between the N-2nd word of the input sentence and the N-1st word of the input sentence. Claim 13 A computer program stored on a computer-readable recording medium comprising a sequence of instructions for providing a speech synthesis service, wherein, when the computer program is executed by a computing device, it analyzes morphemes for an input sentence containing multiple words, compares morphemes included in at least two adjacent words within the input sentence, derives prosodic boundary information between the adjacent words based on the comparison result for the morphemes and a preset prosodic boundary information rule, assigns the derived prosodic boundary information between the adjacent words, performs speech synthesis for the input sentence to which the prosodic boundary information is assigned, determines whether prosodic boundary information corresponding to the right morpheme of the N-1th word of the input sentence and the left morpheme of the Nth word of the input sentence exists in the prosodic boundary information rule, if the prosodic boundary information does not exist in the prosodic boundary information rule, sets the prosodic boundary information as a second symbol (SP, Short Pause) indicating that basic pausing is required, and if the prosodic boundary information exists in the prosodic boundary information rule, the prosodic A computer program stored on a computer-readable recording medium, comprising a sequence of instructions that determine prosodic phrase boundary information between the N-1th word and the Nth word according to boundary information rules.

Images