Lyric processing method, device and system based on agent

Abstract

The invention provides a lyric processing method, device and system based on an agent, and relates to the technical field of computer application, and the method comprises the steps: obtaining a lyric text to be processed; calling a sentence segmentation agent to carry out sentence segmentation processing; identifying an escort mode and an accent position of each lyric clause through a statistical agent; calling an alignment agent to carry out slot filling processing on the lyric clauses; performing syllable segmentation on translations corresponding to the lyric clauses, and filling the slots with the segmented syllables to obtain translation filling information; running a pre-configured dynamic optimization strategy by aligning the intelligent agent, and optimizing the filling condition of each syllable; and embedding the optimized translation filling information into the melody to generate synthetic music. According to the agent-based lyric processing method, device and system provided by the invention, the whole lyric processing process is completed through the cooperation of a plurality of agents, so that the translated text has more rhythm sensation and semantic aesthetic feeling, and the propagation effect and artistic value of the translated text are further improved.

Description

Technical Field

[0001] This invention relates to the technical field of computer applications, and in particular to a method, apparatus and system for processing lyrics based on intelligent agents. Background Technology

[0002] In recent years, generative artificial intelligence (GAI) has made groundbreaking progress in areas such as text generation, speech synthesis, and image processing. This is particularly evident in the large models of OpenAI's GPT (Generative Pre-trained Transformer) series, which have triggered a profound transformation in content creation. Simultaneously, short video platforms have driven a new explosion of content combining music and AI. One highly popular format is the translation or adaptation of Chinese song lyrics into English lyrics, followed by a new re-singing and video production, creating "AI song cover short videos." This format satisfies both visual and auditory senses and has become an important new channel for music dissemination and cultural export.

[0003] However, most lyric translations currently available on the market are still mainly mechanical literal translations, homophonic substitutions, or crude rhymes, lacking rhythm, semantic beauty, and singability, which seriously limits their dissemination and artistic value. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a lyrics processing method, apparatus and system based on intelligent agents to alleviate the above-mentioned technical problems.

[0005] In a first aspect, embodiments of the present invention provide a lyrics processing method based on an intelligent agent, characterized in that it is applied to a lyrics processing system, wherein the lyrics processing system operates a sentence segmentation intelligent agent, a statistical intelligent agent, and an alignment intelligent agent that work in concert; the method includes: acquiring lyrics text to be processed; calling the sentence segmentation intelligent agent to perform sentence segmentation processing on the lyrics text to obtain lyrics phrases, and, marking each lyrics phrase with sentence segmentation, obtaining sentence segmentation results, and inputting the sentence segmentation results to the statistical intelligent agent; identifying the rhyme pattern of each lyrics phrase through the statistical intelligent agent, and, based on the rhyme pattern and the sentence segmentation results, marking the stress position of the lyrics phrase; calling the alignment intelligent agent, and, through the... The alignment agent performs word-by-word slot filling on the lyrics phrases according to pre-set slots to obtain the original placeholder image corresponding to the lyrics phrases; and, the translation corresponding to the lyrics phrases is segmented into syllables to obtain a syllable sequence containing multiple syllables, and each syllable in the syllable sequence is filled into the slot one by one to obtain the translation filling information corresponding to the translation; the alignment agent runs a pre-configured dynamic optimization strategy to optimize the filling of each syllable in the translation filling information based on the accent position and the original placeholder image; the optimized translation filling information is embedded into the melody corresponding to the lyrics text to be processed to generate the synthesized music corresponding to the lyrics text to be processed.

[0006] In conjunction with the first aspect, this embodiment of the invention provides a first possible implementation of the first aspect, wherein the step of optimizing the filling of each syllable in the translated filling information based on the stress position and the original placeholder map by running a pre-configured dynamic optimization strategy through the alignment agent includes: calculating the cost index corresponding to each syllable in the translated filling information in the currently filled slot based on the stress position according to the stress position by the alignment agent based on the original placeholder map; wherein the cost index is the cost index required for the comprehensive cost index of the dynamic optimization strategy; performing a weighted calculation on each cost index to obtain a comprehensive cost index; and gradually optimizing the slot filled by the syllable with the minimum comprehensive cost index as the optimization objective until the comprehensive cost index reaches the minimum value, thereby obtaining the optimized translated filling information.

[0007] In conjunction with the first possible implementation of the first aspect, this embodiment of the invention provides a second possible implementation of the first aspect, wherein the aforementioned cost index includes: a time value matching cost, used to characterize the difference between the syllable duration and the slot duration; an accent matching cost, used to characterize whether the stress of the syllable matches the strong beat of the slot; an overflow and insufficiency penalty cost, used to characterize the penalty cost when the syllable crosses too many slots or exceeds the remaining slots; and a semantic integrity cost, used to characterize whether the semantic segmentation of the syllable is fragmented; the comprehensive cost index is the result obtained by weighted summation of each of the cost indices.

[0008] In conjunction with the first aspect, this embodiment of the invention provides a third possible implementation of the first aspect, wherein the above method further includes: optimizing the translation corresponding to the lyrics according to pre-configured optimization conditions; wherein the optimization conditions include at least one of the following conditions: slot utilization optimization conditions, rhythmic feature optimization conditions, end rhyme optimization conditions, and imagery optimization conditions.

[0009] In conjunction with the third possible implementation of the first aspect, this embodiment of the invention provides a fourth possible implementation of the first aspect, wherein the step of optimizing the translation corresponding to the lyrics according to the pre-configured optimization conditions includes: if the optimization conditions include slot utilization optimization conditions, then calling the alignment agent to calculate the slot utilization rate corresponding to the translation based on the translation filling information; determining whether the slot utilization rate meets the pre-configured threshold range; if yes, determining that the translation meets the slot utilization optimization conditions; if no, generating adjustment information and feeding the adjustment information back to the sentence segmentation agent to prompt the sentence segmentation agent to re-segment the lyrics text.

[0010] In conjunction with the third possible implementation of the first aspect, this embodiment of the invention provides a fifth possible implementation of the first aspect, wherein the above-mentioned lyrics processing system further includes a morpheme agent; the step of optimizing the translation corresponding to the lyrics according to pre-configured optimization conditions further includes: if the optimization conditions include prosodic feature optimization conditions, then calling the morpheme agent to perform morpheme decomposition on each lyric phrase to obtain a first speech distribution feature corresponding to the lyric phrase; wherein the first speech distribution feature is used to characterize the frequency and distribution features of initials and finals in the lyric phrase; and, performing phoneme segmentation on the translation corresponding to the lyric phrase to obtain the translation pair. The second speech distribution feature is used to represent the frequency and distribution characteristics of vowels and consonants in the translation. A preset number of phonetic features are extracted from the first speech distribution feature in descending order of frequency, and a preset number of letter features are extracted from the second speech distribution feature. Based on the phonetic features and the letter features, the similarity between the first speech distribution feature and the second speech distribution feature is calculated. If the similarity reaches a preset similarity threshold, the translation is determined to satisfy a preset prosodic feature; otherwise, the translation is determined not to satisfy the prosodic feature, and an optimization prompt is generated to prompt prosodic adjustment of the translation.

[0011] In conjunction with the fifth possible implementation of the first aspect, this embodiment of the invention provides a sixth possible implementation of the first aspect, wherein the step of optimizing the translation corresponding to the lyrics according to the pre-configured optimization conditions further includes: if the optimization conditions include a rhyme optimization condition, then calling the morpheme agent to determine whether the ending sound of the translation and the lyrics satisfy the rhyme constraint mechanism; if yes, determining that the translation satisfies the rhyme optimization condition; if no, generating optimization prompt information, wherein the optimization prompt information is used to prompt adjustments to the translation.

[0012] In conjunction with the third possible implementation of the first aspect, this embodiment of the invention provides a seventh possible implementation of the first aspect, wherein the above-mentioned lyrics processing system further includes an imagery agent; the step of optimizing the translation corresponding to the lyrics according to pre-configured optimization conditions further includes: if the optimization conditions include imagery optimization conditions, then calling the imagery agent to extract imagery information contained in the lyrics, wherein the imagery information includes the type and number of intentional words contained in the lyrics; and: extracting imagery reproduction information of the translation corresponding to the lyrics; calculating the imagery reproduction rate of the translation based on the imagery reproduction information and the imagery information; if the imagery reproduction rate reaches a pre-set reproduction rate threshold, then determining that the translation meets the imagery optimization conditions; if the imagery reproduction rate is lower than the pre-set reproduction rate threshold, then adjusting the imagery of the translation.

[0013] Secondly, embodiments of the present invention also provide an agent-based lyrics processing device, applied to a lyrics processing system, wherein the lyrics processing system operates a sentence segmentation agent, a statistical agent, and an alignment agent that work in concert; the device includes: an acquisition module, used to acquire lyrics text to be processed; a sentence segmentation agent module, used to call the sentence segmentation agent to perform sentence segmentation processing on the lyrics text to obtain lyrics phrases, and to perform sentence segmentation annotation on each lyrics phrase to obtain sentence segmentation results, and input the sentence segmentation results to the statistical agent; a statistical agent module, used to identify the rhyme pattern of each lyrics phrase through the statistical agent, and to mark the stress position of the lyrics phrase based on the rhyme pattern and the sentence segmentation results; and an alignment agent module, used to adjust... The alignment agent is used to fill the lyrics phrases word by word according to the pre-set slots to obtain the original placeholder map corresponding to the lyrics phrases; and the translation corresponding to the lyrics phrases is segmented into syllables to obtain a syllable sequence containing multiple syllables. Each syllable in the syllable sequence is filled into the slot one by one to obtain the translation filling information corresponding to the translation; the alignment agent runs a pre-configured dynamic optimization strategy to optimize the filling of each syllable in the translation filling information based on the accent position and the original placeholder map; the synthesis module is used to embed the optimized translation filling information into the melody corresponding to the lyrics text to be processed to generate the synthesized music corresponding to the lyrics text to be processed.

[0014] Thirdly, embodiments of the present invention also provide a lyrics processing system, wherein the lyrics processing system operates a synergistic intelligent agent for sentence segmentation, a statistical intelligent agent, and an alignment intelligent agent, for executing the agent-based lyrics processing method described in the first aspect.

[0015] The embodiments of the present invention bring the following beneficial effects: This invention provides a method, apparatus, and system for processing lyrics based on an intelligent agent. The system can acquire lyrics text to be processed, invoke a sentence segmentation intelligent agent to segment the lyrics text, and, through a statistical intelligent agent, identify the rhyme pattern of each lyric phrase. Based on the rhyme pattern and the sentence segmentation results, it marks the stress position of each lyric phrase. Then, it invokes an alignment intelligent agent to fill each lyric phrase into pre-set slots, obtaining the original placeholder image corresponding to each lyric phrase. Finally, it segments the translated text corresponding to each lyric phrase into syllables, obtaining a syllable sequence containing multiple syllables, and fills each syllable in the syllable sequence into a slot, obtaining the translation filling information corresponding to the translated text. Then, by aligning the agents and running pre-configured dynamic optimization strategies, the filling of each syllable in the translation filling information is optimized based on the accent position and the original placeholder map. The optimized translation filling information is then embedded into the melody corresponding to the lyrics text to be processed, generating the synthesized music corresponding to the lyrics text to be processed. The entire lyrics processing process is implemented based on multiple collaborative agents of the lyrics processing system, rather than mechanically translating directly. Moreover, the optimization process of syllable filling can flexibly adjust the correspondence between syllables and slots while ensuring overall singability, making the final translation more rhythmic and semantically beautiful, thereby enhancing the translation's dissemination ability and artistic value.

[0016] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention are realized and obtained in accordance with the structures particularly pointed out in the description, claims and drawings.

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0019] Figure 1 A flowchart of an agent-based lyrics processing method provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a multi-agent collaborative lyrics processing method provided in an embodiment of the present invention; Figure 3This is a schematic diagram of the structure of a lyrics processing device based on an intelligent agent, provided in an embodiment of the present invention. Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Currently, the mainstream methods for translating lyrics include: (1) A method for literal translation of lyrics based on a general machine translation model; This method, currently the most common approach to lyrics translation, relies on general-purpose Neural Machine Translation (NMT) models. These methods are trained on large-scale parallel corpora and utilize encoder-decoder structures and attention mechanisms to translate lyrics line by line. For example, using a Transformer architecture, they demonstrate stable performance across high-resource language pairs such as English-Chinese and English-Japanese. Additionally, there is the DeepL model, which, through optimized neural networks and contextual modeling, outperforms traditional statistical translation methods in maintaining semantic coherence.

[0022] (2) Dedicated translation tools based on lyrical context and musical characteristics; Unlike general translation tools, some platforms have developed specialized translation tools for song lyrics. These methods not only rely on text translation models but also optimize for the rhythm, rhyme, and cultural context of the lyrics. For example, they offer multilingual versions of lyrics, with some translations reviewed by humans to ensure both original meaning and singability. Furthermore, user community contributions and review mechanisms, combined with machine translation and human revision, can enhance the naturalness and artistry of the translations.

[0023] (3) Intelligent translation methods based on multimodal and semantic understanding; In recent years, with the development of multimodal large-scale models, researchers have attempted to apply visual language models (such as CLIP, mBART, and mT5) to lyric translation tasks. These methods not only process the lyric text but also incorporate emotional tags from the song, vocal tone, and even music video visuals for semantic enhancement. For example, some experimental platforms utilize multimodal Transformer models to align lyrics with musical emotional features, thereby generating translations that better match the song's atmosphere.

[0024] However, the machine translation models described in (1) above lack singability. For example, existing lyric translation methods based on neural machine translation (NMT) mainly focus on semantic alignment and syntactic fluency, but ignore the formal constraints of lyrics, such as rhyme, rhythm, number of syllables and stress position. Therefore, although the translation is correct in literal sense, it is difficult to match the melody and lacks "singability". This defect is particularly prominent in cover or adaptation scenarios, resulting in the translation being unusable for singing. The dedicated translation tools described in (2) above mostly rely on user crowdsourcing and human proofreading to make up for the shortcomings of machine translation. This method can guarantee a certain quality for popular songs, but its defects are: uneven quality, with significant differences in the language ability and aesthetic standards of different users, resulting in fragmented translation styles; insufficient coverage, with unpopular songs and lyrics in less common languages ​​often lacking sufficient user contributions and remaining in a state of "no translation" or "low-quality translation" for a long time; and delayed updates, with the translation of lyrics for new songs or live versions often relying on community popularity and unable to guarantee real-time updates. The multimodal and semantic understanding methods described in (3) above often require excessive investment, making mass production and promotion difficult. In recent years, multimodal translation methods have attempted to combine information such as lyrics, audio emotions, and music video visuals to enhance semantic depth and emotional delivery. However, these methods have significant engineering and resource bottlenecks: high computational costs, requiring large-scale computing power for training and inference, making large-scale deployment on commercial platforms difficult; data scarcity, with extremely limited high-quality multimodal lyrics parallel corpora, restricting the generalization ability of the models; and difficulty in implementation, performing well in laboratory environments but struggling to achieve mass production and large-scale service in practical applications.

[0025] Based on this, the present invention provides a lyrics processing method, apparatus and system based on intelligent agents, which can effectively alleviate the above-mentioned technical problems.

[0026] To facilitate understanding of this embodiment, a detailed description of an agent-based lyrics processing method disclosed in this embodiment of the invention will be provided first.

[0027] In a possible implementation manner, an embodiment of the present invention provides an agent-based lyrics processing method, which is applied to a lyrics processing system. The lyrics processing system runs a sentence-breaking agent, a statistics agent, and an alignment agent that cooperate with each other; that is, in the embodiment of the present invention, the processing of lyrics is achieved through the collaborative action of multiple agents.

[0028] Specifically, as Figure 1 shown in the flowchart of an agent-based lyrics processing method, the method includes the following steps: Step S102, obtain the lyrics text to be processed; Step S104, call the sentence-breaking agent to perform sentence-breaking processing on the lyrics text to obtain lyric clauses, and perform sentence-breaking annotation on each lyric clause to obtain a clause result, and input the clause result into the statistics agent; Step S106, identify the rhyming pattern of each lyric clause through the statistics agent, and based on the rhyming pattern, combine the clause result to mark the stress position of the lyric clause; Step S108, call the alignment agent, and through the alignment agent, perform slot filling on each lyric clause word by word according to the pre-set slots to obtain the original placeholder map corresponding to the lyric clause; and, perform syllable segmentation on the translation corresponding to the lyric clause to obtain a syllable sequence containing multiple syllables, and fill each syllable in the syllable sequence into the slots one by one to obtain the translation filling information corresponding to the translation; In actual use, the lyrics text to be processed obtained in the above step S102 usually refers to Chinese lyrics, and the Chinese lyrics need to be mapped word by word to the slots in the subsequent steps to form an original "word-slot" placeholder map, that is, the original placeholder map. Therefore, in the embodiment of the present invention, for the obtained lyrics text, it is necessary to first call the sentence-breaking agent to perform sentence-breaking processing to obtain lyric clauses, and the lyric clauses are also called bars, which represent a sentence of lyrics. For example, the lyric clause "Welcome another dawn" can be used as a bar, and then further call the sentence-breaking agent to perform sentence-breaking annotation to mark the breathing points, beat boundaries, etc. of the lyric clause. Moreover, the statistics agent further counts the frequencies of the initials and finals of each word in the lyrics and identifies the rhyming pattern, and then combines the breathing points, beats, etc. marked by the clause result to mark the stress position of the lyric clause.

[0029] Furthermore, in order to facilitate the slot filling by the alignment agent, usually, the slots matching the lyrics text to be processed can be determined in advance, generally eight slots. For example, the lyric clause "Welcome another dawn" corresponds to "12345677", and each number represents a word. The meaning of the two 7s is that the word "dawn" sings two slots, and the second 7 actually occupies the slot of 8.

[0030] Furthermore, the translation in step S108 typically refers to an English translation, i.e., translated English lyrics. Generally, the method for segmenting the translation into syllables involves breaking down words into their smallest vocal units and filling them sequentially into the corresponding slots. Through this "syllable-slot filling" method, the translated English lyrics maintain rhythmic consistency with the original lyrics. As long as the syllables are correctly inserted into the slots, the translation is guaranteed to be singable. This process is entirely consistent with the logic of filling musical notes in sheet music, i.e., "words filling slots" corresponds to "notes filling beats."

[0031] In practice, the process typically involves dividing the song into measures based on its beat and setting the number of slots, such as eight. Next, an alignment agent is invoked to map the Chinese lyrics word-by-word into the slots according to measures, creating the initial placeholder diagram. Then, the translated text is segmented into syllables to obtain a syllable sequence. Subsequently, the syllables are sequentially filled into the slots to form the aligned translation. Finally, the slot filling results can be checked to verify the completeness of the syllable filling, the rationality of the rhyme and stress positions, etc. If there are insufficient or redundant slots, corrections are made by adjusting the vocabulary or sentence structure of the translated text. This ensures that the final output translation is not only semantically faithful to the original text but also naturally matches the melody in rhythm and rhyme, meeting the needs of cross-language singing.

[0032] Step S110: The alignment agent runs a pre-configured dynamic optimization strategy to optimize the filling of each syllable in the translation filling information based on the accent position and the original placeholder map. Step S112: Embed the optimized translation fill information into the melody corresponding to the lyrics text to be processed, and generate the synthesized music corresponding to the lyrics text to be processed.

[0033] The dynamic optimization strategy in this embodiment of the invention is actually a dynamic programming (DP) method. This algorithm can flexibly adjust the correspondence between syllables and slots while ensuring overall singability.

[0034] In practical implementation, to achieve the above dynamic optimization strategy, a corresponding cost function or cost index needs to be designed to implement the dynamic optimization process. Specifically, the optimization process in step S110 includes the following: the alignment agent calculates the cost index corresponding to each syllable in the translation filling information in the currently filled slot based on the original placeholder map and the stress position; wherein, the cost index is the cost index required for the comprehensive cost index of the dynamic optimization strategy; each cost index is weighted and calculated to obtain the comprehensive cost index; with the minimum comprehensive cost index as the optimization objective, the slots filled by the syllables are gradually optimized until the comprehensive cost index is minimized, and the optimized translation filling information is obtained.

[0035] The cost metrics in this embodiment of the invention include: timing matching cost, which characterizes the difference between the duration of a syllable and the duration of a slot; stress matching cost, which characterizes whether the stress of a syllable matches the strong beat of a slot; overflow and insufficiency penalty cost, which characterizes the penalty cost when a syllable crosses too many slots or exceeds the remaining slots; and semantic integrity cost, which characterizes whether the semantic segmentation of a syllable is fragmented. The above comprehensive cost metrics are obtained by weighted summation of each cost metric.

[0036] Specifically, the goal of the aforementioned timing matching cost is to ensure that the syllable length matches the slot length, avoiding problems such as "not being able to sing" or "running out of time." In practical use, it can be calculated by comparing the estimated duration of the syllable with the available duration of the slot. For example, the absolute value of the difference between the estimated duration of the syllable and the available duration of the slot can be used as a ratio coefficient to the available duration of the slot to represent the degree of matching between the estimated duration of the syllable and the available duration of the slot, i.e., the aforementioned timing matching cost. The smaller this value, the better the estimated duration of the syllable matches the available duration of the slot. The corresponding formula can be expressed as: Cslot = |duration(e_k)-available_duration(j)| / available_duration(j); Here, duration(e_k) represents the estimated duration of the syllable, and available_duration(j) represents the available duration of the slot.

[0037] Furthermore, the aforementioned stress matching cost aims to ensure that the stressed syllables in the translation fall on the strong beats of the original lyrics. Specifically, referring to the stress position, if the stressed syllable matches a strong beat, the difference between the stressed syllable and the strong beat is approximately zero, resulting in a low stress matching cost. If the unstressed syllable falls on a strong beat, the stress matching cost is moderate. If the stressed syllable falls on a weak beat, the translation is less singable, resulting in a high stress matching cost. The corresponding formula can be expressed as: C_stress = 1-match(stress_prob(e_k), beat_weight(j)), where the match() function represents the matching function. When the stressed syllable coincides with the strong beat, the value of the matching function is higher, that is, the matching degree is higher, and the corresponding stress matching cost is lower; conversely, when the value of the matching function is lower, the corresponding stress matching cost is higher. stress_prob(e_k) represents the stress probability of the syllable, and beat_weight(j) represents the beat intensity of slot j.

[0038] Furthermore, the aforementioned overflow and insufficiency penalties are designed to prevent a syllable from spanning too many slots or a slot from accommodating too many syllables, thus preventing "syllable stacking" or "sentence-breaking." Typically, the number of syllables allocated to each slot or the number of slots a syllable spans can be determined based on the original placeholder map and the translation filling information. If a syllable spans two or more slots, or if a slot is allocated more than one syllable, the corresponding overflow and insufficiency penalties are higher.

[0039] Furthermore, the aforementioned semantic integrity cost is to avoid breaking alignment in semantically close phrases and to maintain a natural semantic flow. Specifically, it can be determined whether a sentence is over-splitting and causing semantic breaks based on the contextual semantic relationship of the syllable sequence. If so, the semantic integrity cost is high.

[0040] Furthermore, for each of the above cost indices, a weighted sum can be performed according to the pre-configured weight parameters to obtain the final comprehensive cost index. During the dynamic optimization process, DP[i][j] can be constructed, representing the minimum cost of filling the first i syllables into the first j slots, with an initial DP[0][0]=0. Then, the filling situations of all syllables are enumerated, and the comprehensive cost index is calculated once for each filling situation until the translation filling information with the minimum comprehensive cost index is obtained. At this point, the above step S110 is completed.

[0041] Furthermore, in this embodiment of the invention, before segmenting the translation corresponding to the lyrics into syllables, the translation corresponding to the lyrics can be optimized according to pre-configured optimization conditions; wherein, the optimization conditions include at least one of the following conditions: slot utilization optimization condition, rhythm feature optimization condition, end rhyme optimization condition, and imagery optimization condition; and then the optimized translation is segmented into syllables.

[0042] Specifically, the optimization process for the translation can also be implemented based on multiple collaborative agents in the lyrics processing system. For example, if the optimization conditions include slot utilization optimization, the alignment agent is invoked to calculate the slot utilization corresponding to the translation based on the translation filling information; it is then determined whether the slot utilization meets the pre-configured threshold range; if so, the translation is confirmed to meet the slot utilization optimization conditions; if not, adjustment information is generated and fed back to the sentence segmentation agent to prompt the sentence segmentation agent to re-segment the lyrics text.

[0043] Furthermore, in the lyrics processing system of this embodiment, a morpheme agent can also be run. For example, if the optimization conditions include prosodic feature optimization conditions, the morpheme agent is invoked to perform morpheme decomposition on each lyric phrase to obtain the first speech distribution feature corresponding to the lyric phrase. The first speech distribution feature is used to characterize the frequency and distribution features of initials and finals in the lyric phrase. The translation corresponding to the lyric phrase is phoneme-segmented to obtain the second speech distribution feature corresponding to the translation. The second speech distribution feature is used to represent the frequency and distribution features of vowels and consonants in the translation. A preset number of pinyin features are extracted from the first speech distribution feature according to the order of frequency from high to low, and a preset number of letter features are extracted from the second speech distribution feature. The similarity between the first and second speech distribution features is calculated based on the pinyin and letter features. If the similarity reaches a similarity threshold, the translation is determined to meet the preset prosodic features; otherwise, the translation is determined not to meet the prosodic features, and optimization prompts are generated to prompt prosodic adjustment of the translation.

[0044] Furthermore, if the optimization conditions include end rhyme optimization conditions, the morpheme agent is invoked to determine whether the ending sound of the translation and the lyrics satisfy the end rhyme constraint mechanism; if yes, it is determined that the translation meets the end rhyme optimization conditions; if no, optimization prompt information is generated, wherein the optimization prompt information is used to prompt the translation to adjust the end rhyme.

[0045] Furthermore, the lyrics processing system in this embodiment of the invention also operates an imagery agent; if the above optimization conditions include imagery optimization conditions, the imagery agent can be invoked to extract imagery information contained in the lyrics, wherein the imagery information includes the types and quantities of intentional words contained in the lyrics; and, the imagery reproduction information of the translation corresponding to the lyrics is extracted; the imagery reproduction rate of the translation is calculated based on the imagery reproduction information and the imagery information; if the imagery reproduction rate reaches a preset reproduction rate threshold, the translation is determined to meet the imagery optimization conditions; if the imagery reproduction rate is lower than the preset reproduction rate threshold, the imagery of the translation is adjusted.

[0046] In actual use, one or more of the above optimization conditions can be selected for execution, or all of them can be executed. The specific settings can be made according to the actual usage situation, and the embodiments of the present invention do not impose any restrictions on this.

[0047] Furthermore, to realize the aforementioned multi-agent collaborative lyrics processing system, a lyrics processing server can typically be set up, through which multiple functional agents can run to constitute the lyrics processing system of this embodiment. Furthermore, each agent has independent input / output format definitions, calling tools, natural language prompts, and algorithm execution logic. Processing results are transmitted between agents using a unified data structure, achieving comprehensive optimization in terms of singability, rhythmic consistency, and preservation of artistic conception.

[0048] For ease of understanding, Figure 2 A schematic diagram of a lyrics processing system is shown, such as Figure 2 As shown, it includes acquisition agent 20, sentence segmentation agent 21, statistical agent 22, alignment agent 23, morpheme agent 24, image agent 25, and composite agent 26.

[0049] Specifically, the intelligent agent 20 can perform the process described in step S102 above, that is, obtain the lyrics text to be processed. In addition, it can also obtain the corresponding translation of the lyrics. For example, it can automatically retrieve and output the corresponding lyrics text and sheet music data based on the input song name. The input format type is generally: song name: Chinese / English song title string with exact or fuzzy matching; the output format type adopts structured output (JSON Schema), including, for example: song_meta: {title, artist, album, release_year, version_tag} lyrics: An array containing {line_id, content, timestamp(optional)} jianpu: Simplified musical notation structure and time signature slotting results, including: key_signature: key signature (e.g., "C"); time_signature: time signature (e.g., "4 / 4"); tempo_bpm: tempo (e.g., 92); measures: array, organized by measure, etc.

[0050] In addition, the acquiring agent can also call music database APIs and musical notation data parsing tools, which can be configured according to actual usage.

[0051] Further, the sentence segmentation agent 21 is mainly used for sentence segmentation, that is, dividing the lyrics into clauses. For example, according to the semantics of the lyrics and the singing breathing pattern, the whole lyrics are divided into sentences suitable for singing. Each sentence is equivalent to the lyrics clause or the lyrics measure in the embodiments of the present invention. Its input format type is: string (the whole lyrics text); the output format type is: json (the sequence of sentence texts). Specifically, when implemented, the sentence segmentation agent 21 can call natural language clause segmentation models, such as the BERT model, the HanLP model, etc., and can also call the speech rhythm detection module. Moreover, the sentence segmentation agent will analyze the semantic structure of the lyrics (such as the completeness of the sentence meaning) and the simple music score rhythm (such as the note length that each sentence should contain) at the same time, and then use the language model to calculate the semantic coherence degree between each word, and combine the beat position to calculate the rhythm pause score. The two are weighted and summed to obtain the total score. When the score reaches a local peak, it is regarded as a suitable sentence segmentation point. For example, when there is a natural semantic break point in the lyrics (such as "The wind has stopped | The rain has also fallen asleep"), and there is also a measure boundary in the beat, the sentence segmentation agent will segment the sentence here.

[0052] Further, the above-mentioned statistical agent 22, alignment agent 23, morpheme agent 24, and image agent 25 are mainly used for the above optimization process, that is, before syllable segmentation of the translation, the translation corresponding to the lyrics can be optimized according to the pre-configured optimization conditions.

[0053] For example, the statistical agent 22 can count the slot utilization rate, and can also count the above-mentioned first speech distribution feature and the second speech distribution feature to assist optimization. Taking the statistical agent 22 counting the first speech distribution feature and the second speech distribution feature as an example, the function of the statistical agent 22 at this time is mainly to count the frequency and distribution characteristics of the initials and finals appearing in each sentence of the lyrics, identify the highest-frequency preset quantity, such as the first three initials and finals, and determine the rhyming pattern and beat type. Its input generally comes from the lyrics sentence segmentation sequence output by the sentence segmentation agent, and the output is the first three initials, the first three finals, the rhyming type, and the beat information of each sentence, etc.

[0054] In actual use, the statistical agent 22 can call tools such as the pinyin decomposition module and the beat analysis module, and then convert each Chinese character into pinyin and split it into initials and finals. For example, "wind" is divided into the initial "f" and the final "eng". Further, count the occurrence frequency of each initial and final to obtain the phonetic rhyme distribution table of each sentence of the lyrics. In addition, the statistical agent can also use the rhyming detection algorithm to identify the end rhyme type of each sentence. Finally, judge the beat type according to the simple music score rhythm information, and mark it according to the stress rule corresponding to the beat type to help the subsequent agent determine the stress position, etc.

[0055] Furthermore, the aforementioned alignment agent 23 is used to achieve rhythmic alignment between lyrics and translation based on beat slots, ensuring that the translation is sung at the rhythmic level, and to run the aforementioned pre-configured dynamic optimization strategy.

[0056] Specifically, the input of the alignment agent 23 typically includes Chinese lyrics text and beat information, translation, original placeholder image and syllable sequence, and the output is the alignment of the syllables of the translation with the slots, that is, obtaining the above-mentioned translation filling information in the embodiment of the present invention. In order to realize the alignment function, the alignment agent can call beat detection tools, syllable segmentation tools, beat slot modeling modules, and machine translation models based on neural networks, etc., depending on the actual use case. The embodiment of the present invention does not impose any restrictions on this.

[0057] Furthermore, the aforementioned morpheme agent 24 can perform morpheme decomposition on the lyrics during the prosodic feature optimization process, while the imagery agent 25 is used to optimize the imagery of the translation. In actual use, the morpheme agent 24 and the imagery agent 25 can be performed before syllable segmentation of the translation corresponding to the lyrics, or after optimization processing, depending on the actual usage. This embodiment of the invention does not impose any restrictions on this.

[0058] Specifically, the morpheme agent 24 is mainly used to optimize the pronunciation structure of the translated text, making its initial and final distributions closer to the rhythm and rhyme of the original Chinese lyrics. Furthermore, the morpheme agent 24 can utilize tools such as a phonetic word segmenter, a morpheme analyzer, a pronunciation dictionary, and a rhyme detection module. For example, during morpheme decomposition, the phonetic word segmenter is used to decompose each Chinese character into an initial and a final. The frequency of these components is then counted and sorted by the statistical agent to obtain the aforementioned first phonetic distribution feature. Similarly, the phoneme analyzer is used to segment the translated words into consonants and vowels. Again, the frequency of these components is counted and sorted by the statistical agent to obtain the aforementioned second phonetic distribution feature. In addition, the morpheme agent 24 can also calculate the similarity between the first and second phonetic distribution features; for example, defining a similarity function: similarity_score = (overlap_top3_initials + overlap_top3_finals) / 6; Wherein, overlap_top3_initials = the number of intersections between the top 3 initials / consonants in the lyrics and the translation; overlap_top3_finals = the number of intersections between the top 3 finals / vowels in the lyrics and the translation. The specific similarity function can also be set to other forms according to actual needs, and this embodiment of the invention does not impose any restrictions on this.

[0059] Furthermore, the rhyme constraint mechanism corresponding to the aforementioned rhyme optimization conditions can also invoke the morpheme agent 24. For example, it can extract the final vowels of Chinese lyrics (such as "an") and check whether the vowels of the final words in the English translation rhyme (such as "again" → "an"). If they do not rhyme, it can call the candidate replacement word library (synonyms / near-synonyms) for replacement, which can realize the detection of whether the rhymes of the lyrics and the translation are consistent.

[0060] Furthermore, the aforementioned imagery agent 25 is used to ensure that the translated text maintains consistency with the Chinese lyrics in terms of emotional and artistic conception. Specifically, the imagery agent 25 can invoke natural language understanding models, imagery dictionaries, cultural reference databases, etc., to perform semantic analysis on the Chinese lyrics and extract a set of keywords; then, it matches the imagery dictionary to filter out imagery words with cultural symbolic meaning (such as "moon," "wind," "mountain," etc.); it counts the number and types of imagery words in the original Chinese lyrics; it checks the reproduction of corresponding images in the translated text; it calculates the proportion of reproduced images to original images to obtain the artistic conception retention rate; if the artistic conception retention rate is lower than a set threshold (such as 0.85), it automatically replaces non-keywords in the translated text with words with similar cultural imagery.

[0061] Furthermore, the aforementioned composite agent 26 is used to embed the final optimized translated lyrics into the original melody to generate synthesized music, thus implementing step S112. Typically, the composite agent 26 can call music synthesis engines (such as Synthesizer V, Vocaloid), vocoder modules, etc., to generate the corresponding melody's synthesized audio based on the beat slots and the lyrics. Specifically, it can embed the translated lyrics into the melody slot by slot according to the beat slot order; call synthesis tools to generate audio signals; automatically match pitch and rhythm to ensure the lyrics pronunciation aligns with the melody beat; and finally output a complete music file.

[0062] In summary, in this embodiment of the invention, by processing lyrics through the aforementioned multi-agent collaborative approach, a method for optimizing the singability of lyrics translation based on morpheme similarity and rhyme constraints can be achieved. Specifically, this method performs frequency statistics and alignment at the phonetic level on the original lyrics and the translation, respectively, on a sentence-by-sentence basis, thereby maintaining the consistency of prosodic features in cross-language conversion. Specifically, in the Chinese lyrics processing stage, each line of lyrics is first decomposed into morphemes, and the frequencies of initials and finals are statistically analyzed. The results are then sorted and recorded to obtain the first phonetic distribution features of that line of lyrics. Subsequently, in the corresponding translation, such as the English translation, words are segmented into phonemes, and the frequencies of consonants and vowels are statistically analyzed, generating sorted results in the same way. By comparing the statistical results of the two languages, the frequency distribution of the top three consonants and vowels in the English lyrics is required to be as consistent as possible with the frequency distribution of initials and finals in the Chinese lyrics. Through this cross-language morpheme frequency alignment, the translation is closer to the original lyrics in terms of vocal rhythm and auditory feel, thus significantly improving singability. In addition, considering that Chinese lyrics generally emphasize rhyme in artistic expression, this embodiment of the invention can also introduce a forced rhyme constraint mechanism in the generation and optimization process of English lyrics. Corresponding to the above-mentioned rhyme optimization conditions, at the end of each line of lyrics, words or syllables that can form a rhyme must be selected to ensure that the translation echoes the Chinese lyrics in terms of rhythmic structure. This rhyme constraint not only enhances the musicality of the translation but also makes it more in line with the aesthetic expectations of the audience when sung.

[0063] Furthermore, by processing lyrics through the aforementioned multi-agent collaborative approach, a method for preserving the artistic conception of lyrics translation based on the reproduction of key imagery can be achieved. Specifically, by introducing the concept of poetic imagery, the core images in the original lyrics can be extracted and reproduced during the translation process, thereby ensuring that the translation still conveys the artistic conception and aesthetic characteristics of the original work during cross-language conversion. In this method, semantic analysis and imagery recognition are first performed on the Chinese lyrics. Through the combination of natural language processing technology and an imagery dictionary, recurring or culturally symbolic imagery information in the lyrics is extracted, such as "mountain," "sea," "moon," "wind," "loneliness," and "journey." This imagery information not only carries the emotional atmosphere of the lyrics but also constitutes the core of the poetic expression. Subsequently, during the translation process, it is required to reproduce this key imagery information. Specifically, while maintaining semantic coherence, the translation must prioritize retaining or replacing words with equivalent cultural imagery. For example, in Chinese, the word "moon" often symbolizes longing and loneliness. In English translation, one can choose to directly use "moon" or, if necessary, introduce similar imagery such as "night sky" to maintain the emotional tone of the original text.

[0064] In the reproduction process, embodiments of the present invention can also introduce a consistency check mechanism for imagery. That is, at both the single-sentence and overall levels, the frequency and distribution of imagery are statistically analyzed to ensure that the translation maintains an overall resonance with the imagery of the original lyrics. At the same time, in conjunction with the constraints of rhyme and rhythm, necessary adjustments are made to the imagery vocabulary so that it can both convey the poetic meaning of the original text and naturally blend into the melody.

[0065] The advantages of this imagery reproduction are twofold: firstly, it allows the aesthetics of poetic imagery to be introduced into lyric translation, making the translation not only a semantic conversion but also a reproduction of the artistic conception; secondly, by extracting and reproducing key images, it avoids the cultural loss and emotional dilution problems commonly encountered in literal translation. The final translated lyrics maintain a high degree of consistency with the original text in terms of semantics, artistic conception, and rhythm, thus achieving artistry and singability in cross-language performance.

[0066] Furthermore, embodiments of the present invention also provide an agent-based lyrics processing device, applied to a lyrics processing system, wherein the lyrics processing system operates a cooperative sentence segmentation agent, a statistical agent, and an alignment agent; such as Figure 3 The diagram shows a structural schematic of an agent-based lyrics processing device, which includes: Module 30 is used to acquire the lyrics text to be processed; The sentence segmentation intelligent agent module 32 is used to call the sentence segmentation intelligent agent to perform sentence segmentation processing on the lyrics text to obtain lyrics sentences, and to perform sentence segmentation annotation on each of the lyrics sentences to obtain sentence segmentation results, and input the sentence segmentation results into the statistical intelligent agent; The statistical agent module 34 is used to identify the rhyme pattern of each lyric phrase through the statistical agent, and to mark the stress position of the lyric phrase based on the rhyme pattern and the phrase segmentation result. Alignment agent module 36 is used to invoke the alignment agent to fill the lyrics phrases word by word according to the pre-set slots, thereby obtaining the original placeholder map corresponding to the lyrics phrases; and to segment the translation corresponding to the lyrics phrases into syllables to obtain a syllable sequence containing multiple syllables, and to fill each syllable in the syllable sequence into the slots one by one, thereby obtaining the translation filling information corresponding to the translation; and to optimize the filling of each syllable in the translation filling information based on the stress position and the original placeholder map by running a pre-configured dynamic optimization strategy through the alignment agent. The synthesis module 38 is used to embed the optimized translation fill information into the melody corresponding to the lyrics text to be processed, and generate the synthesized music corresponding to the lyrics text to be processed.

[0067] The lyrics processing device based on intelligent agents provided in this embodiment of the invention has the same technical features as the lyrics processing method based on intelligent agents provided in the above embodiments, so it can also solve the same technical problems and achieve the same technical effects.

[0068] Furthermore, this embodiment of the invention also provides a lyrics processing system, which operates a collaborative sentence segmentation agent, a statistical agent, and an alignment agent to execute the aforementioned agent-based lyrics processing method. The lyrics processing system can refer to... Figure 2 The content shown will not be repeated here.

[0069] Furthermore, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above method.

[0070] This invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the above-described method.

[0071] Furthermore, embodiments of the present invention also provide a schematic diagram of the structure of an electronic device, such as... Figure 4 The diagram shows the structure of the electronic device, which includes a processor 41 and a memory 40. The memory 40 stores computer-executable instructions that can be executed by the processor 41, and the processor 41 executes the computer-executable instructions to implement the above-described method.

[0072] exist Figure 4 In the illustrated embodiment, the electronic device further includes a bus 42 and a communication interface 43, wherein the processor 41, the communication interface 43, and the memory 40 are connected via the bus 42.

[0073] The memory 40 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this system network element and at least one other network element is achieved through at least one communication interface 43 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 42 may be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 42 can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0074] Processor 41 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of processor 41 or by software instructions. Processor 41 can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in the memory, and the processor 41 reads the information in the memory and uses its hardware to complete the aforementioned method.

[0075] The computer program product of the intelligent agent-based lyrics processing method, apparatus and system provided in the embodiments of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods described in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.

[0076] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0077] Furthermore, in the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0078] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0080] Finally, it should be noted that the above embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A lyrics processing method based on intelligent agents, characterized in that, It is applied to a lyrics processing system, which operates a sentence segmentation agent, a statistical agent, and an alignment agent that work together in a coordinated manner; The method includes: Obtain the lyrics text to be processed; The lyric text is segmented by a sentence-segmentation agent to obtain lyric phrases. Each lyric phrase is then segmented and labeled to obtain a segmentation result. The segmentation result is then input into the statistical agent. The statistical agent identifies the rhyme pattern of each lyric phrase, and, based on the rhyme pattern, marks the stress position of the lyric phrase in conjunction with the phrase segmentation results; The alignment agent is invoked to fill the lyrics phrases word by word according to the pre-set slots, thereby obtaining the original placeholder image corresponding to the lyrics phrases; and the translation corresponding to the lyrics phrases is segmented into syllables to obtain a syllable sequence containing multiple syllables, and each syllable in the syllable sequence is filled into the slot one by one to obtain the translation filling information corresponding to the translation. The alignment agent runs a pre-configured dynamic optimization strategy to optimize the filling of each syllable in the translation filling information based on the accent position and the original placeholder map. The optimized translation fill information is embedded into the melody corresponding to the lyrics text to be processed, thereby generating synthesized music corresponding to the lyrics text to be processed.

2. The method according to claim 1, characterized in that, The step of optimizing the filling of each syllable in the translation filling information based on the stress position and the original placeholder map by running a pre-configured dynamic optimization strategy through the alignment agent includes: The alignment agent calculates the cost index corresponding to each syllable in the translation filling information under the currently filled slot based on the original placeholder map and the stress position; wherein, the cost index is the cost index required for the comprehensive cost index of the dynamic optimization strategy. Each of the aforementioned cost indicators is weighted and calculated to obtain a comprehensive cost indicator; With the goal of minimizing the overall cost index, the slots filled by the syllables are gradually optimized until the overall cost index reaches its minimum value, thus obtaining the optimized translation filling information.

3. The method according to claim 2, characterized in that, The cost metrics include: time value matching cost, which characterizes the difference between the duration of the syllable and the duration of the slot; stress matching cost, which characterizes whether the stress of the syllable matches the strong beat of the slot; overflow and insufficiency penalty cost, which characterizes the penalty cost when the syllable crosses too many slots or exceeds the remaining slots; and semantic integrity cost, which characterizes whether the semantic segmentation of the syllable is fragmented. The comprehensive cost index is the result of a weighted summation of each of the cost indices.

4. The method according to claim 1, characterized in that, The method further includes: The translation of the lyrics is optimized according to the pre-configured optimization conditions; wherein the optimization conditions include at least one of the following conditions: slot utilization optimization condition, rhythm feature optimization condition, end rhyme optimization condition, and imagery optimization condition.

5. The method according to claim 4, characterized in that, The steps of optimizing the translation of the lyrics according to pre-configured optimization conditions include: If the optimization conditions include slot utilization optimization conditions, then the alignment agent is invoked to calculate the slot utilization corresponding to the translation based on the translation filling information; Determine whether the slot utilization rate meets the pre-configured threshold range; If so, the translation is determined to meet the slot utilization optimization conditions; If not, generate adjustment information and feed it back to the sentence segmentation agent to prompt the agent to re-segment the lyrics text.

6. The method according to claim 4, characterized in that, The lyrics processing system also runs a morpheme intelligent agent; The step of optimizing the translation of the lyrics according to pre-configured optimization conditions further includes: If the optimization conditions include prosodic feature optimization conditions, then the morpheme agent is invoked to perform morpheme decomposition on each lyric phrase to obtain the first speech distribution feature corresponding to the lyric phrase; wherein, the first speech distribution feature is used to characterize the frequency and distribution features of the initials and finals in the lyric phrase; Furthermore, the translation corresponding to the lyrics phrases is segmented into phonemes to obtain the second speech distribution features corresponding to the translation; wherein, the second speech distribution features are used to represent the frequency and distribution features of vowels and consonants in the translation; According to the order of occurrence frequency from high to low, a preset number of pinyin features are extracted from the first speech distribution features, and a preset number of letter features are extracted from the second speech distribution features. Based on the pinyin features and the letter features, calculate the similarity between the first speech distribution features and the second speech distribution features; If the similarity reaches a preset similarity threshold, the translation is determined to meet the preset prosodic features; otherwise, the translation is determined not to meet the prosodic features, and an optimization prompt is generated to prompt the translation to be adjusted in prosodic terms.

7. The method according to claim 6, characterized in that, The step of optimizing the translation of the lyrics according to pre-configured optimization conditions further includes: If the optimization conditions include end rhyme optimization conditions, then the morpheme agent is invoked to determine whether the ending sound of the translation and the lyrics satisfy the end rhyme constraint mechanism. If so, the translation is determined to satisfy the rhyme optimization condition; If not, generate optimization prompts, which are used to suggest adjustments to the translation.

8. The method according to claim 4, characterized in that, The lyrics processing system also runs an image-based intelligent agent; The step of optimizing the translation of the lyrics according to pre-configured optimization conditions further includes: If the optimization conditions include imagery optimization conditions, then the imagery agent is invoked to extract imagery information contained in the lyrics, wherein the imagery information includes the types and quantities of suggestive words contained in the lyrics; and: Extract the imagery reproduction information of the translated lyrics; The image reproduction rate of the translation is calculated based on the image reproduction information and the image information. If the imagery reproduction rate reaches a preset reproduction rate threshold, then the translation is determined to meet the imagery optimization condition. If the imagery reproduction rate is lower than a preset reproduction rate threshold, the imagery in the translation will be adjusted.

9. A lyrics processing device based on an intelligent agent, characterized in that, It is applied to a lyrics processing system, which operates a sentence segmentation agent, a statistical agent, and an alignment agent that work together in a coordinated manner; The device includes: The acquisition module is used to acquire the lyrics text to be processed; The sentence segmentation intelligent agent module is used to call the sentence segmentation intelligent agent to perform sentence segmentation processing on the lyrics text to obtain lyrics sentences, and to perform sentence segmentation annotation on each lyrics sentence to obtain sentence segmentation results, and input the sentence segmentation results into the statistical intelligent agent; A statistical agent module is used to identify the rhyme pattern of each lyric phrase through the statistical agent, and to mark the stress position of the lyric phrase based on the rhyme pattern and the phrase segmentation result; An alignment agent module is used to invoke the alignment agent to fill the lyrics phrases word by word according to pre-set slots, thereby obtaining the original placeholder image corresponding to the lyrics phrases; and to segment the translation corresponding to the lyrics phrases into syllables, obtaining a syllable sequence containing multiple syllables, and filling each syllable in the syllable sequence into the slots one by one, thereby obtaining the translation filling information corresponding to the translation; and to optimize the filling of each syllable in the translation filling information based on the stress position and the original placeholder image by running a pre-configured dynamic optimization strategy through the alignment agent. The synthesis module is used to embed the optimized translation fill information into the melody corresponding to the lyrics text to be processed, and generate synthesized music corresponding to the lyrics text to be processed.

10. A lyrics processing system, characterized in that, The lyrics processing system operates a collaborative sentence segmentation agent, a statistical agent, and an alignment agent to execute the agent-based lyrics processing method according to any one of claims 1 to 8.

Images