Intelligent education platform interaction method and system based on child voice behavior analysis

Abstract

The invention relates to the technical field of intelligent interaction, and discloses an intelligent education platform interaction method and system based on child voice behavior analysis, and the method comprises the steps: obtaining user interaction content which comprises a first text signal and a voice signal; performing semantic recognition according to the voice signal to obtain a second text signal; performing word segmentation processing according to the first text signal and the second text signal to obtain a text unit; constructing a text content vector according to the text unit, inputting the text content vector into a pre-trained cognitive competence prediction model, and outputting a cognitive competence score; searching an answer content vector in a pre-stored answer knowledge base according to the cognitive competence score; and performing similarity analysis according to the answer content vectors and the text content vectors, and outputting the text content corresponding to the answer content vector with the highest similarity as the optimal answer content. According to the method, the interaction accuracy can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent interaction technology, and in particular to an intelligent education platform interaction method and system based on children's speech behavior analysis. Background Art

[0002] At present, as the resource advantages of families and society continue to tilt towards children's education and development, the role and tasks of early education have been increasingly recognized and valued by parents and society. How to discover and develop children's language ability as early as possible is a major concern for parents and early childhood educators. Generally speaking, early childhood educators have rich knowledge in pedagogy, psychology, etc., and have more professional experience in guiding and discovering children's language ability. Parents are often not professional enough, and they can only turn to early childhood educators for problems encountered by their children in the development of language ability. However, this method is limited by time and region, and there is also a lack of personalized customization. Therefore, intelligent voice interaction devices are a preferred way to resolve this contradiction.

[0003] In one existing technology, by integrating advanced speech recognition technology, the voice input of children is first captured and converted into a processable text form; the system then uses natural language processing technology to semantically understand and analyze these texts, identify the specific needs or emotional states of children, and call the preset knowledge base to generate the most appropriate response information, which is then fed back to the children in the form of speech. When first used, traditional paper-and-pencil tests were used to assess the cognitive development of children in order to give appropriate answers.

[0004] However, the children's cognitive ability assessment method based on voice interaction is relatively independent of the children's cognitive ability test method. It fails to provide real-time scoring feedback for children's voice input during the cognitive ability assessment process, and lacks the correlation and penetration of children's cognitive ability assessment and voice interaction. In general, the interaction accuracy is not high. Summary of the invention

[0005] The present invention provides an intelligent education platform interaction method and system based on children's voice behavior analysis, so as to improve the interaction accuracy of the intelligent education platform.

[0006] In the first aspect, in order to solve the above technical problems, the present invention provides an intelligent education platform interaction method based on children's voice behavior analysis, comprising: Acquire user interaction content, where the user interaction content includes: a first text signal and a voice signal; Performing semantic recognition according to the voice signal to obtain a second text signal; Performing word segmentation processing according to the first text signal and the second text signal to obtain text units; Constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; Retrieving an answer content vector from a pre-stored answer knowledge base according to the cognitive ability score; A similarity analysis is performed between the answer content vector and the text content vector, and the text content corresponding to the answer content vector with the highest similarity is output as the optimal answer content.

[0007] In an optional implementation, performing semantic recognition according to the voice signal to obtain a second text signal includes: Sampling and quantizing the speech signal to obtain a speech digital signal; Extracting features from the speech digital signal to obtain speech signal features; The speech signal feature is input into a pre-trained speech recognition model, and a second text signal is output.

[0008] In an optional implementation manner, performing word segmentation processing according to the first text signal and the second text signal to obtain text units includes: Preprocessing the first text signal and the second text signal to obtain a standard text signal; Matching the standard text signal with a preset dictionary database to obtain vocabulary units; The vocabulary units are post-processed to obtain text units.

[0009] In an optional implementation, the process of constructing the text content vector includes: The text content vector is calculated using the following formula: ; in, is the first i dimensions, For the i In the text unit j The frequency of the word, For the i In the text unit j The inverse frequency of the word, m is the total number of text content vectors, i is the text unit number, j The number of the word in the text unit.

[0010] In an optional implementation, constructing a text content vector according to the text unit, inputting the text content vector into a pre-trained cognitive ability estimation model, and outputting a cognitive ability score includes: The cognitive ability score is calculated using the following formula: ; in, is the cognitive ability score, m is the total number of text content vectors, i is the text unit number, j is the number of the word in the text unit, For the i In the text unit j The conditional probability of a word, For the i Text units, For the j words, For the j The word weight of each word; The cognitive ability prediction model is obtained by training by taking the historical text vector as input and the historical cognitive ability score as output.

[0011] In an optional implementation, the retrieving the answer content vector from a pre-stored answer knowledge base according to the cognitive ability score includes: When the cognitive ability score is greater than the content ability score corresponding to the content vector in the answer knowledge base, using the content vector as the answer content vector; The content ability score is obtained by inputting the content vector into the cognitive ability estimation model in advance, and the content ability score is stored in the answer knowledge base.

[0012] In an optional implementation, performing similarity analysis on the answer content vector and the text content vector, and outputting the text content corresponding to the answer content vector with the highest similarity as the optimal answer content vector, includes: Calculating the cosine similarity between the answer content vector and the text content vector; Sorting the answer content vectors from high to low according to the cosine similarity; The top-ranked answer content vector is used as the optimal answer content vector; The optimal answer content vector is converted into text content, which is output to the user as the optimal answer content.

[0013] In a second aspect, the present invention provides an intelligent education platform interaction system based on children's speech behavior analysis, comprising: A data acquisition module, used to acquire user interaction content, wherein the user interaction content includes: a first text signal and a voice signal; A semantic recognition module, used for performing semantic recognition according to the voice signal to obtain a second text signal; A word segmentation processing module, used for performing word segmentation processing according to the first text signal and the second text signal to obtain text units; an ability calculation module, for constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; An answer retrieval module, used to retrieve an answer content vector from a pre-stored answer knowledge base according to the cognitive ability score; The result output module is used to perform similarity analysis based on the answer content vector and the text content vector, and output the text content corresponding to the answer content vector with the highest similarity as the optimal answer content.

[0014] In a third aspect, the present invention also provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements any one of the above-mentioned intelligent education platform interaction methods based on children's voice behavior analysis.

[0015] In a fourth aspect, the present invention also provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned intelligent education platform interaction methods based on children's voice behavior analysis.

[0016] Compared with the prior art, the present invention has the following beneficial effects: the present invention discloses an intelligent education platform interaction method and system based on children's voice behavior analysis, the method includes obtaining user interaction content, the user interaction content includes: a first text signal and a voice signal; semantic recognition is performed according to the voice signal to obtain a second text signal; word segmentation is performed according to the first text signal and the second text signal to obtain a text unit; a text content vector is constructed according to the text unit, and the text content vector is input into a pre-trained cognitive ability estimation model to output a cognitive ability score; according to the cognitive ability score, an answer content vector is retrieved from a pre-stored answer knowledge base; similarity analysis is performed according to the answer content vector and the text content vector, and the text content corresponding to the answer content vector with the highest similarity is output as the optimal answer content. This method can make up for the shortcomings of text input, such as spelling errors or unclear expressions, by semantic recognition of voice signals, thereby improving the accuracy of understanding user intentions. Especially for children, voice input is more natural than text input, and this step helps to improve user experience, especially for children who are younger and have limited writing skills. Converting text content into vector form allows unstructured text information to be understood by machine learning algorithms. Using vector matching technology, relevant answers can be quickly found in large amounts of data, greatly improving response speed.

[0017] Specifically, when the cognitive ability score is greater than the content ability score corresponding to the content vector, the content vector is used as the answer content vector. This process ensures that the answer provided does not exceed the child's current cognitive level, avoids comprehension barriers caused by overly difficult questions, and also prevents a decline in learning interest caused by overly simple content. By inputting the content vector into the cognitive ability prediction model in advance to obtain the content ability score and storing it in the answer knowledge base, the user's cognitive ability and content difficulty can be quickly compared, improving the system's response speed and efficiency. Calculate the cosine similarity between the text content vector entered by the user and the content vector of each candidate answer. Cosine similarity measures the similarity between two vectors by calculating the cosine value of the angle between them. It is not affected by the length of the vector and can more accurately reflect the relevance of the content.

[0018] In summary, the present invention can improve the interaction accuracy of the intelligent education platform by analyzing children's voice behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of an intelligent education platform interaction method based on children's voice behavior analysis provided by the first embodiment of the present invention; Figure 2It is a schematic diagram of the structure of an intelligent education platform interactive system based on children's voice behavior analysis provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] At present, as the resource advantages of families and society continue to tilt towards children's education and development, the role and tasks of early education have been increasingly recognized and valued by parents and society. How to discover and develop children's language ability as early as possible is a major concern for parents and early childhood educators. Generally speaking, early childhood educators have rich knowledge in pedagogy, psychology, etc., and have more professional experience in guiding and discovering children's language ability. Parents are often not professional enough, and they can only turn to early childhood educators for problems encountered by their children in the development of language ability. However, this method is limited by time and region, and there is also a lack of personalized customization. Therefore, intelligent voice interaction devices are a preferred way to resolve this contradiction.

[0022] In one existing technology, by integrating advanced speech recognition technology, the voice input of children is first captured and converted into a processable text form; the system then uses natural language processing technology to semantically understand and analyze these texts, identify the specific needs or emotional states of children, and call the preset knowledge base to generate the most appropriate response information, which is then fed back to the children in the form of speech. When first used, traditional paper-and-pencil tests were used to assess the cognitive development of children in order to give appropriate answers.

[0023] However, the children's cognitive ability assessment method based on voice interaction is relatively independent of the children's cognitive ability test method. It fails to provide real-time scoring feedback for children's voice input during the cognitive ability assessment process, and lacks the correlation and penetration of children's cognitive ability assessment and voice interaction. In general, the interaction accuracy is not high.

[0024] To solve the above problems, refer to Figure 1 The first embodiment of the present invention provides an intelligent education platform interaction method based on children's voice behavior analysis, comprising the following steps: S11, acquiring user interaction content, where the user interaction content includes: a first text signal and a voice signal; S12, performing semantic recognition according to the voice signal to obtain a second text signal; S13, performing word segmentation processing according to the first text signal and the second text signal to obtain text units; S14, constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; S15, searching for an answer content vector in a pre-stored answer knowledge base according to the cognitive ability score; S16, performing similarity analysis between the answer content vector and the text content vector, and outputting the text content corresponding to the answer content vector with the highest similarity as the optimal answer content.

[0025] In step S11, user interaction content is acquired, where the user interaction content includes: a first text signal and a voice signal.

[0026] In one embodiment, the present invention provides a text input box, allowing children to input text content through a keyboard or a touch screen. When the child enters text in the text input box, the system captures the text content in real time as the first text content. At the same time, a voice input button is provided, and the child can start the voice input function by clicking the button. When the child clicks the voice input button, the system starts the microphone to record the child's voice. After the recording is completed, the system generates a voice file in MP3 format for subsequent voice recognition.

[0027] In step S12, semantic recognition is performed according to the voice signal to obtain a second text signal.

[0028] In one implementation, the speech signal is sampled and quantized to obtain a speech digital signal; feature extraction is performed on the speech digital signal to obtain speech signal features; the speech signal features are input into a pre-trained speech recognition model to output a second text signal.

[0029] It is worth noting that the present invention directly uses the pydub library for sampling and quantization to process audio files, wherein a sampling rate of 16 kHz and a quantization bit number of 16 bits are selected.

[0030] It is worth noting that the feature extraction of the speech digital signal to obtain the speech signal feature includes: enhancing the high frequency part through pre-emphasis processing to make the spectrum of the speech digital signal flatter; frame processing to divide the signal into short-time frames; using Hamming window for windowing processing; performing fast Fourier transform on each frame signal; filtering the spectrum through Mel filter group; taking the logarithm of the filtered energy; performing discrete cosine transform on the logarithmic energy; extracting the first several coefficients as Mel frequency cepstrum coefficients as speech signal features.

[0031] In one embodiment, the present invention uses the open source speech recognition model Kaldi. First, the pre-trained model is loaded and used, and then the historical children's speech is used as input data, and the recognized historical children's speech text is used as output data to train the speech recognition model. In order to integrate Kaldi into the children's intelligent education platform, the present invention creates a Python interface to call Kaldi's decoding script and return the recognition result. The Kaldi model uses Mel frequency cepstral coefficient features. The training process of the model includes: first training a monophonic sub-model as the basis for subsequent training; then training a triphonic sub-model using triangular smoothing technology; using linear discriminant analysis (LDA) and maximum likelihood linear transform (MLLT) to further improve the model performance; and using time-delay neural network (TDNN) and chain model (Chain) for final training.

[0032] In step S13, word segmentation is performed according to the first text signal and the second text signal to obtain text units.

[0033] In one implementation, the first text signal and the second text signal are preprocessed to obtain a standard text signal; the standard text signal is matched with a preset dictionary database to obtain a vocabulary unit; and the vocabulary unit is post-processed to obtain a text unit.

[0034] It is worth noting that the purpose of the preprocessing is to remove noise in the text, unify the text format, and convert the text into structured data. The input text information can be obtained through the child's oral narration. For example, the child inputs "What color is the big tiger?" The text units obtained after word segmentation of the text information include "big", "tiger", "what color", "of", and "is". The processing process includes: removing useless symbols in the text, such as punctuation marks, special characters, etc., through regular expressions. Since the text is captured from a web page, it contains HTML tags. The BeautifulSoup library can be used to remove these tags. Then convert all text to lowercase. Stop words refer to words that appear frequently in the text but do not contribute much to the meaning of the text, including "of", "le", etc. Use a predefined stop word list to remove these words.

[0035] It is worth noting that dictionary database matching is to match the preprocessed text with a preset dictionary to identify vocabulary units. A dictionary is a database containing vocabulary and related information. This information includes part of speech, interpretation, synonyms, antonyms, etc. This method pre-stores a dictionary database. For Chinese text, this method can use jieba for word segmentation. The text is divided into separate words, and the text after word segmentation is compared with the vocabulary in the dictionary to find matching vocabulary units. The present invention adopts accurate matching to ensure the accuracy of subsequent cognitive ability assessment.

[0036] It is worth noting that the post-processing includes word order adjustment and semantic disambiguation. Word order adjustment adjusts the order of vocabulary units according to grammatical rules or context information. Semantic disambiguation solves the ambiguity of vocabulary units and ensures that the semantics of each vocabulary unit is correct. Disambiguation is performed based on context information. The processed vocabulary units are combined into the final text unit.

[0037] In step S14, a text content vector is constructed according to the text unit, and the text content vector is input into a pre-trained cognitive ability estimation model to output a cognitive ability score.

[0038] In one implementation, the text content vector is calculated using the following formula: ; in, is the first i dimensions, For the i In the text unit j The frequency of the word, For the i In the text unit j The inverse frequency of the word, m is the total number of text content vectors, i is the text unit number, j The number of the word in the text unit.

[0039] The cognitive ability score is calculated using the following formula: ; in, is the cognitive ability score, m is the total number of text content vectors, i is the text unit number, j is the number of the word in the text unit, For the i In the text unit j The conditional probability of a word, For the i Text units, For the j words, For the j The word weight of each word; The cognitive ability prediction model is obtained by training by taking the historical text vector as input and the historical cognitive ability score as output.

[0040] It is worth noting that the text content vector formula means: for each text unit, the weighted average word frequency of all its words is calculated, where the weight is the inverse word frequency. The purpose of this is to highlight those words that appear frequently in the current text unit but are uncommon in the entire document collection, so as to better capture the characteristics of the text. The calculation formula for the cognitive ability score means that for each text unit, the sum of the product of the conditional probability of all its words and the word weight is calculated, and then the results of all text units are added to obtain the final cognitive ability score. Among them, word frequency refers to the ratio of the number of times a word appears in a document to the sum of the number of times all words appear in the document. Inverse document frequency is used to measure the prevalence of a word in a document collection. It is calculated using the following formula: ; in, Represents the inverse document frequency of a term. Represents the total number of documents in the document collection. Indicates that it contains words The total number of documents, for j Document No. for i For example: for document 1: "this is the first document". "this" appears once, so the term frequency is 1 / 5. For document 2: "thisis the second second document". "second" appears twice, the total number of words is 6, so the term frequency is 1 / 3. The inverse term frequency is .

[0041] It is worth noting that conditional probability refers to the probability of a word appearing given a context, which is calculated by dividing the number of times the word appears in the context by the total number of times the word appears in the context. Word weight is used to measure the importance or influence of a word in a specific task. This method uses the product of the word frequency and the inverse word frequency for calculation.

[0042] It is worth noting that the construction of the cognitive ability estimation model includes the following steps: Select a neural network model that can learn the mapping relationship between the input text vector and the cognitive ability score. Collect a large amount of historical text data and its corresponding historical cognitive ability scores. Divide the data into a training set and a validation set. The data comes from various channels such as educational assessment and psychological testing. Use the above formula to convert the text data into a vector representation as the input data of the model, and then use the corresponding historical cognitive ability score as the output data to train the model. Use the mean square error as the evaluation indicator.

[0043] In step S15, the answer content vector is retrieved from a pre-stored answer knowledge base according to the cognitive ability score.

[0044] In one embodiment, when the cognitive ability score is greater than the content ability score corresponding to the content vector in the answer knowledge base, the content vector is used as the answer content vector; wherein the content ability score is obtained by inputting the content vector into the cognitive ability prediction model in advance, and the content ability score is stored in the answer knowledge base.

[0045] It is worth mentioning that the answer knowledge base stores answers with different scores. For example, "When the earth revolves around the sun, sometimes the sun shines on more places and sometimes less, thus forming the four seasons of spring, summer, autumn and winter." has a capability score of 20 points, and "The earth's axis is tilted at an angle of 23.5 degrees. This tilt angle causes the sun's direct point to move between the equator and the tropic of cancer during the year. This movement causes different latitudes to receive different amounts of solar radiation, thus forming the change of seasons." has a capability score of 50 points, which will be replied to different users.

[0046] In step S16, a similarity analysis is performed based on the answer content vector and the text content vector, and the text content corresponding to the answer content vector with the highest similarity is output as the optimal answer content.

[0047] In one implementation, the cosine similarity between the answer content vector and the text content vector is calculated; the answer content vectors are sorted from high to low according to the cosine similarity; the answer content vector with the highest sorting value is used as the optimal answer content vector; the optimal answer content vector is converted into text content and output to the user as the optimal answer content.

[0048] It is worth noting that the cosine similarity is calculated by the following formula: ; in, represents the cosine similarity, is the answer content vector, is the text content vector, is the modulus of the answer content vector, is the modulus of the text content vector.

[0049] The answer content vectors are sorted by a sorting algorithm. The answer content vector at the top of the sorting is used as the optimal answer content vector, which means that the answer content vector with the highest cosine similarity is selected as the best match. This is because the higher the cosine similarity, the closer the directions of the two vectors are, that is, the semantics or themes between the answer content vector and the text content vector are closer, so such an answer can better meet the needs of the user. The optimal answer content vector is converted into text content and output to the user as the optimal answer content. This step is to reversely convert the data in vector form back to text form. This can be done by looking up a pre-established vocabulary, where each vector dimension corresponds to a word or phrase in the vocabulary. By mapping each non-zero element in the optimal answer content vector back to the corresponding word, the original text content can be reconstructed, and finally the text content is presented to the user as an answer.

[0050] In order to facilitate the understanding of the present invention, some preferred embodiments of the present invention are further described below.

[0051] It is worth noting that for the question: "Why does the earth have four seasons?", this method can adjust the way and depth of the answer according to the different ages of children and the corresponding cognitive ability scores. For users whose cognitive ability scores are lower than the first threshold, for example, the first threshold is set to 30 points. For the question "Why does the earth have four seasons?" Answer: "When the earth revolves around the sun, sometimes the sun shines more and sometimes less, thus forming the four seasons of spring, summer, autumn and winter. In summer, the sun shines for a long time, and the weather is hot; in winter, the sun shines for a short time, and the weather is cold." In this example, very basic concepts are used to explain the changes of the four seasons, avoiding the use of overly professional terms to ensure that users can understand. It is assumed here that the user's cognitive ability score is low, so a more intuitive and simple explanation is provided. For users above the first threshold, the answer is: "While the earth revolves around the sun, it also maintains an inclination angle (about 23.5 degrees). Due to this inclination angle, when the earth moves on its orbit, the sun's direct point will move back and forth between the equator and the tropic of cancer, causing changes in the amount of solar radiation received by different regions, thus forming the four seasons of spring, summer, autumn and winter. In summer, the sun's direct point is close to the northern hemisphere, and the northern hemisphere receives more solar radiation and the temperature rises; in winter, the opposite is true, the sun's direct point is far away from the northern hemisphere, the northern hemisphere receives less solar radiation and the temperature drops." In this version of the answer, concepts such as "earth tilt" and "orbital revolution" are introduced, and a detailed explanation is given of how the movement of the sun's direct point affects the climate in different regions. This is more suitable for high-level users because they have higher cognitive ability scores and can understand more complex and abstract concepts.

[0052] It is worth noting that the present invention is based on a children's education intelligent platform and belongs to the data processing link. The data is analyzed by this method to achieve efficient natural language interaction. The children's education intelligent platform is a comprehensive system integrating multiple technologies, aiming to improve the efficiency and quality of children's education. The data input link of the platform accurately captures the user's scanning behavior through the user intention recognition technology of the scanning pen, reduces misoperation, and obtains accurate input data. The data processing link includes an intelligent recommendation system for children's education content based on knowledge graphs, which analyzes children's learning data and preferences and recommends personalized education content for children; and an adaptive adjustment system for the platform interaction interface based on image recognition, which identifies the environment and equipment when children interact with the platform, automatically adjusts the interaction interface, and improves the user experience. In addition, there is an intelligent education platform interaction system based on children's voice behavior analysis, which understands children's language habits and expressions and provides a natural and friendly interactive experience. The data output link analyzes children's learning performance and progress through a children's mathematical thinking training effect evaluation system based on data mining, evaluates the effect of the education platform, and helps educators and parents adjust education strategies in a timely manner. Overall, the platform covers the entire process from data input to processing to output, providing children with a personalized, interactive and measurable educational environment.

[0053] In summary, the present invention discloses an intelligent education platform interaction method based on children's voice behavior analysis, which aims to improve the interaction accuracy of the intelligent education platform, and realizes intelligent analysis and response to children's voice and text input by comprehensively using speech recognition, natural language processing and machine learning technology. This method obtains the user's text and voice interaction content, performs semantic recognition on the voice signal, converts it into a text signal, and then performs word segmentation processing together with the user's text input to form a text unit. These text units are then constructed into text content vectors and input into a pre-trained cognitive ability prediction model, which can output a score that matches the child's cognitive ability. Based on this score, the system can retrieve the answer content vector that is suitable for the child's cognitive level in the pre-stored answer knowledge base, and through similarity analysis, find the answer content that is most similar to the user input, and finally output the optimal answer content.

[0054] In terms of semantic recognition, the present invention uses sampling and quantization technology to convert speech signals into digital signals, extracts speech signal features, and then uses a pre-trained speech recognition model to output text signals. This process involves preprocessing of speech signals, such as pre-emphasis, framing, windowing, fast Fourier transform, Mel filter bank filtering, logarithmic energy calculation and discrete cosine transform, etc., to extract Mel frequency cepstral coefficients as speech signal features. The word segmentation processing involves preprocessing of text signals, including removing useless symbols, converting to lowercase, removing stop words, etc., as well as matching and post-processing with a dictionary database to obtain text units. This step ensures the accuracy of vocabulary units, solves the problem of ambiguity, and ensures the semantic correctness of each vocabulary unit. The construction process of the text content vector includes calculating the weighted average word frequency of words in each text unit, where the weight is the inverse word frequency, to highlight those words that appear frequently in the current text unit but are uncommon in the entire document collection. The calculation of cognitive ability scores is based on the sum of the product of the conditional probabilities of all words in each text unit and the word weights, and then adding the results of all text units. The cognitive ability estimation model is trained by taking historical text vectors as input and historical cognitive ability scores as output. This model can learn the mapping relationship between text vectors and cognitive ability scores, thereby providing a personalized interactive experience for each child. In the answer retrieval and result output stage, this method retrieves the appropriate answer content vector from the answer knowledge base based on the child's cognitive ability score, and calculates the cosine similarity to find the answer content that is most similar to the user input. Cosine similarity is an indicator that measures the similarity between two vectors. It is achieved by calculating the cosine value of the angle between the two vectors. It is not affected by the length of the vector and can more accurately reflect the relevance of the content.

[0055] In general, the present invention provides an accurate intelligent education platform interaction method by combining children's voice behavior analysis and cognitive ability estimation, thereby improving interaction efficiency.

[0056] Reference Figure 2 The second embodiment of the present invention provides an intelligent education platform interaction system based on children's voice behavior analysis, including: A data acquisition module, used to acquire user interaction content, wherein the user interaction content includes: a first text signal and a voice signal; A semantic recognition module, used for performing semantic recognition according to the voice signal to obtain a second text signal; A word segmentation processing module, used for performing word segmentation processing according to the first text signal and the second text signal to obtain text units; an ability calculation module, for constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; An answer retrieval module, used to retrieve an answer content vector from a pre-stored answer knowledge base according to the cognitive ability score; The result output module is used to perform similarity analysis based on the answer content vector and the text content vector, and output the text content corresponding to the answer content vector with the highest similarity as the optimal answer content.

[0057] Preferably, the data acquisition module is used to: User interaction content is acquired, where the user interaction content includes: a first text signal and a voice signal.

[0058] Preferably, the semantic recognition module is used to: Performing semantic recognition according to the voice signal to obtain a second text signal includes: Sampling and quantizing the speech signal to obtain a speech digital signal; Extracting features from the speech digital signal to obtain speech signal features; The speech signal feature is input into a pre-trained speech recognition model, and a second text signal is output.

[0059] Preferably, the word segmentation processing module is used to perform word segmentation processing according to the first text signal and the second text signal to obtain text units, including: Preprocessing the first text signal and the second text signal to obtain a standard text signal; Matching the standard text signal with a preset dictionary database to obtain vocabulary units; The vocabulary units are post-processed to obtain text units.

[0060] Preferably, the ability calculation module is used to construct a text content vector according to the text unit, and input the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score, including: The text content vector is calculated using the following formula: ; in, is the first i dimensions, For the i In the text unit j The frequency of the word, For the i In the text unit jThe inverse frequency of the word, m is the total number of text content vectors, i is the text unit number, j is the number of the word in the text unit; The cognitive ability score is calculated using the following formula: ; in, is the cognitive ability score, m is the total number of text content vectors, i is the text unit number, j is the number of the word in the text unit, For the i In the text unit j The conditional probability of a word, For the i Text units, For the j words, For the j The word weight of each word; The cognitive ability prediction model is obtained by training by taking the historical text vector as input and the historical cognitive ability score as output.

[0061] Preferably, the answer retrieval module is used to retrieve the answer content vector from a pre-stored answer knowledge base according to the cognitive ability score, including: When the cognitive ability score is greater than the content ability score corresponding to the content vector in the answer knowledge base, using the content vector as the answer content vector; The content ability score is obtained by inputting the content vector into the cognitive ability estimation model in advance, and the content ability score is stored in the answer knowledge base.

[0062] Preferably, the result output module is used to perform similarity analysis based on the answer content vector and the text content vector, and output the text content corresponding to the answer content vector with the highest similarity as the optimal answer content, including: Calculating the cosine similarity between the answer content vector and the text content vector; Sorting the answer content vectors from high to low according to the cosine similarity; The top-ranked answer content vector is used as the optimal answer content vector; The optimal answer content vector is converted into text content, which is output to the user as the optimal answer content.

[0063] It should be noted that the intelligent education platform interaction system based on children's voice behavior analysis provided in an embodiment of the present invention is used to execute all the process steps of the intelligent education platform interaction method based on children's voice behavior analysis in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, so they will not be repeated here.

[0064] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, the steps in the above-mentioned embodiments of the intelligent education platform interaction method based on children's voice behavior analysis are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data acquisition module.

[0065] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0066] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0067] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0068] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0069] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0070] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0071] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent education platform interaction method based on children's voice behavior analysis, characterized in that: include: Acquire user interaction content, where the user interaction content includes: a first text signal and a voice signal; Performing semantic recognition according to the voice signal to obtain a second text signal; Performing word segmentation processing according to the first text signal and the second text signal to obtain text units; Constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; Retrieving an answer content vector from a pre-stored answer knowledge base according to the cognitive ability score; A similarity analysis is performed between the answer content vector and the text content vector, and the text content corresponding to the answer content vector with the highest similarity is output as the optimal answer content.

2. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The performing semantic recognition according to the voice signal to obtain a second text signal includes: Sampling and quantizing the speech signal to obtain a speech digital signal; Extracting features from the speech digital signal to obtain speech signal features; The speech signal feature is input into a pre-trained speech recognition model, and a second text signal is output.

3. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The performing word segmentation processing according to the first text signal and the second text signal to obtain text units includes: Preprocessing the first text signal and the second text signal to obtain a standard text signal; Matching the standard text signal with a preset dictionary database to obtain vocabulary units; The vocabulary units are post-processed to obtain text units.

4. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The process of constructing the text content vector includes: The text content vector is calculated using the following formula: ； in, is the i-th dimension of the text content vector, is the frequency of the jth word in the i-th text unit, is the inverse word frequency of the jth word in the ith text unit, m is the total number of text content vectors, i is the text unit number, and j is the word number in the text unit.

5. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The step of constructing a text content vector according to the text unit, inputting the text content vector into a pre-trained cognitive ability estimation model, and outputting a cognitive ability score includes: The cognitive ability score is calculated using the following formula: ； in, is the cognitive ability score, m is the total number of text content vectors, i is the text unit number, j is the word number in the text unit, is the conditional probability of the jth word in the i-th text unit, is the i-th text unit, is the jth word, is the word weight of the jth word; The cognitive ability prediction model is obtained by training by taking the historical text vector as input and the historical cognitive ability score as output.

6. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The step of retrieving the answer content vector from a pre-stored answer knowledge base according to the cognitive ability score includes: When the cognitive ability score is greater than the content ability score corresponding to the content vector in the answer knowledge base, using the content vector as the answer content vector; The content ability score is obtained by inputting the content vector into the cognitive ability estimation model in advance, and the content ability score is stored in the answer knowledge base.

7. The intelligent education platform interaction method based on children's voice behavior analysis according to claim 1 is characterized in that: The performing similarity analysis on the answer content vector and the text content vector, and outputting the text content corresponding to the answer content vector with the highest similarity as the optimal answer content vector, comprises: Calculating the cosine similarity between the answer content vector and the text content vector; Sorting the answer content vectors from high to low according to the cosine similarity; The top-ranked answer content vector is used as the optimal answer content vector; The optimal answer content vector is converted into text content, which is output to the user as the optimal answer content.

8. An intelligent education platform interactive system based on children's voice behavior analysis, characterized in that: include: A data acquisition module, used to acquire user interaction content, wherein the user interaction content includes: a first text signal and a voice signal; A semantic recognition module, used for performing semantic recognition according to the voice signal to obtain a second text signal; A word segmentation processing module, used for performing word segmentation processing according to the first text signal and the second text signal to obtain text units; an ability calculation module, for constructing a text content vector according to the text unit, and inputting the text content vector into a pre-trained cognitive ability estimation model to output a cognitive ability score; An answer retrieval module, used to retrieve an answer content vector from a pre-stored answer knowledge base according to the cognitive ability score; The result output module is used to perform similarity analysis based on the answer content vector and the text content vector, and output the text content corresponding to the answer content vector with the highest similarity as the optimal answer content.

9. An electronic device, characterized in that: It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements the intelligent education platform interaction method based on children's voice behavior analysis as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the intelligent education platform interaction method based on children's voice behavior analysis as described in any one of claims 1 to 7.

Images