Voice processing method, device and medium

Abstract

The embodiment of the present invention provides a speech processing method and device, and a device for speech processing, wherein the method specifically includes: determining the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users; according to the entry features, determining personalized entries from the user vocabulary; the personalized entries are used for speech input. The embodiment of the present invention can improve the distinction and accuracy of personalized entries, and can improve the accuracy of speech recognition.

Description

Technical Field

[0001] The present invention relates to the field of speech processing technology, and in particular to a speech processing method and device, and a machine-readable medium. Background Art

[0002] Speech recognition technology can convert speech into corresponding characters or codes, and is widely used in voice input, intelligent chat, voice translation and other fields.

[0003] Current speech recognition models usually use acoustic models and language models, giving priority to identifying common words and words with a higher probability of appearing in the corpus.

[0004] In the process of implementing the embodiments of the present invention, the inventors found that commonly used words and words with a higher probability of appearing in the corpus reflect the language usage habits of most users, while individual users usually have their own personalized language usage habits. If the language usage habits of most users are contrary to the personalized language usage habits, erroneous speech recognition results will be provided. Summary of the invention

[0005] In view of the above problems, embodiments of the present invention are proposed to provide a speech processing method, a speech processing device and a device for speech processing that overcome the above problems or at least partially solve the above problems. The embodiments of the present invention can improve the distinctiveness and accuracy of personalized entries, and can improve the accuracy of speech recognition.

[0006] In order to solve the above problems, the present invention discloses a speech processing method, comprising:

[0007] Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0008] According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input.

[0009] In order to solve the above problems, the present invention discloses a speech processing method, comprising:

[0010] Receiving voice data to be recognized;

[0011] Decoding the speech data to be recognized;

[0012] According to the personalized entry corresponding to the user, the score of the decoding path corresponding to the speech data to be recognized is adjusted; the personalized entry corresponding to the user is obtained according to the above method;

[0013] Determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path

[0014] On the other hand, an embodiment of the present invention discloses a speech processing device, including:

[0015] The entry feature determination module is used to determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0016] The personalized entry determination module is used to determine a personalized entry from the user vocabulary based on the entry features; the personalized entry is used for voice input.

[0017] On the other hand, an embodiment of the present invention discloses a speech processing device, including:

[0018] A receiving module, used for receiving voice data to be recognized;

[0019] A decoding module, used for decoding the speech data to be recognized;

[0020] An adjustment module, configured to adjust the score of the decoding path corresponding to the speech data to be recognized according to the personalized entry corresponding to the user; the personalized entry corresponding to the user is obtained according to the aforementioned method; and

[0021] The speech recognition result determination module is used to determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path.

[0022] In another aspect, an embodiment of the present invention discloses a device for speech processing, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations:

[0023] Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0024] According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input.

[0025] The embodiment of the present invention further discloses one or more machine-readable media, characterized in that instructions are stored thereon, which, when executed by one or more processors, enable the device to execute the aforementioned method.

[0026] The embodiments of the present invention include the following advantages:

[0027] The embodiment of the present invention comprehensively utilizes the entry features such as personal word frequency, public word frequency, and the number of users to mine personalized entries from the user vocabulary. The comprehensive utilization of the entry features such as personal word frequency, public word frequency, and the number of users can improve the distinguishability of personalized entries between different users and the representativeness of user entries for the current user, thereby improving the distinguishability and accuracy of personalized entries.

[0028] The embodiment of the present invention applies personalized entries to the speech recognition process, which can make the speech recognition result conform to the personalized habits of the user, thereby improving the accuracy of speech recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a schematic diagram of the process of a speech processing method according to an embodiment of the present invention;

[0030] Figure 2 is a flow chart of the steps of Embodiment 1 of a speech processing method of the present invention;

[0031] Figure 3 is a flow chart of the steps of Embodiment 2 of a speech processing method of the present invention;

[0032] Figure 4 is a flow chart of the steps of Embodiment 3 of a speech processing method of the present invention;

[0033] Figure 5 is a flowchart of the steps of Embodiment 4 of a speech processing method of the present invention;

[0034] Figure 6 is a flow chart of the steps of Embodiment 5 of a speech processing method of the present invention;

[0035] Figure 7 is a flowchart of the steps of Embodiment 6 of a speech processing method of the present invention;

[0036] Figure 8 is a structural block diagram of a speech processing device of the present invention;

[0037] Fig. 9 is a structural block diagram of a speech processing device of the present invention;

[0038] Fig.10 is a block diagram of a device 1300 for speech processing of the present invention; and

[0039] Fig.11 It is a structural schematic diagram of a server of the present invention. DETAILED DESCRIPTION

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] The embodiment of the present invention can be applied to speech recognition scenarios. The speech processing scenario is used to convert speech into text, and the speech processing scenario can include: speech input scenario, intelligent chat scenario, speech translation scenario, etc.

[0042] The acoustic model uses a deep neural network model to model the mapping relationship between acoustic pronunciation and basic acoustic units (usually phonemes); phonemes are the smallest speech units divided according to the natural properties of speech. The acoustic model can receive input speech features and output a phoneme sequence corresponding to the speech features.

[0043] Reference Figure 1 , which shows a schematic diagram of the process of a speech recognition method in an embodiment of the present invention. The models used in the speech recognition method may include: an acoustic model, a language model and a decoder.

[0044] The process of determining the acoustic model may include: extracting features from speech corpora in a speech database, and training the acoustic model based on the extracted features.

[0045] The process of determining the language model may include: training the language model based on the text corpus in the text database.

[0046] The decoder is used to find the best decoding path given a phoneme sequence, and then obtain the speech recognition result.

[0047] Figure 1 The speech recognition process shown may include: extracting features from the input speech to obtain speech features, and inputting the speech features into a decoder. The decoder first uses an acoustic model to determine the phoneme sequence corresponding to the speech features; then, speech decoding is performed on the phoneme sequence according to a language model to obtain a speech recognition result, and the text corresponding to the speech recognition result is output.

[0048] The acoustic model may include: a neural network model and a hidden Markov model, wherein the neural network model may provide an acoustic modeling unit to the hidden Markov model, and the granularity of the acoustic modeling unit may include: a word, a syllable, a phoneme, or a state, etc.; and the hidden Markov model may determine the phoneme sequence based on the acoustic modeling unit provided by the neural network model. A state mathematically represents the state of a Markov process.

[0049] The decoder in the speech recognition process searches for the best decoding path in the search space composed of knowledge sources such as acoustic models, dictionaries and language models based on the speech data to be recognized, and obtains the speech recognition result based on the word sequence corresponding to the best decoding path.

[0050] In the process of speech recognition, there are often problems with homophones or similar-sounding words. For example, the syllable "shuangzai" can correspond to words such as "shuangzai", "shuangzai", "shuangzai", and "shuangzai".

[0051] Current decoders usually use general acoustic models and language models, and give priority to identifying common words and words with a high probability of appearing in the corpus. Common words and words with a high probability of appearing in the corpus can be called popular words.

[0052] When users speak, they usually include their own personalized terms, such as names of people, places, institutions, or words in their professional fields, etc. Since these personalized terms are not popular words, the possibility of accurate recognition is low.

[0053] The embodiment of the present invention mines personalized entries from the user vocabulary based on the entry features of user entries in the user vocabulary, and applies the personalized entries to the speech recognition process to make the speech recognition results conform to the user's personalized habits, thereby improving the accuracy of speech recognition.

[0054] In conventional technology, the following two methods are usually used to determine personalized entries. One method provides a field for the user to select, and after the user selects a field, the entry in the field is used as the personalized entry. Another method provides an adding interface to receive the personalized entry added by the user through the adding interface. The above two methods usually cost the user a setting cost, and the obtained personalized entry usually has limitations.

[0055] In order to improve the accuracy of speech recognition results, an embodiment of the present invention provides a speech processing solution, which specifically includes: determining the entry features corresponding to the user entries in the user vocabulary; the above-mentioned entry features include: personal word frequency, public word frequency, and the number of users; based on the above-mentioned entry features, determining personalized entries from the above-mentioned user vocabulary.

[0056] Those skilled in the art mine personalized entries from the user vocabulary based on entry features such as personal word frequency, public word frequency, and the number of users using the entry. Among them, personal word frequency can be used to represent the number of times the user entry appears in the historical input data of the current user (or logged-in user); public word frequency can be used to represent the number of times the user entry appears in the historical input data of users across the entire network; and the number of users using the entry can be used to represent the number of users using the user entry.

[0057] The historical input data may represent the historical input content of the user. In practical applications, the input method program may record the user's upper screen content to obtain the user's historical input content. Optionally, in order to improve the timeliness of the personalized entry, the historical input data may be updated, or the historical input data within a preset time period may be selected. The length of the preset time period may be determined by those skilled in the art, for example, the preset time period may be half a year, three months, etc.

[0058] The embodiment of the present invention comprehensively utilizes the entry features such as personal word frequency, public word frequency, and the number of users to mine personalized entries from the user vocabulary. The comprehensive utilization of the entry features such as personal word frequency, public word frequency, and the number of users can improve the distinguishability of personalized entries between different users and the representativeness of user entries for the current user, thereby improving the distinguishability and accuracy of personalized entries.

[0059] The voice processing method provided by the embodiment of the present invention can be applied to an application environment of a client and a server. The client and the server are located in a wired or wireless network, and the client and the server perform data interaction through the wired or wireless network.

[0060] Optionally, the client may run on a terminal. For example, the client may be a program (APP, Application) running on the terminal, such as a speech transcription APP, a speech translation APP, or an intelligent interaction APP.

[0061] Taking the voice transcription APP as an example, the client can collect the voice data to be recognized and send the voice data to be recognized to the server. The server can use the solution of the embodiment of the present invention to process the voice data to be recognized and return the voice recognition result to the client.

[0062] Taking the voice translation APP as an example, the client can collect the voice data to be recognized and send the voice data to be recognized to the server. The server can use the solution of the embodiment of the present invention to process the voice data to be recognized, and perform machine translation on the obtained voice recognition results to obtain machine translation results, and return the machine translation results to the client.

[0063] Optionally, the terminal may include: a smart phone, a tablet computer, an e-book reader, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer, a car computer, a desktop computer, a set-top box, a smart TV, a wearable device, a smart speaker, etc. It is understood that the embodiments of the present invention are not limited to specific terminals.

[0064] The embodiments of the present invention can be applied to input method programs of various input methods such as keyboard symbols, handwriting, and voice. Taking the keyboard symbol input method as an example, the user can input text through a coded string, and the input string can refer to the coded string input by the user. In the field of input methods, for input method programs such as Chinese, Japanese, Korean, or other languages, the input string input by the user can usually be converted into a candidate of the corresponding language. The following is mainly explained by taking Chinese as an example, and other languages ​​such as Japanese and Korean can be referenced to each other. It can be understood that the above-mentioned Chinese input method can include but is not limited to full spelling, simplified spelling, strokes, Wubi, etc., and the embodiments of the present invention do not limit the specific input method program corresponding to a certain language.

[0065] Taking Chinese input as an example, the types of encoded character strings may include: Pinyin string, character string (such as Wubi string, etc.) Taking English input as an example, the types of encoded character strings may include: letter string, etc.

[0066] In actual applications, for the input method of keyboard symbols, the user can input the above input string through a physical keyboard or a virtual keyboard. For example, for a terminal with a touch screen, a virtual keyboard can be set in the input interface to input the input string by triggering the virtual keys included in the above virtual keyboard. Optionally, examples of the above virtual keyboard may include: a 9-key keyboard and a 26-key keyboard. In addition, it can be understood that in addition to the virtual keys corresponding to letters, the above input interface can also be provided with symbol keys, number keys, such as function keys for switching between Chinese and English, or toolbar keys, etc. It can be understood that the embodiment of the present invention does not limit the specific keys included in the input interface.

[0067] According to some embodiments, the input string may include, but is not limited to, a key symbol or a combination of multiple key symbols input by the user through keys. The key symbol may specifically include pinyin, strokes, kana, and the like.

[0068] In the embodiment of the present invention, the candidate may be used to represent one or more characters provided by the input method program to be selected by the user. The candidate may be characters in languages ​​such as Chinese characters, English characters, and Japanese characters, and the candidate may also be a combination of symbols in the form of emoticons, pictures, and the like. The emoticons include but are not limited to pictures composed of lines, symbols, and words. For example, examples of the emoticons may include: “:P”, “:-o”, “:-)”, and the like.

[0069] The screen-up refers to the operation of outputting the content in the preview window to the application window. Specifically for the input method program, the screen-up can refer to the operation of outputting the candidates in the candidate window to the application window. The screen-up content can represent the candidates for the screen-up.

[0070] Method Example 1

[0071] Reference Figure 2 , shows a flow chart of the steps of a speech processing method embodiment 1 of the present invention, which may specifically include the following steps:

[0072] Step 201: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0073] Step 202: Determine a personalized entry from the user vocabulary based on the entry features; the personalized entry is used for voice input.

[0074] Figure 2 At least one step included in the method shown may be executed by the client and / or the server. It can be understood that the embodiment of the present invention does not limit the specific execution subject of the steps included in the method.

[0075] The vocabulary of an input method program usually includes: a system vocabulary and a user vocabulary.

[0076] The system vocabulary can be obtained by applying statistical learning methods to the corpus and then extracting words with high usage frequency.

[0077] In order to improve the user experience, the client can record the user's screen content into the user vocabulary. The user vocabulary can include the first entry in the system vocabulary, and can also include: the second entry newly input by the user that is not in the system vocabulary. The entries in the user vocabulary can usually be user entries.

[0078] The user vocabulary may correspond to a single user, and the user corresponding to the user vocabulary may refer to the current user as a logged-in user.

[0079] The embodiment of the present invention comprehensively utilizes the entry features such as personal word frequency, public word frequency, and the number of users to mine personalized entries from the user vocabulary. The comprehensive utilization of the entry features such as personal word frequency, public word frequency, and the number of users can improve the distinguishability of personalized entries between different users and the representativeness of user entries for the current user, thereby improving the distinguishability and accuracy of personalized entries.

[0080] Personal word frequency can be used to represent the number of times a user term appears in the historical input data of the current user (or logged-in user); public word frequency can be used to represent the number of times a user term appears in the historical input data of users across the entire network; and the number of users can be used to represent the number of users who use the user term.

[0081] Entry features such as personal word frequency, public word frequency, and the number of users can characterize the distinctiveness of a user entry between different users and the representativeness of the user entry for the current user.

[0082] The embodiment of the present invention can provide the following technical solution for determining personalized entries from the above user vocabulary:

[0083] Technical solution 1,

[0084] In technical solution 1, the above-mentioned determining personalized entries from the above-mentioned user vocabulary specifically includes: filtering out user entries whose number of users exceeds a first threshold from the above-mentioned user vocabulary.

[0085] The larger the number of commonly used users is, the more users use the user term, which means that the user term is less distinctive among different users and the user term is less representative of the current user. Conversely, the smaller the number of commonly used users is, the fewer users use the user term, which means that the user term is more distinctive among different users and the user term is more representative of the current user.

[0086] In an optional embodiment of the present invention, the user entry of the embodiment of the present invention may be an entry generated by a user within a preset time period. The preset time period may be determined by those skilled in the art according to actual application requirements. For example, the length, start time, and end time of the preset time period may be determined by those skilled in the art according to actual application requirements. For example, the duration between the end time and the current time may not exceed the preset duration, such as the preset duration is one day. For example, the length of the preset time period may be 1 month, 2 months, 3 months, etc.

[0087] In another optional embodiment of the present invention, the personalized entry may be located within a preset time period.

[0088] In another optional embodiment of the present invention, the corresponding weight can be determined according to the generation time of the user term, and the weight is used in the process of determining the personalized term from the above user vocabulary. Generally, the higher the weight, the greater the probability that the user term is a personalized term.

[0089] Those skilled in the art can determine the first threshold value according to actual application requirements, for example, the first threshold value is 1, or 2. For example, when the first threshold value is 1, it means that the user entry has been used by more than one user, so the user entry can be filtered out to improve the distinguishability of the personalized entry.

[0090] Technical solution 2,

[0091] In technical solution 2, the above-mentioned determining personalized entries from the above-mentioned user vocabulary specifically includes: obtaining user entries whose personal word frequency exceeds a second threshold from the above-mentioned user vocabulary.

[0092] The personal word frequency can represent the representativeness of the user term to the current user. Generally, the higher the personal word frequency, the more commonly used the user term is to the current user, which means that the user term is more representative to the current user.

[0093] Those skilled in the art can determine the second threshold value according to actual application requirements. For example, the second threshold value is 30, 100, or other values.

[0094] Technical solution 3,

[0095] In technical solution 3, the above-mentioned determining personalized entries from the above-mentioned user vocabulary specifically includes: obtaining personalized entries from the user vocabulary based on the relative relationship between personal word frequency and public word frequency.

[0096] The relative relationship between personal word frequency and public word frequency can also illustrate the distinctiveness of a user term between different users and the representativeness of the user term for the current user.

[0097] Taking the relative relationship as a ratio as an example, if the ratio is small, it means that the personal word frequency is much smaller than the public word frequency, so it means that in addition to the current user, there are other users who have input a large number of user terms. Therefore, the smaller the difference between different users, the weaker the representativeness of the user terms for the current user. On the contrary, if the ratio is larger, it means that the personal word frequency is equivalent to the public word frequency, and the number of times the current user inputs the user term accounts for a large proportion of the number of times the user inputs the user term on the entire network. Therefore, it means that the difference between different users is greater, and the user term is more representative of the current user.

[0098] Technical solution 4,

[0099] In technical solution 4, the above-mentioned determination of personalized entries from the above-mentioned user vocabulary specifically includes: fusing the above-mentioned personal word frequency, the inverse of the above-mentioned public word frequency and the inverse of the above-mentioned number of users; and obtaining personalized entries from the above-mentioned user vocabulary based on the fusion characteristics.

[0100] Generally, the greater the personal word frequency, the smaller the inverse of the public word frequency, and the smaller the inverse of the number of users, it means that the current user often uses the user term, the number of times the current user inputs the user term accounts for a larger proportion of the number of times the user inputs the user term across the entire network, and the user term is used by fewer users; therefore, it means that the user term is more distinctive between different users and is more representative for the current user.

[0101] Optionally, a function with the personal word frequency, the inverse of the public word frequency and the inverse of the number of users as independent variables may be constructed, and the fusion feature may be determined based on the function. The function may perform a weighted average of the personal word frequency, the inverse of the public word frequency and the inverse of the number of users, or may multiply the personal word frequency, the inverse of the public word frequency and the inverse of the number of users, etc. It is understood that the function may correspond to parameters. It is understood that the embodiments of the present invention do not limit the specific parameters.

[0102] For example, the personal word frequency, the inverse of the public word frequency and the inverse of the number of users are multiplied, and a fusion feature is obtained based on the multiplication result. Then, personalized entries can be obtained from the user vocabulary in descending order of the fusion features.

[0103] Technical solution 5,

[0104] In technical solution 5, the above-mentioned term features may also include: hit features for the system vocabulary; the hit features may indicate whether the user term exists in the system vocabulary, if so, it indicates that the user term is a popular term, otherwise it indicates that the user term is a self-created term.

[0105] The above-mentioned determining personalized entries from the above-mentioned user vocabulary specifically includes: obtaining user entries whose hit features are misses from the above-mentioned user vocabulary, so that the personalized entries can be limited to the scope of self-created entries.

[0106] Technical Solution 6,

[0107] In technical solution 6, the above-mentioned term features may also include: entity features, which may indicate whether the user term corresponds to a named entity.

[0108] An entity is an object that exists in the real world and can be distinguished from other objects. Named entities refer to entities with specific meanings in text, such as names of people, places, organizations, proper nouns, etc.

[0109] When a user speaks, the personalized terms usually correspond to named entities such as names of people, places, and organizations. Therefore, the embodiment of the present invention can use entity features to mine personalized terms.

[0110] In practical applications, user terms can be matched with named entities in the entity library, or named entity recognition can be performed on user terms to determine whether the user terms correspond to named entities. Named entity recognition methods can include: rule-based and dictionary-based methods, statistical-based methods, or neural network-based methods.

[0111] Optionally, the determining of personalized entries from the user vocabulary specifically includes: acquiring user entries whose entity features are named entities from the user vocabulary.

[0112] The above technical solutions 1 to 5 have introduced in detail the process of determining personalized entries from the above-mentioned user vocabulary. It can be understood that technical personnel in this field can adopt any one or a combination of the above-mentioned technical solutions 1 to 5 according to actual application requirements, or can also adopt other technical solutions. The embodiment of the present invention does not limit the specific process of determining personalized entries from the above-mentioned user vocabulary.

[0113] For example, in other solutions of the present invention, user terms such as stop words and single-character words may also be filtered out from the user vocabulary.

[0114] In information retrieval, in order to save storage space and improve search efficiency, certain words are automatically filtered out before or after processing natural language data (or text). These words are called Stop Words. Examples of stop words may include: "I", "we", "he", "want", "better", "night", "so", "wrong", etc.

[0115] The embodiment of the present invention can save the personalized terms mined in step 202 to the user's personalized word library. It is understandable that the personalized terms can also be mined from the updated user terms according to the update event of the user word library, thereby realizing the update of the personalized word library.

[0116] In the embodiment of the present invention, optionally, user entries in the user vocabulary that are different from personalized entries may be deleted.

[0117] In summary, the speech processing method of the embodiment of the present invention comprehensively utilizes the entry features such as personal word frequency, public word frequency, and the number of users to mine personalized entries from the user vocabulary. The comprehensive utilization of the entry features such as personal word frequency, public word frequency, and the number of users can improve the distinguishability of personalized entries between different users and the representativeness of user entries for the current user, thereby improving the distinguishability and accuracy of personalized entries.

[0118] Method Example 2

[0119] Reference Figure 3 , shows a flow chart of the steps of Embodiment 2 of a speech processing method of the present invention, which may specifically include the following steps:

[0120] Step 301: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features specifically include: hit features, entity features, personal word frequency, public word frequency, and the number of users of the system vocabulary;

[0121] Step 302: Obtain the first user term with a hit feature of being a miss from the user word library;

[0122] Step 303: Delete stop words and single-character words from the first user entry to obtain a second user entry;

[0123] Step 304: Obtain a third user entry whose entity feature is a named entity from the second user entry;

[0124] Step 305: Filter out fourth user entries whose number of users exceeds a first threshold from the third user entries;

[0125] Step 306: Acquire a fifth user entry whose personal word frequency exceeds a second threshold from the fourth user entry;

[0126] Step 307: fuse the personal word frequency, the inverse of the public word frequency and the inverse of the number of users; obtain a personalized entry from the fifth user entry based on the fusion feature; and use the personalized entry for voice input.

[0127] The embodiment of the present invention sequentially utilizes entry features such as hit features, stop word features, entity features, number of users, personal word frequency, and fusion features to filter and screen user entries. The comprehensive utilization of multiple entry features can improve the distinction between personalized entries and the representativeness of user entries for the current user, thereby improving the distinction and accuracy of personalized entries.

[0128] It can be understood that in practical applications, the embodiments of the present invention do not limit the specific order of using term features. For example, entity features can be used first and then hit features.

[0129] The embodiment of the present invention can determine the score of the personalized entry based on the entry features, and sort the personalized entries in descending order of the scores. For example, the score of the personalized entry can be determined based on the aforementioned fusion features. It can be understood that the embodiment of the present invention does not limit the specific determination process of the score of the personalized entry.

[0130] Method Example 3

[0131] Reference Figure 4 , shows a flow chart of the steps of Embodiment 3 of a speech processing method of the present invention, which may specifically include the following steps:

[0132] Step 401: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0133] Step 402: Determine a personalized entry from the user vocabulary based on the entry features; the personalized entry is used for voice input;

[0134] Relative to Figure 2 The method of the first embodiment shown in the figure may further include:

[0135] Step 403: When the user is not logged in, output login prompt information, where the login prompt information is used to guide the user to start the personalized voice recognition service.

[0136] In the embodiment of the present invention, when the user status is not logged in, the login prompt information is output to guide the user to start the personalized voice recognition service, and the personalized voice recognition service can represent a service that uses personalized terms to perform voice recognition. Since it is impossible to locate a specific user when the user status is not logged in, the login prompt information is output to guide the user to log in and start the personalized voice recognition service.

[0137] The embodiment of the present invention can play the login prompt information, or can display the login prompt information on the interface. Optionally, the login prompt information can include: a login entrance, so that the user logs in through the login entrance and then starts the personalized voice recognition service.

[0138] Optionally, the login prompt information may correspond to a closing entrance. For example, a closing control may be provided around the login prompt information to close the login prompt information.

[0139] In an optional embodiment of the present invention, when the user is in a non-logged-in state, it can be determined whether a closing operation for the login prompt information has been received, and if not, the login prompt information is output.

[0140] If a close operation has been received for the login prompt information, the number of close operations is determined. If the number of close operations is greater than the fourth threshold, this indicates that the user's willingness to enable personalized voice service is weak, so no processing is performed, that is, the login prompt information is not output.

[0141] If the number of closing operations is less than (the fourth threshold + 1), it can be determined whether the time since the closing operation is less than the first duration. If not, a login prompt message is output, which can reduce the frequency of the login prompt message disturbing the user.

[0142] If the time from the close operation is less than the first time length, it is determined whether the time from the close operation is greater than the second time length. If not, no processing is performed, that is, the login prompt information is not output, which can reduce the frequency of the login prompt information disturbing the user. If the time from the close operation is greater than the second time length, it can be determined whether the user has received modification behavior for the voice recognition result. If so, it is considered that the user has a need to improve the accuracy of the voice recognition result, so the login prompt information is output.

[0143] Those skilled in the art can determine the fourth threshold, the first duration and the second duration according to actual application requirements. For example, the fourth threshold is 1, the first duration is 4 days, the second duration is 1 day, and so on.

[0144] Method Example 4

[0145] Reference Figure 5 , shows a flowchart of the steps of a fourth embodiment of a speech processing method of the present invention, which may specifically include the following steps:

[0146] Step 501: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0147] Step 502: according to the above-mentioned entry features, a personalized entry is determined from the above-mentioned user vocabulary; the above-mentioned personalized entry is used for voice input;

[0148] Relative to Figure 2 The method of the first embodiment shown in the figure may further include:

[0149] Step 503: When the user status is updated from the never logged in status to the logged in status, a service activation prompt message is output, where the service activation prompt message is used to prompt the user to activate the personalized voice recognition service.

[0150] In the embodiment of the present invention, when the user status is updated from the non-logged-in status to the logged-in status, service activation prompt information is output, and personalized voice recognition service activation feedback can be provided according to the user's login behavior, thereby improving the user experience.

[0151] According to an embodiment, the service activation prompt information may be the first service activation prompt information, and the first service activation prompt information may have an informing function, for example, the corresponding text may be "personalized voice recognition will be activated soon, and input words will be recognized more accurately" and the like.

[0152] According to another embodiment, the service activation prompt information may be the second service activation prompt information, and the second service activation prompt information may include: the number of personalized entries, or personalized entries whose personal word frequency exceeds the third threshold. For example, the text corresponding to the service activation prompt information may be "Personalized voice recognition is enabled, and learning of words such as "Lu Ye" and "Li Qian" is strengthened." Alternatively, the text corresponding to the service activation prompt information may be "Personalized voice recognition is enabled, and learning of X words is strengthened," etc.

[0153] Optionally, the first service activation prompt information may be displayed first, and then the second service activation prompt information may be displayed based on the user's confirmation information for the first service activation prompt information.

[0154] Method Example 5

[0155] Reference Figure 6 , shows a flowchart of the steps of Embodiment 5 of a speech processing method of the present invention, which may specifically include the following steps:

[0156] Step 601: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0157] Step 602: Determine a personalized entry from the user vocabulary based on the entry features.

[0158] Relative to Figure 2 The method of the first embodiment shown in the figure may further include:

[0159] Step 603: When the user is in the logged-in state, output update prompt information, where the update prompt information is used to prompt the update of the personalized entry.

[0160] In the embodiment of the present invention, when the user is in the logged-in state, update prompt information may be output to prompt the update of the personalized entry.

[0161] The embodiment of the present invention can output update prompt information according to a preset period. The length of the preset period can be determined by those skilled in the art according to actual application requirements, for example, the length of the preset period is 7 days.

[0162] Optionally, the update prompt information may include: the number of newly added personalized entries, such as the text of the update prompt information being "Y more exclusive words for you have been learned".

[0163] Method Example 6

[0164] Reference Figure 7 , shows a flowchart of the steps of Embodiment 6 of a speech processing method of the present invention, which may specifically include the following steps:

[0165] Step 701: receiving voice data to be recognized;

[0166] Step 702: Decode the speech data to be recognized;

[0167] Step 703: adjusting the score of the decoding path corresponding to the speech data to be recognized according to the personalized entry corresponding to the user;

[0168] The personalized entry corresponding to the above user can be obtained according to the above method;

[0169] Step 704: Determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path.

[0170] The embodiment of the present invention can use the personalized entry corresponding to the logged-in user to provide a personalized speech recognition service to the logged-in user.

[0171] In the process of speech recognition, homophones or similar-sounding words often appear. For example, the syllable "shuangzai" can correspond to words such as "shuangzai", "shuangzai", "shuangzai", and "shuangzai". Assuming that the personalized entry corresponding to a certain user includes "shuangzai", the score of the decoding path where "shuangzai" is located can be increased, thereby increasing the selection probability of the decoding path where the personalized entry is located, so that some keywords in the speech recognition results obtained by decoding are more matched with the personalized entry.

[0172] The embodiment of the present invention can use a decoder to decode the above-mentioned speech data to be recognized. Optionally, the decoder can use an acoustic model to determine the phoneme sequence corresponding to the speech features of the speech data to be recognized; then, the phoneme sequence is voice decoded according to the language model, and the voice decoding is used to find the best decoding path under the condition of a given phoneme sequence, so as to obtain a speech recognition result.

[0173] In the embodiment of the present invention, the language model and the acoustic model in the decoding process may adopt a general language model and an acoustic model to obtain a decoding path corresponding to the above-mentioned speech data to be recognized.

[0174] Furthermore, the score of the decoding path corresponding to the speech data to be recognized can be adjusted according to the personalized entry corresponding to the user. Specifically, the score of the decoding path passing through the personalized entry can be increased.

[0175] According to an embodiment, the increase score corresponding to the decoding path where the personalized term is located can be determined according to the score of the personalized term. For example, the higher the score of the personalized term, the greater the increase score.

[0176] By adjusting the decoding path score, the score of the decoding path where the personalized entry is located is improved, which can increase the selection probability of the decoding path where the personalized entry is located, make some keywords in the decoded speech recognition results more matched with the personalized entry, and improve the accuracy of speech recognition.

[0177] In summary, the speech recognition method of the embodiment of the present invention adjusts the score of the decoding path according to the personalized entry, which can improve the score of the decoding path where the personalized entry is located, and further improve the speech recognition accuracy corresponding to the personalized entry.

[0178] It should be noted that, for the method embodiments, for the sake of simplicity of description, they are all described as a series of motion action combinations, but those skilled in the art should know that the embodiments of the present invention are not limited by the order of the described motion actions, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the motion actions involved are not necessarily required by the embodiments of the present invention.

[0179] Device Embodiment

[0180] Reference Figure 8 , shows a structural block diagram of an embodiment of a speech processing device of the present invention, which may specifically include:

[0181] The entry feature determination module 801 is used to determine the entry features corresponding to the user entry in the user vocabulary; the entry features specifically include: personal word frequency, public word frequency, and the number of users;

[0182] The personalized entry determination module 802 is used to determine a personalized entry from the user vocabulary based on the entry features; the personalized entry is used for voice input.

[0183] Optionally, the personalized entry determination module may include:

[0184] A filtering module, used to filter out user entries whose number of users exceeds a first threshold from the user vocabulary; and / or

[0185] A first acquisition module is used to acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or

[0186] The second acquisition module is used to acquire personalized entries from the user vocabulary according to the relative relationship between the personal word frequency and the public word frequency.

[0187] Optionally, the personalized entry determination module may include:

[0188] A fusion module, used for fusing the personal word frequency, the inverse of the public word frequency and the inverse of the number of users;

[0189] The third acquisition module is used to acquire personalized entries from the user vocabulary based on the fusion features.

[0190] Optionally, the above-mentioned entry features may also include: hit features for the system vocabulary;

[0191] The personalized entry determination module may include:

[0192] The fourth acquisition module is used to acquire user terms with hit features as misses from the user vocabulary.

[0193] Optionally, the above-mentioned entry features may also include: entity features;

[0194] The personalized entry determination module may include:

[0195] The fifth acquisition module is used to acquire user terms whose entity features are named entities from the user vocabulary.

[0196] Optionally, the above device may further include:

[0197] The first prompt module is used to output login prompt information when the user is not logged in, and the login prompt information is used to guide the user to start the personalized voice recognition service.

[0198] Optionally, the above device may further include:

[0199] The second prompt module is used to output service activation prompt information when the user status is updated from the non-logged-in status to the logged-in status. The service activation prompt information is used to prompt the user to activate the personalized voice recognition service.

[0200] Optionally, the service activation prompt information may include: the number of personalized entries, or personalized entries whose personal word frequency exceeds a third threshold.

[0201] Optionally, the above device may further include:

[0202] The third prompt module is used to output update prompt information when the user is in the logged-in state, and the update prompt information is used to prompt the update of the personalized entry.

[0203] Reference Fig. 9 , shows a structural block diagram of an embodiment of a speech processing device of the present invention, which may specifically include:

[0204] Receiving module 901, used for receiving voice data to be recognized;

[0205] A decoding module 902, used for decoding the speech data to be recognized;

[0206] An adjustment module 903, configured to adjust the score of the decoding path corresponding to the speech data to be recognized according to the personalized entry corresponding to the user; the personalized entry corresponding to the user is obtained by the aforementioned method; and

[0207] The speech recognition result determination module 904 is used to determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path.

[0208] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0209] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0210] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0211] Fig.10 1 is a block diagram of a device 1300 for speech processing according to an exemplary embodiment. For example, the device 1300 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0212] Reference Fig.10, the device 1300 may include one or more of the following components: a processing component 1302 , a memory 1304 , a power component 1306 , a multimedia component 1308 , an audio component 1310 , an input / output (I / O) interface 1312 , a sensor component 1314 , and a communication component 1316 .

[0213] The processing component 1302 generally controls the overall operation of the device 1300, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 1302 may include one or more processors 1320 to execute instructions to perform all or part of the steps of the above-described method. In addition, the processing component 1302 may include one or more modules to facilitate interaction between the processing component 1302 and other components. For example, the processing component 1302 may include a multimedia module to facilitate interaction between the multimedia component 1308 and the processing component 1302.

[0214] The memory 1304 is configured to store various types of data to support operations on the device 1300. Examples of such data include instructions for any application or method operating on the device 1300, contact data, phone book data, messages, pictures, videos, etc. The memory 1304 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0215] The power supply component 1306 provides power to the various components of the device 1300. The power supply component 1306 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 1300.

[0216] The multimedia component 1308 includes a screen that provides an output interface between the device 1300 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 1308 includes a front camera and / or a rear camera. When the device 1300 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0217] The audio component 1310 is configured to output and / or input audio signals. For example, the audio component 1310 includes a microphone (MIC), and when the device 1300 is in an operation mode, such as a call mode, a recording mode, and a voice data processing mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 1304 or sent via the communication component 1316. In some embodiments, the audio component 1310 also includes a speaker for outputting audio signals.

[0218] I / O interface 1312 provides an interface between processing component 1302 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0219] The sensor assembly 1314 includes one or more sensors for providing various aspects of the status assessment of the device 1300. For example, the sensor assembly 1314 can detect the open / closed state of the device 1300, the relative positioning of components, such as the display and keypad of the device 1300, the sensor assembly 1314 can also detect the position change of the device 1300 or a component of the device 1300, the presence or absence of user contact with the device 1300, the orientation or acceleration / deceleration of the device 1300, and the temperature change of the device 1300. The sensor assembly 1314 can include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 1314 can also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 1314 can also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0220] The communication component 1316 is configured to facilitate wired or wireless communication between the device 1300 and other devices. The device 1300 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof. In an exemplary embodiment, the communication component 1316 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 1316 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency data processing (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0221] In an exemplary embodiment, the device 1300 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above-mentioned methods.

[0222] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 1304 including instructions, and the instructions can be executed by the processor 1320 of the device 1300 to perform the above method. For example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0223] Fig.11 1 is a schematic diagram of the structure of the server in an embodiment of the present invention. The server 1900 may have relatively large differences due to different configurations or performances, and may include one or more central processing units (CPU) 1922 (for example, one or more processors) and memory 1932, and one or more storage media 1930 (for example, one or more mass storage devices) storing application programs 1942 or data 1944. Among them, the memory 1932 and the storage medium 1930 can be short-term storage or permanent storage. The program stored in the storage medium 1930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the server. Furthermore, the central processing unit 1922 can be configured to communicate with the storage medium 1930 and execute a series of instruction operations in the storage medium 1930 on the server 1900.

[0224] The server 1900 may also include one or more power supplies 1926, one or more wired or wireless network interfaces 1950, one or more input and output interfaces 1958, one or more keyboards 1956, and / or one or more operating systems 1941, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, etc.

[0225] Those skilled in the art will readily appreciate other embodiments of the invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in this disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0226] A non-temporary computer-readable storage medium, when the instructions in the storage medium are executed by a processor of an intelligent terminal, enables the intelligent terminal to execute a voice processing method, the method comprising: determining the entry features corresponding to user entries in a user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users; based on the entry features, determining personalized entries from the user vocabulary; the personalized entries are used for voice input.

[0227] It should be understood that the present invention is not limited to the precise structure described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

[0228] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0229] The embodiment of the present invention discloses A1, a speech processing method, comprising:

[0230] Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0231] According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input.

[0232] A2. The method according to A1, wherein the determining of personalized entries from the user vocabulary comprises:

[0233] Filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or

[0234] Acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or

[0235] According to the relative relationship between the personal word frequency and the public word frequency, the personalized word entry is obtained from the user word library.

[0236] A3. According to the method described in A1, the step of determining the personalized entry from the user vocabulary comprises:

[0237] Merging the personal word frequency, the inverse of the public word frequency and the inverse of the number of users;

[0238] According to the fusion features, personalized entries are obtained from the user vocabulary.

[0239] A4. According to the method described in A1, the entry features further include: hit features for the system vocabulary;

[0240] Determining a personalized entry from the user vocabulary includes:

[0241] Obtain user terms whose hit features are missed from the user word library.

[0242] A5. According to the method described in A1, the term feature further comprises: entity feature;

[0243] Determining a personalized entry from the user vocabulary includes:

[0244] A user term whose entity feature is a named entity is obtained from the user vocabulary.

[0245] A6. The method according to any one of A1 to A5, further comprising:

[0246] When the user is in a non-logged-in state, a login prompt message is output, where the login prompt message is used to guide the user to start the personalized voice recognition service.

[0247] A7. The method according to any one of A1 to A5, further comprising:

[0248] When the user status is updated from the never logged in status to the logged in status, service activation prompt information is output, where the service activation prompt information is used to prompt the user to activate the personalized speech recognition service.

[0249] A8. According to the method described in A7, the service activation prompt information includes: the number of personalized entries, or personalized entries whose personal word frequency exceeds a third threshold.

[0250] A9. The method according to any one of A1 to A5, further comprising:

[0251] When the user is in the logged-in state, update prompt information is output, where the update prompt information is used to prompt the update of the personalized entry.

[0252] The embodiment of the present invention discloses B10, a speech processing method, including:

[0253] Receiving voice data to be recognized;

[0254] Decoding the speech data to be recognized;

[0255] According to the personalized entry corresponding to the user, the score of the decoding path corresponding to the speech data to be recognized is adjusted; the personalized entry corresponding to the user is obtained according to any one of the methods described in A1 to A9;

[0256] A speech recognition result corresponding to the speech data to be recognized is determined according to the adjusted score of the decoding path.

[0257] The embodiment of the present invention discloses C11, a speech processing device, comprising:

[0258] The entry feature determination module is used to determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0259] The personalized entry determination module is used to determine a personalized entry from the user vocabulary based on the entry features; the personalized entry is used for voice input.

[0260] C12. According to the device described in C11, the personalized entry determination module includes:

[0261] A filtering module, used to filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or

[0262] A first acquisition module is used to acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or

[0263] The second acquisition module is used to acquire personalized entries from the user vocabulary according to the relative relationship between the personal word frequency and the public word frequency.

[0264] C13. According to the device described in C11, the personalized entry determination module includes:

[0265] A fusion module, used for fusing the personal word frequency, the inverse of the public word frequency and the inverse of the number of users;

[0266] The third acquisition module is used to acquire personalized entries from the user vocabulary based on the fusion features.

[0267] C14. The device according to C11, wherein the entry feature further comprises: a hit feature for a system vocabulary;

[0268] The personalized entry determination module includes:

[0269] The fourth acquisition module is used to acquire user terms with hit features as misses from the user vocabulary.

[0270] C15. According to the device described in C11, the entry feature further includes: entity feature;

[0271] The personalized entry determination module includes:

[0272] The fifth acquisition module is used to acquire user terms whose entity features are named entities from the user vocabulary.

[0273] C16. The device according to any one of C11 to C15, further comprising:

[0274] The first prompt module is used to output login prompt information when the user is in a non-logged-in state, and the login prompt information is used to guide the user to start the personalized voice recognition service.

[0275] C17. The device according to any one of C11 to C15, further comprising:

[0276] The second prompt module is used to output service activation prompt information when the user status is updated from the non-logged-in status to the logged-in status, wherein the service activation prompt information is used to prompt the user to activate the personalized voice recognition service.

[0277] C18. According to the device described in C17, the service activation prompt information includes: the number of personalized entries, or personalized entries whose personal word frequency exceeds a third threshold.

[0278] C19. The device according to any one of C11 to C15, further comprising:

[0279] The third prompt module is used to output update prompt information when the user is in the logged-in state, and the update prompt information is used to prompt the update of the personalized entry.

[0280] The embodiment of the present invention discloses D20, a speech processing device, comprising:

[0281] A receiving module, used for receiving voice data to be recognized;

[0282] A decoding module, used for decoding the speech data to be recognized;

[0283] An adjustment module, configured to adjust the score of the decoding path corresponding to the speech data to be recognized according to a personalized entry corresponding to the user; the personalized entry corresponding to the user is obtained according to any one of the methods described in A1 to A9; and

[0284] The speech recognition result determination module is used to determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path.

[0285] The embodiment of the present invention discloses E21, a device for speech processing, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations:

[0286] Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users;

[0287] According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input.

[0288] E22. According to the apparatus of E21, the step of determining the personalized entry from the user vocabulary comprises:

[0289] Filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or

[0290] Acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or

[0291] According to the relative relationship between the personal word frequency and the public word frequency, the personalized word entry is obtained from the user word library.

[0292] E23. According to the apparatus of E21, the step of determining the personalized entry from the user vocabulary comprises:

[0293] Merging the personal word frequency, the inverse of the public word frequency and the inverse of the number of users;

[0294] According to the fusion features, personalized entries are obtained from the user vocabulary.

[0295] E24. The apparatus according to E21, wherein the entry feature further comprises: a hit feature for a system vocabulary;

[0296] Determining a personalized entry from the user vocabulary includes:

[0297] Obtain user terms whose hit features are missed from the user word library.

[0298] E25. The apparatus according to E21, wherein the term feature further comprises: entity feature;

[0299] Determining a personalized entry from the user vocabulary includes:

[0300] A user term whose entity feature is a named entity is obtained from the user vocabulary.

[0301] 26. The device according to any one of E21 to E25, further comprising:

[0302] When the user is in a non-logged-in state, a login prompt message is output, where the login prompt message is used to guide the user to start the personalized voice recognition service.

[0303] E27. The device according to any one of E21 to E25, further comprising:

[0304] When the user status is updated from the never logged in status to the logged in status, service activation prompt information is output, where the service activation prompt information is used to prompt the user to activate the personalized speech recognition service.

[0305] E28. According to the device described in E27, the service activation prompt information includes: the number of personalized entries, or personalized entries whose personal word frequency exceeds a third threshold.

[0306] E29. The device according to any one of E21 to E25, further comprising:

[0307] When the user is in the logged-in state, update prompt information is output, where the update prompt information is used to prompt the update of the personalized entry.

[0308] The embodiment of the present invention discloses F30, a device for speech processing, comprising a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors. The one or more programs include instructions for performing the following operations:

[0309] Receiving voice data to be recognized;

[0310] Decoding the speech data to be recognized;

[0311] According to the personalized entry corresponding to the user, the score of the decoding path corresponding to the speech data to be recognized is adjusted; the personalized entry corresponding to the user is obtained according to any one of the methods described in claims 1 to 9;

[0312] A speech recognition result corresponding to the speech data to be recognized is determined according to the adjusted score of the decoding path.

[0313] An embodiment of the present invention discloses G31, one or more machine-readable media having instructions stored thereon, which, when executed by one or more processors, enable the device to execute one or more of the methods described in A1 to A9.

[0314] The embodiment of the present invention discloses H32, one or more machine-readable media, on which instructions are stored, which, when executed by one or more processors, enable the device to execute the method described in B10.

[0315] The speech processing method, speech processing device and device for speech processing provided by the present invention are introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A speech processing method, characterized in that: include: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users; the public word frequency represents the number of times the user entry appears in the historical input data of all network users; According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input; When the user status is updated from the never logged in status to the logged in status, outputting service activation prompt information, wherein the service activation prompt information is used to prompt the user to activate the personalized speech recognition service; Determining a personalized entry from the user vocabulary includes: Filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or Acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or According to the relative relationship between the personal word frequency and the public word frequency, the personalized word entry is obtained from the user word library.

2. The method according to claim 1, characterized in that Determining a personalized entry from the user vocabulary includes: Merging the personal word frequency, the inverse of the public word frequency and the inverse of the number of users; According to the fusion features, personalized entries are obtained from the user vocabulary.

3. A speech processing method, characterized in that: include: Receiving voice data to be recognized; Decoding the speech data to be recognized; According to the personalized entry corresponding to the user, the score of the decoding path corresponding to the speech data to be recognized is adjusted; the personalized entry corresponding to the user is obtained according to the method described in any one of claims 1 to 2; A speech recognition result corresponding to the speech data to be recognized is determined according to the adjusted score of the decoding path.

4. A speech processing device, characterized in that: include: The entry feature determination module is used to determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users; the public word frequency represents the number of times the user entry appears in the historical input data of all network users; A personalized entry determination module, used to determine a personalized entry from the user vocabulary according to the entry features; the personalized entry is used for voice input; A second prompt module is used to output service activation prompt information when the user status is updated from the non-logged-in status to the logged-in status, wherein the service activation prompt information is used to prompt the user to activate the personalized voice recognition service; Determining a personalized entry from the user vocabulary includes: Filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or Acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or According to the relative relationship between the personal word frequency and the public word frequency, the personalized word entry is obtained from the user word library.

5. A speech processing device, characterized in that: include: A receiving module, used for receiving voice data to be recognized; A decoding module, used for decoding the speech data to be recognized; An adjustment module, configured to adjust the score of the decoding path corresponding to the speech data to be recognized according to the personalized entry corresponding to the user; the personalized entry corresponding to the user is obtained according to the method described in any one of claims 1 to 2; as well as The speech recognition result determination module is used to determine the speech recognition result corresponding to the speech data to be recognized according to the adjusted score of the decoding path.

6. A device for speech processing, characterized in that: The system comprises a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, wherein the one or more programs include instructions for performing the following operations: Determine the entry features corresponding to the user entry in the user vocabulary; the entry features include: personal word frequency, public word frequency, and the number of users; the public word frequency represents the number of times the user entry appears in the historical input data of all network users; According to the feature of the entry, a personalized entry is determined from the user vocabulary; the personalized entry is used for voice input; When the user status is updated from the never logged in status to the logged in status, outputting service activation prompt information, wherein the service activation prompt information is used to prompt the user to activate the personalized speech recognition service; Determining a personalized entry from the user vocabulary includes: Filter out user terms whose number of users exceeds a first threshold from the user vocabulary; and / or Acquire user entries whose personal word frequency exceeds a second threshold from the user word library; and / or According to the relative relationship between the personal word frequency and the public word frequency, the personalized word entry is obtained from the user word library.

7. A device for speech processing, characterized in that: The system comprises a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by one or more processors, wherein the one or more programs include instructions for performing the following operations: Receiving voice data to be recognized; Decoding the speech data to be recognized; According to the personalized entry corresponding to the user, the score of the decoding path corresponding to the speech data to be recognized is adjusted; the personalized entry corresponding to the user is obtained according to the method described in any one of claims 1 to 2; A speech recognition result corresponding to the speech data to be recognized is determined according to the adjusted score of the decoding path.

8. One or more machine-readable media, characterized in that Instructions are stored thereon, which, when executed by one or more processors, cause the apparatus to perform the method as claimed in any one of claims 1 to 2.

9. One or more machine-readable media, characterized in that Instructions are stored thereon, which, when executed by one or more processors, cause the apparatus to perform the method as claimed in claim 3.

Images