CN107632718B - Method, device and readable medium for recommending digital information in voice input - Google Patents

Method, device and readable medium for recommending digital information in voice input Download PDF

Info

Publication number
CN107632718B
CN107632718B CN201710655037.1A CN201710655037A CN107632718B CN 107632718 B CN107632718 B CN 107632718B CN 201710655037 A CN201710655037 A CN 201710655037A CN 107632718 B CN107632718 B CN 107632718B
Authority
CN
China
Prior art keywords
recommendation
pronunciation
digital
string
recommendation information
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201710655037.1A
Other languages
Chinese (zh)
Other versions
CN107632718A (en
Inventor
张鑫
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Baidu Netcom Science and Technology Co Ltd
Original Assignee
Beijing Baidu Netcom Science and Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Baidu Netcom Science and Technology Co Ltd filed Critical Beijing Baidu Netcom Science and Technology Co Ltd
Priority to CN201710655037.1A priority Critical patent/CN107632718B/en
Publication of CN107632718A publication Critical patent/CN107632718A/en
Application granted granted Critical
Publication of CN107632718B publication Critical patent/CN107632718B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Abstract

The invention provides a method and a device for recommending digital information in voice input and a readable medium. The method comprises the following steps: receiving a voice input request which carries the pronunciation of the numeric string and is input by a user; respectively acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model and a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by a preset digital model; if the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score; and displaying second recommendation information corresponding to the second recommendation score to the user. According to the technical scheme, if the numeric string is a number such as a telephone number, the recommendation information can be guaranteed to be pure numbers, the accuracy of the recommendation information of the numeric string can be improved, and a user can input the numeric string directly according to the recommendation information when inputting the numeric string, so that the input operation cost of the numeric string can be effectively improved, and the input efficiency of the numeric string is improved.

Description

Method, device and readable medium for recommending digital information in voice input
[ technical field ] A method for producing a semiconductor device
The invention relates to the technical field of computer application, in particular to a method and a device for recommending digital information in voice input and a readable medium.
[ background of the invention ]
With the development of science and technology, intelligent devices have become indispensable portable devices in people's lives. The user can send and receive mails, browse news, make comments and communicate with friends or chat through instant messaging installed on the intelligent equipment at any time and any place.
In the prior art, an information input device such as an Application (App) of an input method is installed on a smart device, and many apps of the existing input methods support voice input. The user may send a voice input request to the information input device by voice, and then the information input device may obtain recommended input information corresponding to the input request from the language model according to the voice input request and a pre-stored language model, and display the recommended input information to the user. Therefore, the user can directly input information according to the recommended input information, so that the operation steps of the user for inputting the information can be reduced, the operation cost for inputting the information is reduced, and the efficiency for inputting the information is improved. The prior art language model is generated from an existing corpus, and may include each word and the probability of occurrence of each word in the corpus, as well as the probability of occurrence of the word along with the corresponding context word. If a voice input request of a user carries a numeric string such as a telephone number, a language model recommends a corresponding recommendation result according to the pronunciation of each digit in the numeric string, but does not take the whole numeric string as a whole, so that the recommendation result of a certain digit in the whole numeric string may be a character with the same pronunciation as the digit, but not a digit, and the recommendation result of the whole numeric string is a mixture of the character and the digit.
[ summary of the invention ]
The invention provides a method, a device and a readable medium for recommending digital information in voice input, which are used for improving the accuracy of recommended information of a digital string and further improving the input operation cost and the input efficiency of the digital string.
The invention provides a method for recommending digital information in voice input, which comprises the following steps:
receiving a voice input request which carries the pronunciation of the numeric string and is input by a user;
respectively acquiring a first recommendation score of first recommendation information fed back by a preset language model to the pronunciation of the digital string and a second recommendation score of second recommendation information fed back by the preset language model to the pronunciation of the digital string;
if the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score;
and displaying the second recommendation information corresponding to the second recommendation score to the user.
Further optionally, in the method as described above, if the length of the number string is not equal to the preset number length threshold, the method includes:
judging and determining that the second recommendation score is smaller than the first recommendation score;
and displaying the first recommendation information corresponding to the first recommendation score to the user.
Further optionally, in the method, obtaining a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model specifically includes:
acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from the language model;
according to the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
arranging the first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form the first recommendation information;
acquiring a first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
and calculating the first recommendation score of the first recommendation information fed back to the pronunciation of the digital string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string.
Further optionally, in the method as described above, before obtaining the first recommendation score of the first recommendation information fed back to the pronunciation of the numeric string by the preset language model, the method further includes:
collecting a plurality of sentences from the Internet as corpora to form a corpus;
normalizing the format of the numbers in the linguistic data carrying the numbers in the plurality of linguistic data;
performing word segmentation processing on each language material to obtain a plurality of words;
forming the words into a dictionary base;
calculating the first occurrence probability of each word in the dictionary database in the corpus and the first occurrence probability of each word in the corpus together with the context word in the corresponding corpus;
and generating the language model according to the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus.
Further optionally, in the method, obtaining a second recommendation score of second recommendation information fed back to the pronunciation of the digital string by a preset digital model specifically includes:
acquiring a second recommendation result corresponding to each digit of the reading of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
judging whether the number of digits of the second recommendation result continuously acquired from the digital model based on the digital string is equal to the preset number length threshold value or not;
if so, arranging the second recommendation results corresponding to all the bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information;
and calculating the second recommendation score of the second recommendation information fed back to the pronunciation of the digital string by the digital model according to the second occurrence probability of the second recommendation result corresponding to each bit in the pronunciation of the digital string.
Further optionally, in the method described above, if the number of bits of the second recommendation result continuously obtained from the digital model based on the number string is not equal to the preset number length threshold, the method further includes:
and stopping processing the voice input request, and setting the second recommendation score to be 0.
Further optionally, in the method, before obtaining a second recommendation score of second recommendation information fed back to the reading of the digital string by a preset digital model, the method further includes:
setting the second probability of occurrence of an N-gram; the N-ary number represents a consecutive N numbers, and the second probability of occurrence of the N-ary number represents a probability of occurrence that a last bit of the consecutive N numbers occurs based on a preceding N-1 bits; n is a positive integer greater than or equal to 1;
generating the digital model according to the N-gram and the second probability of occurrence of the N-gram.
The invention provides a recommendation device of digital information in voice input, which comprises:
the receiving module is used for receiving a voice input request which carries the pronunciation of the numeric string and is input by a user;
the acquisition module is used for respectively acquiring a first recommendation score of first recommendation information fed back by a preset language model to the pronunciation of the digital string and a second recommendation score of second recommendation information fed back by the preset language model to the pronunciation of the digital string;
the determining module is used for judging and determining that the second recommendation score is larger than the first recommendation score if the length of the number string is equal to a preset number length threshold;
and the display module is used for displaying the second recommendation information corresponding to the second recommendation score to the user.
Further optionally, in the apparatus as described above, the determining module is further configured to determine and determine that the second recommendation score is smaller than the first recommendation score if the length of the number string is not equal to the preset number length threshold;
the display module is further configured to display the first recommendation information corresponding to the first recommendation score to the user.
Further optionally, in the apparatus as described above, the obtaining module includes:
the acquisition unit based on the language model is used for acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the digital string by a preset language model;
the acquisition unit based on the digital model is used for acquiring a second recommendation score of second recommendation information fed back by the preset digital model to the pronunciation of the digital string;
further, the language model-based obtaining unit is specifically configured to:
acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from the language model;
according to the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
arranging the first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form the first recommendation information;
acquiring a first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
and calculating the first recommendation score of the first recommendation information fed back to the pronunciation of the digital string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string.
Further optionally, the apparatus as described above further includes:
the system comprises an acquisition module, a database module and a database module, wherein the acquisition module is used for acquiring a plurality of sentences from the Internet as corpora to form a corpus;
the normalization module is used for normalizing the format of the number in the linguistic data carrying the number in the plurality of linguistic data;
the word segmentation module is used for carrying out word segmentation processing on the speech materials to obtain a plurality of words; and forming a dictionary base by the plurality of words;
a calculation module, configured to calculate the first occurrence probability of each word in the dictionary repository in the corpus and the first occurrence probability of each word appearing in the corpus together with a context word in the corresponding corpus;
and the language model generating module is used for generating the language model according to the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus which appear in the corpus together.
Further optionally, in the apparatus as described above, the obtaining unit based on the digital model is specifically configured to:
acquiring a second recommendation result corresponding to each digit of the reading of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
judging whether the number of digits of the second recommendation result continuously acquired from the digital model based on the digital string is equal to the preset number length threshold value or not;
if so, arranging the second recommendation results corresponding to all the bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information;
and calculating the second recommendation score of the second recommendation information fed back to the pronunciation of the digital string by the digital model according to the second occurrence probability of the second recommendation result corresponding to each bit in the pronunciation of the digital string.
Further optionally, in the apparatus as described above, the obtaining unit based on a digital model is specifically further configured to suspend processing of the voice input request and set the second recommendation score to 0 if the number of bits of the second recommendation result continuously obtained from the digital model based on the digital string is not equal to the preset number length threshold.
Further optionally, in the apparatus described above, the apparatus further includes:
a setting module for setting the second probability of occurrence of an N-gram; the N-ary number represents a consecutive N numbers, and the second probability of occurrence of the N-ary number represents a probability of occurrence that a last bit of the consecutive N numbers occurs based on a preceding N-1 bits; n is a positive integer greater than or equal to 1;
a digital model generating module, configured to generate the digital model according to the N-ary number and the second occurrence probability of the N-ary number.
The present invention also provides a computer apparatus, the apparatus comprising:
one or more processors;
a memory for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement a method of recommending digital information in a speech input as described above.
The present invention also provides a computer-readable medium having stored thereon a computer program which, when executed by a processor, implements the method of recommending digital information in a speech input as described above.
The invention relates to a method, a device and a readable medium for recommending digital information in voice input, which receive a voice input request which is input by a user and carries pronunciation of a digital string; respectively acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model and a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by a preset digital model; if the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score; and displaying second recommendation information corresponding to the second recommendation score to the user. By adopting the technical scheme of the invention, if the numeric string is a number such as a telephone number, the recommendation information can be ensured to be composed of pure numbers, so that the accuracy of the recommendation information of the numeric string can be improved, and thus, when a user inputs the numeric string, the input can be realized directly according to the recommendation information, the input operation cost of the numeric string can be effectively improved, and the input efficiency of the numeric string is improved.
[ description of the drawings ]
Fig. 1 is a flowchart of an embodiment of a method for recommending digital information in voice input according to the present invention.
Fig. 2 is a block diagram of a first embodiment of a device for recommending digital information during voice input according to the present invention.
Fig. 3 is a block diagram of a second embodiment of the apparatus for recommending digital information in voice input according to the present invention.
FIG. 4 is a block diagram of an embodiment of a computer device of the present invention.
Fig. 5 is an exemplary diagram of a computer device provided by the present invention.
[ detailed description ] embodiments
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a flowchart of an embodiment of a method for recommending digital information in voice input according to the present invention. As shown in fig. 1, the method for recommending digital information in voice input according to this embodiment may specifically include the following steps:
100. receiving a voice input request which carries the pronunciation of the numeric string and is input by a user;
the execution main body of the method for recommending digital information in voice input according to this embodiment may be a device for recommending digital information in voice input, for example, the device for recommending digital information in voice input may be disposed on a server side of an information input device supporting voice input, such as an input method, and implement recommendation of digital information in voice input to a user according to a voice input request of the user carrying a numeric string.
The method for recommending digital information in voice input according to the embodiment can be applied to Chinese input, and can also be applied to other languages of characters with the same or similar pronunciation as that of the digits, such as Japanese.
101. Respectively acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model and a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by a preset digital model;
in this embodiment, besides the language model, a digital model needs to be provided, and the digital model of this embodiment is used for specifically identifying and recommending the numbers corresponding to the digital pronunciations, thereby improving the accuracy of digital information recommendation.
For example, before step 101, specifically, the following language model generation may be further included, and specifically, the following steps may be included:
(a1) collecting a plurality of sentences from the Internet as corpora to form a corpus;
the language model of the present embodiment is a full-network language model generated based on the internet. In this embodiment, the number of the plurality of corpora collected from the internet for generating the corpus may be tens of thousands, including internet sentences of the whole network.
(b1) Normalizing the format of the numbers in the linguistic data with the numbers in the plurality of linguistic data;
because part of the language material includes digital information, namely, the sentence represented by the character carries a number string such as a telephone number or other numbers, some number strings are in full-angle format in the character, and some number strings are in half-angle format, in order to facilitate the unification of the number strings, the numbers in full-angle format are all uniformly adjusted to be the numbers in half-angle format. Or the numbers in the half-angle format can be uniformly adjusted to the numbers in the full-angle format.
(c1) Performing word segmentation processing on each corpus to obtain a plurality of words;
(d1) forming a dictionary base by a plurality of words;
the word segmentation process in this embodiment may refer to word segmentation process in the language model generation process in the related prior art, and is not described herein again. After word segmentation processing is carried out on each corpus in the corpus, a plurality of words can be obtained, and the number of the words is far larger than that of expected corpora in the corpus. For example, the number of words comprising the lexicon library may be up to 20-30 ten thousand.
(e1) Calculating a first occurrence probability of each word in the dictionary database in the corpus and a first occurrence probability of each word and a context word in the corresponding corpus which appear in the corpus together;
(f1) and generating a language model according to the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus.
After the above processing, a first occurrence probability of each word in the dictionary database in the corpus and a first occurrence probability of each word in the corpus together with the context word in the corresponding corpus can be calculated according to the dictionary database and the corpus. The first probability of occurrence of each word in the lexicon library in the corpus is equal to the number of occurrences of the word in the corpus divided by the total number of occurrences of all words in the lexicon library in the corpus. The first probability of occurrence of a word in the corpus together with the context word in the corresponding corpus is equal to the number of times of occurrence of the word together with the context word in all the corpora in the corpus divided by the total number of times of occurrence of the word together with all the context words in all the corpora in the corpus. Where the number of contextual words may be one, two, or more. The first probability of occurrence of a word in the corpus together with the context word in the corresponding corpus may also be referred to as the first probability of occurrence of the N-gram of the word. The N-gram of the term may include the term and N-1 context terms of the term.
For example, a 1-gram of the word includes only the word, a 2-gram of the word may include the word and a context word before or after the word, and the first probability of occurrence of the 2-gram of the word is the probability of occurrence of the word, together with a context word before or after the word, in the corpus. The 3-gram of the word may include 2 context words before and/or after the word, and the first probability of occurrence of the 3-gram of the word is a first probability of occurrence that 2 context words before and/or after the word occur together in the corpus. By analogy, the first occurrence probability of the N-gram of the word can be obtained.
And finally, storing the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus in a large information table to obtain the language model. That is, the language model is similar to a large information table in which the occurrence probability of each word and the corresponding N-gram is stored.
The language model generated in this embodiment normalizes the format of the numbers in the corpus carrying the numbers, so that the probability of the words formed by the numbers is not dispersed, the first occurrence probability of the words formed by the numbers in the language model is improved, and the probability that the numbers corresponding to the digital readings are recommended is improved.
Based on the generated language model, step 101 may obtain a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by the language model, and for example, may specifically include the following steps:
(a2) acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from a language model;
(b2) according to the size of the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
(c2) arranging first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form first recommendation information;
(d2) acquiring a first occurrence probability of a first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
(e2) and calculating a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the numeric string.
When the language model of this embodiment processes the numeric string, since the language input request is received, the numeric string in the voice input request also corresponds to the pronunciation of the numeric string, and in the language model, the pronunciation of each digit in the numeric string may include some characters in addition to the corresponding digit. For example, in Chinese, the pronunciation of the number 1 correspondingly has the Chinese characters "one", "clothing", "doctor", etc. Therefore, for the pronunciation of each digit from front to back in the digit string, the plurality of candidate recommendation information and the first occurrence probability corresponding to each candidate recommendation information can be obtained from the language model. Then, the candidate information with the highest first occurrence probability can be obtained from the plurality of candidate information as the first recommendation result corresponding to the digit of the bit. Then, the first recommendation results corresponding to each digit in the pronunciation of the digit string can be arranged according to the sequence of the digits from front to back, so as to form the first recommendation information corresponding to the digit string. The recommendation information is the overall recommendation result output by the language model for the numeric string.
Then, the first occurrence probability of the first deduction result corresponding to each bit in the pronunciation of the numeric string can be obtained from the language model; and calculating a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the numeric string. For example, the first occurrence probability of each bit in the digital string from the front to the back may be multiplied, and the logarithm after the multiplication result may be taken as the first recommendation score corresponding to the first recommendation information. Or in practical application, the first recommendation score of the first recommendation information may be obtained in other manners, as long as a proportional relationship between the score and the multiplication result is ensured, which is not repeated here.
Further optionally, in this embodiment, before step 101, the following generation of a digital model may be further included, and specifically, the following steps may be included:
(a3) setting a second occurrence probability of the N-ary number; the N-digit number represents continuous N digits, and the second occurrence probability of the N-digit number represents the occurrence probability that the last digit of the continuous N digits appears based on the previous N-1 digits; n is a positive integer greater than or equal to 1;
(b3) and generating the digital model according to the N-ary numbers and the second occurrence probability of the N-ary numbers.
The numerical model in this embodiment is used only for the recommended numbers. For example, for a 1-digit number, 10 digits from 0 to 9, the probability of occurrence for each digit is the same, with a probability equal to 0.1. For binary numbers, i.e., two consecutive numbers, any of a total of 100 combinations of 00, 01, and up to 99 are possible. In this embodiment 00 indicates that the first digit 0 is followed by a digit 0, and 01 indicates that the first digit 0 is followed by a digit 1. 99 indicates that the first digit 9 is followed by the digit 9. But all represent no single value. In a binary number, 00 is combined, i.e. the first number 0 has an occurrence probability of 0.1, and the number following the number 0 may be any one of 0 to 9, i.e. the probability of 0 appearing again is also 0.1, so the probability of 0 appearing after the first number 0 is equal to 0.1 x 0.1 — 0.01. In the same way, the probability of any 2-element number combination is 0.01. For ternary numbers, i.e., representing three consecutive numbers, any one of a total of 1000 combinations of 000, 001, up to 999 is possible. In this embodiment, 000 indicates that 0 is present in the second digit after the first digit 0 and 0 is present in the third digit, and 001 indicates that 0 is present in the second digit after the first digit 0 and 1 is present in the third digit. 999 indicates that the 2 nd digit after the first digit 9 appears as a digit 9 and the third digit also appears as a digit 9. But all represent no single value. In the ternary number, the probability of the occurrence of the second digit 0 after the first digit 0 in the 000 combinations is 0.01, and the probability of the occurrence of the third digit after the first digit 0 in the 000 combinations may be any one of 0 to 9, and there is a probability of 0.1, so that the probability of the occurrence of the 000 combinations is equal to 0.001 × 0.1 — 0.001. In the same way, the probability of any 3-element number combination is 0.001. By analogy, the probability of the occurrence of any N-digit number can be obtained according to the number of digits needing to be stored. And are not described in detail herein. For example, in this embodiment, it may be preferable to take the example of acquiring a 1-digit number, a 2-digit number, and a 3-digit number.
And finally, taking the probability of the obtained N-element number combination as a second occurrence probability of the N-element number. The N-grams, together with the second probability of occurrence of the N-grams, are stored in a large table of information as a numerical model.
Based on the generated digital model, step 101 may obtain a second recommendation score of second recommendation information fed back to the pronunciation of the string by the preset digital model, for example, the method may specifically include the following steps:
(a4) acquiring a second recommendation result corresponding to each digit of the pronunciation of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
(b4) judging whether the digit of a second recommendation result continuously acquired from the digital model based on the digit string is equal to a preset number length threshold value or not; if yes, executing step (c 4); otherwise, performing step (e 4);
(c4) arranging second recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information; performing step ((d 4);
(d4) and calculating a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by the digital model according to a second occurrence probability of a second recommendation result corresponding to each digit in the pronunciation of the numeric string, and ending.
(e4) The processing of the voice input request is suspended and the second recommendation score is set to 0.
In this embodiment, when the digital string is identified and recommended according to the digital model, since the digital string includes consecutive multi-digit numbers, the digital model needs to perform consecutive self-jumping when processing the digital string, so as to recommend the second recommendation result for each digit in the digital string. Wherein the number of consecutive self-jumps of the digital model is the number of bits included in the digital string, i.e. the number of bits of the second recommendation which is consecutively obtained from the digital model based on the digital string. And if the digital model finishes continuous self-jumping, judging whether the digit of a second recommendation result continuously acquired from the digital model based on the digital string is equal to a preset number length threshold value or not. If yes, the whole number string can be the number of the preset mode at the moment, and the number string can be recommended as a whole. Correspondingly, second recommendation results corresponding to all bits in the pronunciation of the numeric string are arranged from front to back in the pronunciation of the numeric string according to the corresponding bits to form second recommendation information, and the second recommendation information is recommended by the numeric model, is a numeric string and does not contain other character information. The preset number length threshold in this embodiment is preset, for example, if the preset number length threshold is used for identifying a phone number, since a general phone number is 11 digits, the preset number length threshold may be 11 digits, so that the accuracy of a recommended phone number may be improved, and the problem of missing digits and the like may be avoided.
In addition, there are also some service phones like 10086 or 10010, which have only 5 bits. In this embodiment, a plurality of preset number length thresholds may be set. If the lengths of the 5-digit and 11-digit telephone numbers are set simultaneously, if the digit model has 5 successive self-jumps, the digit string is 5 digits, which can be a telephone number, and then the second recommendation information can be formed in the manner described above. If the number of consecutive self-hops of the digital model is 11, it means that the digital string is 11 digits, and may be a telephone number, and in this case, the second recommendation information may be formed as described above. And calculating a second recommendation score of second recommendation information fed back by the digital model to the pronunciation of the numeric string according to a second occurrence probability of a second recommendation result corresponding to each digit in the pronunciation of the numeric string, for example, multiplying the second occurrence probabilities of the second recommendation results of each digit corresponding to the pronunciation of the numeric string, and further taking the logarithm of the obtained product as the second recommendation score of the second recommendation information. The same can be done in other ways as long as it is ensured that the second recommendation score is proportional to the resulting product.
If the number of consecutive self-hops of the digital model is not 5 times or 11 times, the number string is represented as a non-telephone number, and the recommended score can be set to 0 directly.
In the above embodiments, the numerical model is used for recommending the telephone number as an example. In practical application, the method can also be used for recommending other numbers, such as patent application numbers or other numbers in a patent management system, and at the moment, only the preset number length threshold value needs to be set as the length of the number to be recommended, so that the method can be used for recommending the number of the category when a voice input request carrying the number of the category is received, and the recommendation accuracy of the number of the category is improved.
102. If the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score;
103. and displaying second recommendation information corresponding to the second recommendation score to the user.
In this embodiment, if the length of the numeric string is equal to the preset number length threshold, the second recommendation score is a non-zero result. Because the number of words in the language model is much greater than the number of N-grams in the numerical model, the second probability of occurrence of each number is much greater than the first probability of occurrence of each word in the language model. Therefore, the second recommendation score obtained based on the second recommendation information recommended for the digital string by the digital model is far greater than the first recommendation score obtained based on the first recommendation information recommended for the digital string by the language model. Therefore, the second recommendation score can be judged and determined to be larger than the first recommendation score; and displaying second recommendation information corresponding to the second recommendation score to the user.
In addition, optionally, if the length of the number string is not equal to the preset number length threshold, which may be only a string of digits of a non-telephone number or a special number, and the second recommendation score is 0, this may further include: judging and determining that the second recommendation score is smaller than the first recommendation score; and displaying first recommendation information corresponding to the first recommendation score to the user.
In voice input in japanese, for example, the user inputs the telephone number 0942458280 by voice, because the pronunciation of "person" and number 2 of japanese in japanese is similar, the recommendation result corresponding to the language model may be "0942458 persons 80", which results in poor accuracy and price of the recommendation information of the number string, and with the technical solution of this embodiment, the length of the numeric string is exactly equal to the length of the preset telephone number, the second recommendation score of '0942458280' recommended based on the numeric model is larger than the first recommendation score of '0942458 people 80' recommended by the language model, finally, the telephone number '0942458280' corresponding to the second recommendation score is displayed to the user, therefore, when the numeric string is a telephone number, the accuracy of the recommendation information of the numeric string is improved, and a user can directly input the recommendation information, so that the input operation cost of the numeric string is effectively improved, and the input efficiency of the numeric string is improved.
In the method for recommending digital information in voice input of the embodiment, a voice input request carrying pronunciation of a digital string and input by a user is received; respectively acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model and a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by a preset digital model; if the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score; and displaying second recommendation information corresponding to the second recommendation score to the user. By adopting the technical scheme of the embodiment, if the numeric string is a number such as a telephone number, the recommendation information can be guaranteed to be composed of pure numbers, so that the accuracy of the recommendation information of the numeric string can be improved, and thus, when a user inputs the numeric string, the user can input the numeric string directly according to the recommendation information, the input operation cost of the numeric string can be effectively improved, and the input efficiency of the numeric string is improved.
Fig. 2 is a block diagram of a first embodiment of a device for recommending digital information during voice input according to the present invention. As shown in fig. 2, the apparatus for recommending digital information in voice input according to this embodiment may specifically include: the device comprises a receiving module 10, an obtaining module 11, a determining module 12 and a displaying module 13.
The receiving module 10 is configured to receive a voice input request carrying a pronunciation of a numeric string, which is input by a user;
the obtaining module 11 is configured to obtain a first recommendation score of first recommendation information fed back by a preset language model to a pronunciation of a numeric string in the voice input request received by the receiving module 10, and obtain a second recommendation score of second recommendation information fed back by the preset language model to the pronunciation of the numeric string in the voice input request received by the receiving module 10;
the determining module 12 is configured to determine and determine that the second recommendation score acquired by the acquiring module 11 is greater than the first recommendation score if the length of the numeric string in the voice input request received by the receiving module 10 is equal to a preset number length threshold;
the display module 13 is configured to display, to the user, second recommendation information corresponding to the second recommendation score acquired by the acquisition module 11.
The implementation principle and technical effect of implementing digital information recommendation in voice input by using the module in the apparatus for recommending digital information in voice input according to this embodiment are the same as those of the related method embodiment, and reference may be made to the description of the related method embodiment in detail, which is not described herein again.
Fig. 3 is a block diagram of a second embodiment of the apparatus for recommending digital information in voice input according to the present invention. As shown in fig. 3, the recommendation apparatus for digital information in voice input according to the present embodiment further introduces the technical solution of the present invention in more detail based on the technical solution of the above embodiment shown in fig. 2.
In the device for recommending digital information during voice input according to this embodiment, the determining module 12 is further configured to determine and determine that the second recommendation score obtained by the obtaining module 11 is smaller than the first recommendation score if the length of the digital string is not equal to the preset number length threshold;
the display module 13 is further configured to display, to the user, first recommendation information corresponding to the first recommendation score acquired by the acquisition module 11.
Further optionally, as shown in fig. 3, in the apparatus for recommending digital information in audio input of this embodiment, the obtaining module 11 includes:
the language model-based obtaining unit 111 is configured to obtain a first recommendation score of first recommendation information fed back by a preset language model to the pronunciation of the numeric string;
the digital model-based obtaining unit 112 is configured to obtain a second recommendation score of second recommendation information fed back to the reading of the string by a preset digital model;
the language model-based obtaining unit 111 is specifically configured to:
acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from a language model;
according to the size of the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
arranging first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form first recommendation information;
acquiring a first occurrence probability of a first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
and calculating a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the numeric string.
Further optionally, as shown in fig. 3, the apparatus for recommending digital information in audio input according to this embodiment further includes:
the collecting module 14 is used for collecting a plurality of sentences from the internet as corpora to form a corpus;
the normalization module 15 is configured to normalize a format of a number in a corpus carrying numbers in a plurality of corpora collected and generated by the collection module 14;
the word segmentation module 16 is configured to perform word segmentation processing on each corpus in the corpus collected and generated by the collection module 14 to obtain a plurality of words; and forming a dictionary base by a plurality of words;
the calculating module 17 is configured to calculate a first occurrence probability of each word in the dictionary database in the corpus, which is processed and obtained by the word segmentation module 16, and a first occurrence probability of each word and a context word in the corresponding corpus occurring together in the corpus;
the language model generating module 18 is configured to generate a language model according to the first occurrence probability of each word in the corpus, which is calculated by the calculating module 17, and the first occurrence probability of each word and the context word in the corresponding corpus which appear in the corpus together.
Further optionally, in the apparatus for recommending digital information in audio input according to this embodiment, the obtaining unit 112 based on a digital model is specifically configured to:
acquiring a second recommendation result corresponding to each digit of the pronunciation of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
judging whether the digit of a second recommendation result continuously acquired from the digital model based on the digit string is equal to a preset number length threshold value or not;
if so, arranging second recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information;
and calculating a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by the digital model according to a second occurrence probability of a second recommendation result corresponding to each digit in the pronunciation of the numeric string.
Further optionally, in the apparatus for recommending digital information in audio input according to this embodiment, the obtaining unit 112 based on a digital model is further specifically configured to, if the number of bits of the second recommendation result continuously obtained from the digital model based on the digital string is not equal to the preset number length threshold, suspend processing of the voice input request, and set the second recommendation score to 0.
Further optionally, as shown in fig. 3, the apparatus for recommending digital information in audio input according to this embodiment further includes:
the setting module 19 is used for setting a second occurrence probability of the N-ary number; the N-digit number represents continuous N digits, and the second occurrence probability of the N-digit number represents the occurrence probability that the last digit of the continuous N digits appears based on the previous N-1 digits; n is a positive integer greater than or equal to 1;
the digital model generating module 20 is configured to set the second occurrence probability according to the N-gram and the N-gram set by the setting module 19 to generate the digital model.
At this time, correspondingly, the language model-based obtaining unit 111 obtains the first recommendation score of the first recommendation information fed back by the language model generated by the language model generation module 18 to the reading of the numeric string in the voice input request received by the reception module 10.
At this time, correspondingly, the digital model-based obtaining unit 112 obtains the second recommendation score of the second recommendation information fed back by the digital model generated by the digital model generation module 20 to the reading of the digital string in the voice input request received by the receiving module 10.
The determining module 12 is configured to determine and determine that the second recommendation score obtained by the obtaining unit 112 based on the digital model is greater than the first recommendation score obtained by the obtaining unit 111 based on the language model if the length of the numeric string in the voice input request received by the receiving module 10 is equal to the preset number length threshold;
the display module 13 is configured to display, to the user, second recommendation information corresponding to the second recommendation score obtained by the obtaining unit 112 based on the digital model according to the determination result of the determining module 12.
If the determining module 12 is configured to determine and determine that the first recommendation score obtained by the language model-based obtaining unit 111 is greater than the second recommendation score obtained by the digital model-based obtaining unit 112 if the length of the numeric string in the voice input request received by the receiving module 10 is not equal to the preset number length threshold;
the display module 13 is configured to display, to the user, second recommendation information corresponding to the first recommendation score acquired by the language model-based acquisition unit 111.
The implementation principle and technical effect of implementing digital information recommendation in voice input by using the module in the apparatus for recommending digital information in voice input according to this embodiment are the same as those of the related method embodiment, and reference may be made to the description of the related method embodiment in detail, which is not described herein again.
FIG. 4 is a block diagram of an embodiment of a computer device of the present invention. As shown in fig. 4, the computer device of the present embodiment includes: one or more processors 30, and a memory 40, the memory 40 for storing one or more programs, when the one or more programs stored in the memory 40 are executed by the one or more processors 30, cause the one or more processors 30 to implement the method for recommending digital information in a speech input as in the embodiment shown in fig. 1-3 above. The embodiment shown in fig. 4 is exemplified by including a plurality of processors 30. The computer apparatus of the present embodiment may be a server apparatus of an information input device such as an input method.
For example, fig. 5 is an exemplary diagram of a computer device provided by the present invention. FIG. 5 illustrates a block diagram of an exemplary computer device 12a suitable for use in implementing embodiments of the present invention. The computer device may be a server device for an information input apparatus such as an input method. The computer device 12a shown in fig. 5 is only an example and should not bring any limitations to the functionality or scope of use of the embodiments of the present invention.
As shown in FIG. 5, computer device 12a is in the form of a general purpose computing device. The components of computer device 12a may include, but are not limited to: one or more processors 16a, a system memory 28a, and a bus 18a that connects the various system components (including the system memory 28a and the processors 16 a).
Bus 18a represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, and a processor or local bus using any of a variety of bus architectures. By way of example, such architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, micro-channel architecture (MAC) bus, enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.
Computer device 12a typically includes a variety of computer system readable media. Such media may be any available media that is accessible by computer device 12a and includes both volatile and nonvolatile media, removable and non-removable media.
The system memory 28a may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM)30a and/or cache memory 32 a. Computer device 12a may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34a may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 5, commonly referred to as a "hard drive"). Although not shown in FIG. 5, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18a by one or more data media interfaces. System memory 28a may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of the various embodiments of the invention described above in fig. 1-3.
A program/utility 40a having a set (at least one) of program modules 42a may be stored, for example, in system memory 28a, such program modules 42a including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may include an implementation of a network environment. Program modules 42a generally perform the functions and/or methodologies described above in connection with the various embodiments of fig. 1-3 of the present invention.
Computer device 12a may also communicate with one or more external devices 14a (e.g., keyboard, pointing device, display 24a, etc.), with one or more devices that enable a user to interact with computer device 12a, and/or with any devices (e.g., network card, modem, etc.) that enable computer device 12a to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface 22 a. Also, computer device 12a may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN) and/or a public network, such as the Internet) through network adapter 20 a. As shown, network adapter 20a communicates with the other modules of computer device 12a via bus 18 a. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with computer device 12a, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.
The processor 16a executes various functional applications and data processing by executing programs stored in the system memory 28a, for example, to implement the recommendation method of digital information in voice input shown in the above-described embodiment.
The present invention also provides a computer-readable medium on which a computer program is stored, which when executed by a processor implements the method of recommending digital information in a voice input as shown in the above embodiments.
The computer-readable media of this embodiment may include RAM30a, and/or cache memory 32a, and/or storage system 34a in system memory 28a in the embodiment illustrated in fig. 5 described above.
With the development of technology, the propagation path of computer programs is no longer limited to tangible media, and the computer programs can be directly downloaded from a network or acquired by other methods. Accordingly, the computer-readable medium in the present embodiment may include not only tangible media but also intangible media.
The computer-readable medium of the present embodiments may take any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
In the embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. For example, the above-described device embodiments are merely illustrative, and for example, the division of the units is only one logical functional division, and other divisions may be realized in practice.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.
The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (16)

1. A method for recommending digital information in a voice input, the method comprising:
receiving a voice input request which carries the pronunciation of the numeric string and is input by a user;
respectively acquiring a first recommendation score of first recommendation information fed back by a preset language model to the pronunciation of the digital string and a second recommendation score of second recommendation information fed back by the preset language model to the pronunciation of the digital string;
if the length of the numeric string is equal to a preset number length threshold, judging and determining that the second recommendation score is larger than the first recommendation score;
and displaying the second recommendation information corresponding to the second recommendation score to the user.
2. The method of claim 1, wherein if the length of the digit string is not equal to the predetermined number length threshold, the method comprises:
judging and determining that the second recommendation score is smaller than the first recommendation score;
and displaying the first recommendation information corresponding to the first recommendation score to the user.
3. The method according to claim 1, wherein obtaining a first recommendation score of first recommendation information fed back to the pronunciation of the numeric string by a preset language model specifically comprises:
acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from the language model;
according to the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
arranging the first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form the first recommendation information;
acquiring a first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
and calculating the first recommendation score of the first recommendation information fed back to the pronunciation of the digital string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string.
4. The method according to claim 3, wherein before obtaining the first recommendation score of the first recommendation information fed back to the pronunciation of the numeric string by the preset language model, the method further comprises:
collecting a plurality of sentences from the Internet as corpora to form a corpus;
normalizing the format of the numbers in the linguistic data carrying the numbers in the plurality of linguistic data;
performing word segmentation processing on each language material to obtain a plurality of words;
forming the words into a dictionary base;
calculating the first occurrence probability of each word in the dictionary database in the corpus and the first occurrence probability of each word in the corpus together with the context word in the corresponding corpus;
and generating the language model according to the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus.
5. The method according to claim 1, wherein obtaining a second recommendation score of second recommendation information fed back to the pronunciation of the numeric string by a preset numeric model specifically comprises:
acquiring a second recommendation result corresponding to each digit of the reading of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
judging whether the number of digits of the second recommendation result continuously acquired from the digital model based on the digital string is equal to the preset number length threshold value or not;
if so, arranging the second recommendation results corresponding to all the bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information;
and calculating the second recommendation score of the second recommendation information fed back to the pronunciation of the digital string by the digital model according to the second occurrence probability of the second recommendation result corresponding to each bit in the pronunciation of the digital string.
6. The method of claim 5, wherein if the number of bits of the second recommendation obtained from the digital model based on the number string is not equal to the preset number length threshold, the method further comprises:
and stopping processing the voice input request, and setting the second recommendation score to be 0.
7. The method according to claim 5 or 6, wherein before obtaining a second recommendation score of second recommendation information fed back to the reading of the numeric string by a preset numeric model, the method further comprises:
setting the second probability of occurrence of an N-gram; the N-ary number represents a consecutive N numbers, and the second probability of occurrence of the N-ary number represents a probability of occurrence that a last bit of the consecutive N numbers occurs based on a preceding N-1 bits; n is a positive integer greater than or equal to 1;
generating the digital model according to the N-gram and the second probability of occurrence of the N-gram.
8. An apparatus for recommending digital information in a voice input, said apparatus comprising:
the receiving module is used for receiving a voice input request which carries the pronunciation of the numeric string and is input by a user;
the acquisition module is used for respectively acquiring a first recommendation score of first recommendation information fed back by a preset language model to the pronunciation of the digital string and a second recommendation score of second recommendation information fed back by the preset language model to the pronunciation of the digital string;
the determining module is used for judging and determining that the second recommendation score is larger than the first recommendation score if the length of the number string is equal to a preset number length threshold;
and the display module is used for displaying the second recommendation information corresponding to the second recommendation score to the user.
9. The apparatus of claim 8, wherein:
the determining module is further configured to determine and determine that the second recommendation score is smaller than the first recommendation score if the length of the number string is not equal to the preset number length threshold;
the display module is further configured to display the first recommendation information corresponding to the first recommendation score to the user.
10. The apparatus of claim 8, wherein the obtaining module comprises:
the acquisition unit based on the language model is used for acquiring a first recommendation score of first recommendation information fed back to the pronunciation of the digital string by a preset language model;
the acquisition unit based on the digital model is used for acquiring a second recommendation score of second recommendation information fed back by the preset digital model to the pronunciation of the digital string;
further, the language model-based obtaining unit is specifically configured to:
acquiring a plurality of alternative recommendation information corresponding to each digit of the pronunciation of the digital string from front to back and a first occurrence probability corresponding to each alternative recommendation information from the language model;
according to the first occurrence probability of the multiple candidate recommendation information, obtaining a first recommendation result with the maximum first occurrence probability as a corresponding bit from the multiple candidate recommendation information;
arranging the first recommendation results corresponding to all bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form the first recommendation information;
acquiring a first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string from the language model;
and calculating the first recommendation score of the first recommendation information fed back to the pronunciation of the digital string by the language model according to the first occurrence probability of the first recommendation result corresponding to each digit in the pronunciation of the digital string.
11. The apparatus of claim 10, further comprising:
the system comprises an acquisition module, a database module and a database module, wherein the acquisition module is used for acquiring a plurality of sentences from the Internet as corpora to form a corpus;
the normalization module is used for normalizing the format of the number in the linguistic data carrying the number in the plurality of linguistic data;
the word segmentation module is used for carrying out word segmentation processing on the speech materials to obtain a plurality of words; and forming a dictionary base by the plurality of words;
a calculation module, configured to calculate the first occurrence probability of each word in the dictionary repository in the corpus and the first occurrence probability of each word appearing in the corpus together with a context word in the corresponding corpus;
and the language model generating module is used for generating the language model according to the first occurrence probability of each word in the corpus and the first occurrence probability of each word and the context word in the corresponding corpus which appear in the corpus together.
12. The apparatus according to claim 10, wherein the digital model-based obtaining unit is specifically configured to:
acquiring a second recommendation result corresponding to each digit of the reading of the digital string from front to back and a second occurrence probability of the second recommendation result corresponding to each digit from the digital model;
judging whether the number of digits of the second recommendation result continuously acquired from the digital model based on the digital string is equal to the preset number length threshold value or not;
if so, arranging the second recommendation results corresponding to all the bits in the pronunciation of the numeric string according to the sequence of the corresponding bits from front to back in the pronunciation of the numeric string to form second recommendation information;
and calculating the second recommendation score of the second recommendation information fed back to the pronunciation of the digital string by the digital model according to the second occurrence probability of the second recommendation result corresponding to each bit in the pronunciation of the digital string.
13. The apparatus according to claim 12, wherein the digital model-based obtaining unit is further configured to suspend processing of the voice input request and set the second recommendation score to 0 if the number of digits of the second recommendation result obtained from the digital model based on the digital string is not equal to the preset number length threshold.
14. The apparatus of claim 12 or 13, further comprising:
a setting module for setting the second probability of occurrence of an N-gram; the N-ary number represents a consecutive N numbers, and the second probability of occurrence of the N-ary number represents a probability of occurrence that a last bit of the consecutive N numbers occurs based on a preceding N-1 bits; n is a positive integer greater than or equal to 1;
a digital model generating module, configured to generate the digital model according to the N-ary number and the second occurrence probability of the N-ary number.
15. A computer device, the device comprising:
one or more processors;
a memory for storing one or more programs,
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-7.
16. A computer-readable medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-7.
CN201710655037.1A 2017-08-03 2017-08-03 Method, device and readable medium for recommending digital information in voice input Active CN107632718B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710655037.1A CN107632718B (en) 2017-08-03 2017-08-03 Method, device and readable medium for recommending digital information in voice input

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710655037.1A CN107632718B (en) 2017-08-03 2017-08-03 Method, device and readable medium for recommending digital information in voice input

Publications (2)

Publication Number Publication Date
CN107632718A CN107632718A (en) 2018-01-26
CN107632718B true CN107632718B (en) 2021-01-22

Family

ID=61099498

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710655037.1A Active CN107632718B (en) 2017-08-03 2017-08-03 Method, device and readable medium for recommending digital information in voice input

Country Status (1)

Country Link
CN (1) CN107632718B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1212404A (en) * 1997-09-19 1999-03-31 国际商业机器公司 Method for identifying character/numeric string in Chinese speech recognition system
CN104102720A (en) * 2014-07-18 2014-10-15 上海触乐信息科技有限公司 Efficient input prediction method and device
CN104412219A (en) * 2012-06-25 2015-03-11 谷歌公司 Visual confirmation of voice recognized text input
CN106257441A (en) * 2016-06-30 2016-12-28 电子科技大学 A kind of training method of skip language model based on word frequency

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6922669B2 (en) * 1998-12-29 2005-07-26 Koninklijke Philips Electronics N.V. Knowledge-based strategies applied to N-best lists in automatic speech recognition systems
US8095365B2 (en) * 2008-12-04 2012-01-10 At&T Intellectual Property I, L.P. System and method for increasing recognition rates of in-vocabulary words by improving pronunciation modeling
CN102722525A (en) * 2012-05-15 2012-10-10 北京百度网讯科技有限公司 Methods and systems for establishing language model of address book names and searching voice

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1212404A (en) * 1997-09-19 1999-03-31 国际商业机器公司 Method for identifying character/numeric string in Chinese speech recognition system
CN104412219A (en) * 2012-06-25 2015-03-11 谷歌公司 Visual confirmation of voice recognized text input
CN104102720A (en) * 2014-07-18 2014-10-15 上海触乐信息科技有限公司 Efficient input prediction method and device
CN106257441A (en) * 2016-06-30 2016-12-28 电子科技大学 A kind of training method of skip language model based on word frequency

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于层级策略的连续数字串识别的研究;汤霖 等;《计算机工程与应用》;20030721(第21期);第83-86页 *
讯飞语音输入法怎样说才是阿拉伯数字;匿名用户;《wenwen.sogou.com/z/q747374485》;20130420;第1页 *

Also Published As

Publication number Publication date
CN107632718A (en) 2018-01-26

Similar Documents

Publication Publication Date Title
CN107220235B (en) Speech recognition error correction method and device based on artificial intelligence and storage medium
US10192545B2 (en) Language modeling based on spoken and unspeakable corpuses
CN107220232B (en) Keyword extraction method and device based on artificial intelligence, equipment and readable medium
CN107193973B (en) Method, device and equipment for identifying field of semantic analysis information and readable medium
US10803241B2 (en) System and method for text normalization in noisy channels
CN108388674B (en) Method and device for pushing information
CN107729313B (en) Deep neural network-based polyphone pronunciation distinguishing method and device
CN109241286B (en) Method and device for generating text
CN112395385B (en) Text generation method and device based on artificial intelligence, computer equipment and medium
CN110377750B (en) Comment generation method, comment generation device, comment generation model training device and storage medium
CN109815481B (en) Method, device, equipment and computer storage medium for extracting event from text
CN108595412B (en) Error correction processing method and device, computer equipment and readable medium
CN111143556A (en) Software function point automatic counting method, device, medium and electronic equipment
CN111930792A (en) Data resource labeling method and device, storage medium and electronic equipment
CN110704608A (en) Text theme generation method and device and computer equipment
CN111666379B (en) Event element extraction method and device
CN110738056A (en) Method and apparatus for generating information
CN111666405B (en) Method and device for identifying text implication relationship
CN112581297A (en) Information pushing method and device based on artificial intelligence and computer equipment
CN113158656A (en) Ironic content identification method, ironic content identification device, electronic device, and storage medium
CN112417875A (en) Configuration information updating method and device, computer equipment and medium
CN107632718B (en) Method, device and readable medium for recommending digital information in voice input
CN115620726A (en) Voice text generation method, and training method and device of voice text generation model
CN111401009B (en) Digital expression character recognition conversion method, device, server and storage medium
CN109036379B (en) Speech recognition method, apparatus and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant