WO2018121275A1

WO2018121275A1 - Method and apparatus for error connection of voice recognition in smart hardware device

Info

Publication number: WO2018121275A1
Application number: PCT/CN2017/116165
Authority: WO
Inventors: 杨英; 张倩倩
Original assignee: 北京奇虎科技有限公司
Priority date: 2016-12-29
Filing date: 2017-12-14
Publication date: 2018-07-05
Also published as: CN106710592B; CN106710592A

Abstract

A method and an apparatus for error connection of voice recognition in a smart hardware device. The method comprises: converting a voice signal received by a smart hardware device into text information by means of voice recognition technology (S110); extracting key words from the text information (S120); matching the extracted key words with a word list related to a smart hardware service, and selecting from the word list one or a plurality of candidate terms matching the key words (S130); and, on the basis of the selected one or plurality of candidate terms, performing error correction processing of the key words in the text information (S140). The present method takes the functional performance of the smart hardware into full consideration, and uses a preset service-related word list to perform smart error correction of key words parsed from voice recognition results, significantly increasing the accuracy of voice recognition and occupying few resources, and thereby meeting the requirements for low smart hardware device energy consumption.

Description

Speech recognition error correction method and device in intelligent hardware device

Technical field

The present invention relates to the field of voice recognition technology, and in particular, to a voice recognition error correction method and apparatus in an intelligent hardware device.

Background technique

The development of speech recognition technology has made it easier for users to interact with smart hardware devices such as smart watches, mobile phones, and driving recorders. Here are some examples of the interaction between the user and the intelligent hardware device using the voice recognition technology in the prior art:

1) Convert the user's instructions into words by voice recognition technology;

2) Understand user intent through semantic analysis techniques;

3) Convert the found text resources into voice through speech synthesis technology and feed back to the user.

Among them, the process of semantic analysis relies on the accuracy of speech recognition, and the accuracy of speech recognition is difficult to reach 100%. For example, when the user is a child, the unclearness of the speech makes various errors in speech recognition.

Summary of the invention

In view of the above problems, the present invention has been made in order to provide a speech recognition error correction method and apparatus in an intelligent hardware device that overcomes the above problems or at least partially solves the above problems.

According to an aspect of the present invention, a voice recognition error correction method in an intelligent hardware device is provided, including:

Transmitting the voice signal received by the intelligent hardware device into text information through voice recognition technology;

Extracting keywords from the text information;

Matching the extracted keywords with a vocabulary related to the intelligent hardware service, and selecting one or more candidate words matching the keyword from the vocabulary;

The error correction processing is performed on the keywords in the text information according to the selected one or more candidate words.

According to another aspect of the present invention, a speech recognition error correction apparatus in an intelligent hardware device is provided, including:

a voice recognition unit, configured to convert a voice signal received by the smart hardware device into text information by using a voice recognition technology;

a keyword extracting unit adapted to extract a keyword from the text information;

a matching unit, configured to match the extracted keyword with a vocabulary related to the intelligent hardware service, and select one or more candidate words matching the keyword from the vocabulary;

The error correction unit is adapted to perform error correction processing on the keywords in the text information according to the selected one or more candidate words.

According to an aspect of the invention there is provided a computer program comprising computer readable code, when said computer readable code is run on a computing device, causing said computing device to perform an intelligent hardware device as described above Speech recognition error correction method.

According to still another aspect of the present invention, a computer readable medium storing a computer program as described above is provided.

It can be seen from the above that the technical solution of the present invention first uses voice recognition technology to perform voice recognition on a voice signal received by an intelligent hardware device, converts it into text information, and further analyzes the text information to obtain a plurality of keywords, and these The keywords are matched by a vocabulary related to the intelligent hardware service, one or more candidate words are determined, and finally the keywords are corrected using the obtained candidate words. The technical solution fully considers the functional characteristics of the intelligent hardware, and uses the preset business-related vocabulary to intelligently correct the keywords parsed in the speech recognition result, which significantly improves the accuracy of the speech recognition and consumes less resources. In line with the low energy consumption requirements of intelligent hardware devices.

The above description is only an overview of the technical solutions of the present invention, and the above-described and other objects, features and advantages of the present invention can be more clearly understood. Specific embodiments of the invention are set forth below.

DRAWINGS

Various other advantages and benefits will become apparent to those skilled in the art from a The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting. Throughout the drawings, the same reference numerals are used to refer to the same parts. In the drawing:

1 is a flow chart showing a voice recognition error correction method in an intelligent hardware device according to an embodiment of the present invention;

2 is a schematic structural diagram of a speech recognition error correction apparatus in an intelligent hardware device according to an embodiment of the present invention;

Figure 3 schematically shows a block diagram of a computing device for performing the method according to the invention;

Fig. 4 schematically shows a storage unit for holding or carrying program code implementing the method according to the invention.

Specific embodiment

Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While the embodiments of the present invention have been shown in the drawings, the embodiments Rather, these embodiments are provided so that this disclosure will be more fully understood and the scope of the disclosure will be fully disclosed.

1 is a schematic flowchart of a voice recognition error correction method in an intelligent hardware device according to an embodiment of the present invention. As shown in FIG. 1, the method includes:

Step S110: Convert the voice signal received by the smart hardware device into text information by using a voice recognition technology.

Among them, the smart hardware device can be a smart phone, a smart watch, an intelligent robot, and the like. Speech recognition technology is a technology that is gradually being improved and put into use. For example, the Siri function in Apple's mobile phone can realize the voice recognition of the user's voice to perform functions such as opening the camera and map navigation in the Apple mobile phone.

In step S120, keywords are extracted from the text information.

For example, the user wants to use the storytelling function in the children's watch to send out the voice signal "I want to hear the story - Little Red Riding Hood", then the "story" is the keyword corresponding to the function, and "Little Red Riding Hood" is with this function. The corresponding keywords of the story classification. In other words, the keywords are related to the intelligent hardware business. So continue with the following steps:

Step S130, matching the extracted keywords with the vocabulary related to the intelligent hardware service, and selecting one or more candidate words matching the keywords from the vocabulary.

In actual use, the user's pronunciation may not be standard, which may cause the sound to be unsatisfactory. For example, if the child says "Little Red Riding Hood" as "Shaw Red Hat", then the speech recognition technology converts the voice signal. The text message may be "Shaw Red Hat." Obviously, the word does not match the "Little Red Riding Hood" completely, but the similarity between the two is very high. It can be concluded from the manual judgment that the child actually wants to express "Little Red Riding Hood". But in fact, the text message into which the speech signal is converted can blur the matching of the keywords. For example, there may be a story of "Little Red Cat" in the story library. Then for a keyword, the candidate words selected from the vocabulary may be one or more.

Step S140: Perform error correction processing on the keywords in the text information according to the selected one or more candidate words.

It can be seen that the method shown in FIG. 1 firstly uses the speech recognition technology to perform speech recognition on the speech signal received by the intelligent hardware device, converts it into text information, and further analyzes the text information to obtain a plurality of keywords, and these keys are The words are matched by a vocabulary related to the intelligent hardware service, one or more candidate words are determined, and finally the keywords are corrected using the obtained candidate words. The technical solution fully considers the functional characteristics of the intelligent hardware, and uses the preset business-related vocabulary to intelligently correct the keywords parsed in the speech recognition result, which significantly improves the accuracy of the speech recognition and consumes less resources. In line with the low energy consumption requirements of intelligent hardware devices.

In an embodiment of the present invention, the method shown in FIG. 1 further includes: presetting one or more fixed sentence patterns associated with the business voice interaction of the smart hardware device; marking in each fixed sentence pattern Position of the keyword; extracting the keyword from the text information includes: matching the text information with one or more fixed sentence patterns; extracting from the corresponding position of the text information according to the position of the keyword marked in the matched fixed sentence pattern Key words.

In the foregoing embodiment, "I want to listen to the story - Xiao Hong Cap" is a fixed sentence style, which can be summarized as "I want to hear the story XXX", wherein "XXX" corresponds to a keyword. For another example, the semantics of the sentence can also be expressed as: "I want to hear the story of Xiao Hong Cap", then the fixed sentence is "I want to hear the story of XXX." When the smart hardware supports song playback, the user can use the phrase "I want to listen to XXX's XXX", for example, "I want to listen to Andy Lau's forgotten water", and so on.

Obviously, by matching the text message of "I want to hear the story of Xiao Hong Cap" and the fixed sentence of "I want to hear the story of XXX", we can confirm that "Shaw Red Hat" is a keyword. Further, it can also be determined that "Shaw Red Hat" corresponds to the story name. Therefore, in an embodiment of the present invention, the method further includes: marking type information for each keyword in the fixed sentence; determining type information of the vocabulary related to the intelligent hardware service; and extracting the extracted keyword and the intelligent hardware The matching of the vocabulary related to the service includes: determining the type information of the extracted keyword according to the type information of the keyword in the matched fixed sentence pattern, and extracting the extracted keyword and type according to the type information of the extracted keyword. Match the vocabulary to match.

For the business function of “storytelling”, the vocabulary related to the business can be the name of the story; for the business function of “song playing”, the vocabulary related to the business can be the name of the song, the style of the song, the singer Name and so on. In the previous embodiment, the position of each keyword in the fixed sentence pattern can be determined. Then, by using the type information corresponding to the keyword of each position, it can be determined which service-related vocabulary should be used to match the candidate words. . For the "Red Riding Hood" in the previous embodiment, you can use the story noun table to For "Andy Lau", you can use the singer noun table to match; for "forget the water", you can use the song noun table to match.

After the keywords and corresponding candidate words are obtained, the true semantics of the keywords can be determined according to the candidate words. For example, the story that a child wants to hear is the story of "Little Red Cat" or the story of "Little Red Riding Hood". In an embodiment of the present invention, in the method shown in FIG. 1, the error correction processing on the keywords in the text information according to the selected one or more candidate words includes: matching the selected keywords with the keywords Each candidate word is scored according to the matching degree of the extracted keyword and the candidate word; if the score of the highest score candidate word of the keyword is higher than or equal to the first confidence value, the highest score candidate is used The word corrects the keyword; if the score of the highest score candidate of the keyword is higher than the second confidence value but lower than the first confidence value, a further voice dialogue is performed with the user to confirm whether the highest score candidate is needed The keyword is corrected; if the score of the highest score candidate of the keyword is lower than or equal to the second confidence value, no correction is made.

For example, the candidates for "Red Riding Hood" have the words "Little Red Cat" and "Little Red Riding Hood". "Red Riding Hood" and "Little Red Riding Hood" have the same two words, only one word is different; "Shaw Red Hat" has the same word as "Little Red Cat", and the other two words have different voices. Then the score result may be: "Little Red Riding Hood" has a score of 0.6, and "Little Red Cat" has a score of 0.5. Then, taking the first confidence value of 0.45 as an example, since the scores of the two candidate words are all higher than 0.45, then the candidate word with the highest score, that is, "Little Red Riding Hood" is selected to correct the keyword "Red Hood".

An example of a strategy for scoring candidate term words is given below. In an embodiment of the present invention, the scoring of the candidate words according to the matching degree between the extracted keywords and the candidate words includes: From high to low, it is divided into three ranges of high, medium and low; if the keyword is the same as the pinyin of the candidate, but the pitch is different, it is scored in the high range; if the keyword is in the pinyin of the candidate If the initial or final part is the same, the score is scored in the middle range; if the key is not the same as the initial and final in the pinyin of the candidate, the score is scored in the low range.

For example, the high-end score range is [0.45, 1], the mid-range score is (0.4, 0.45), and the low-range score is [0, 0.4). Since "Little Red Cat" and "Little Red Riding Hood" and "Red Hood" belong to "the same pinyin, but the tone is different", then the high-grade scoring standard is adopted. Let's look at another example: the pronunciation of the initials of some users "n" and "l" is not divided, resulting in the "beef" being said to be "flowing meat", then the keyword "beef" and the word "flowing meat" The partial initials are different, the finals are the same, then the mid-range scoring standard is adopted. Let's look at another example: the user recorded "Andy Lau" as "Zhang Dehua" due to memory deviation, then the initials and finals of "Liu" and "Zhang" are different, but the other two words are the same. In this case, Use low-grade scoring standards.

Then, after the score of the candidate word is obtained, the confidence of the candidate word of the highest score can be determined according to the method in the previous embodiment. For example, if the score for "beef" is 0.44, which is lower than the first confidence value of 0.45 but higher than the second confidence value of 0.4, then you can ask the user: "Do you want to say 'beef'?" to confirm if you need to use "Beef" corrects "flowing meat." If the score of the highest score candidate of the keyword is lower than or equal to the second confidence value, no correction is made, because even if correction is made at this time, the user's original intention may be deviated.

In an embodiment of the present invention, the method further includes: if a plurality of keywords are extracted from the text information, multiplying the scores of the highest score candidates of each keyword to obtain a score of the plurality of keywords; If the scores of the plurality of keywords are higher than or equal to the third confidence value, correcting each keyword with the highest score candidate of each keyword; if the score of the plurality of keywords is higher than the fourth confidence value but lower than the first The three-confidence value is further voiced with the user to confirm whether the highest score candidate of each keyword is needed to correct each keyword; if the score of the plurality of keywords is lower than or equal to the fourth confidence value, no correct.

In the present embodiment, an example of how to calculate a candidate word score when a keyword is plural is given. For example, the score of "Zhang Dehua" is 0.3, the pronunciation of the user who "forgets the water" is very standard, and the score of the corresponding candidate for "forgetting water" is 1, then the scores of these two keywords are 0.3 × 1 = 0.3.

In an embodiment of the present invention, the method further includes: outputting, according to the result of the correction processing, a corresponding service service of the smart hardware device.

For example, without performing the technical solution of the present invention, the child said to the intelligent accompanying robot: "Do you tell the story of Xiao Hong Cap?", because "Shaw Red Hat" does not match "Little Red Riding Hood", the child gets The reply is "I will not do this." At this time, the mother re-directed the smart accompanying robot to "tell the story of Little Red Riding Hood", and the intelligent accompanying robot correctly obtained the resources of the story "Little Red Riding Hood" for voicetelling. .

In the present embodiment, since the child said "Will you tell the story of Xiaohongxi?", it is correct to correct "Shaw Red Hat" to "Little Red Riding Hood" by correcting the voice recognition. Get the resources of the story "Little Red Riding Hood" and tell the story.

FIG. 2 is a schematic structural diagram of a voice recognition error correction apparatus in an intelligent hardware device according to an embodiment of the present invention. As shown in FIG. 2, the voice recognition error correction apparatus 200 in the smart hardware device includes:

The voice recognition unit 210 is adapted to convert the voice signal received by the smart hardware device into text information by using a voice recognition technology.

The keyword extracting unit 220 is adapted to extract keywords from the text information.

The matching unit 230 is adapted to match the extracted keywords with the vocabulary related to the intelligent hardware service, and select one or more candidate words that match the keywords from the vocabulary.

The error correction unit 240 is adapted to perform error correction processing on the keywords in the text information according to the selected one or more candidate words.

It can be seen that the device shown in FIG. 2 firstly uses the voice recognition technology to perform voice recognition on the voice signal received by the intelligent hardware device by using the mutual recognition of each unit, converts it into text information, and further analyzes the text information to obtain some of them. The keywords are matched by the vocabulary related to the intelligent hardware service to determine one or more candidate words, and finally the keywords are corrected by using the obtained candidate words. The technical solution fully considers the functional characteristics of the intelligent hardware, and uses the preset business-related vocabulary to intelligently correct the keywords parsed in the speech recognition result, which significantly improves the accuracy of the speech recognition and consumes less resources. In line with the low energy consumption requirements of intelligent hardware devices.

In an embodiment of the present invention, the apparatus further includes: a configuration unit, configured to preset one or more fixed sentence patterns associated with the business voice interaction of the smart hardware device; in each fixed sentence pattern Marking the location of the keyword; the keyword extracting unit 220 is adapted to match the text information with one or more fixed sentence patterns; and extracting from the corresponding position of the text information according to the position of the keyword marked in the matched fixed sentence pattern Key words.

In an embodiment of the present invention, in the foregoing apparatus, the configuration unit is further adapted to mark type information for the keywords in each fixed sentence; determine type information of the vocabulary related to the intelligent hardware service; and the matching unit 230 The type information of the extracted keyword is determined according to the type information of the keyword in the matched fixed sentence pattern, and the extracted keyword is matched with the vocabulary matching the type according to the type information of the extracted keyword.

In an embodiment of the present invention, in the foregoing apparatus, the error correcting unit 240 is adapted to select, for each selected candidate word that matches the keyword, according to the matching degree of the extracted keyword and the candidate word, the candidate Word scoring; if the score of the highest score candidate of the keyword is higher than or equal to the first confidence value, correct the keyword with the highest score candidate; if the score of the highest score candidate of the keyword is higher than the second confidence If the value is lower than the first confidence value, a further voice dialogue is performed with the user to confirm whether the keyword needs to be corrected with the highest score candidate; if the score of the highest score candidate of the keyword is lower than or equal to the second confidence Degree value, no correction.

In an embodiment of the present invention, in the foregoing apparatus, the error correction unit 240 is adapted to divide the score from high to low into three ranges of high, medium, and low;

If the keyword is the same as the pinyin of the candidate word, but the pitch is different, the score is scored in the high-end range;

If the keyword is the same as the initial or final part of the pinyin of the candidate word, it is scored in the mid-range range; if the keyword is not the same as the initial or final in the pinyin of the candidate word, it is in the low range hit Minute.

In an embodiment of the present invention, in the foregoing apparatus, the error correction unit 240 is further adapted to multiply the scores of the highest score candidate words of each keyword when the plurality of keywords are extracted from the text information, to obtain the a score of a plurality of keywords; if the scores of the plurality of keywords are higher than or equal to the third confidence value, correcting each keyword with the highest score candidate of each keyword; if the score of the plurality of keywords is higher than the fourth If the confidence value is lower than the third confidence value, a further voice dialogue is performed with the user to confirm whether the highest score candidate of each keyword is needed to correct each keyword; if the scores of the multiple keywords are lower than or equal to the first Four confidence values are not corrected.

In an embodiment of the present invention, the apparatus further includes: a service service unit adapted to output a corresponding service service of the smart hardware device according to the result of the correction process.

It should be noted that the specific implementation manners of the foregoing device embodiments are the same as the specific implementation manners of the foregoing corresponding method embodiments, and details are not described herein again.

In summary, the technical solution of the present invention firstly uses voice recognition technology to perform voice recognition on a voice signal received by an intelligent hardware device, converts it into text information, and further analyzes the text information to obtain a plurality of keywords thereof. These keywords are matched by a vocabulary related to the intelligent hardware service to determine one or more candidate words, and finally the keywords are corrected using the obtained candidate words. The technical solution fully considers the functional characteristics of the intelligent hardware, and uses the preset business-related vocabulary to intelligently correct the keywords parsed in the speech recognition result, which significantly improves the accuracy of the speech recognition and consumes less resources. In line with the low energy consumption requirements of intelligent hardware devices.

It should be noted:

The algorithms and displays provided herein are not inherently related to any particular computer, virtual device, or other device. Various general purpose devices can also be used with the teaching based on the teachings herein. The structure required to construct such a device is apparent from the above description. Moreover, the invention is not directed to any particular programming language. It is to be understood that the invention may be embodied in a variety of programming language, and the description of the specific language has been described above in order to disclose the preferred embodiments of the invention.

In the description provided herein, numerous specific details are set forth. However, it is understood that the embodiments of the invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of the description.

Similarly, the various features of the invention are sometimes grouped together into a single embodiment, in the above description of the exemplary embodiments of the invention, Figure, or a description of it. However, the method disclosed should not be interpreted as reflecting It is intended that the claimed invention be characterized by more features than those described in the appended claims. Rather, as the following claims reflect, inventive aspects reside in less than all features of the single embodiments disclosed herein. Therefore, the claims following the specific embodiments are hereby explicitly incorporated into the embodiments, and each of the claims as a separate embodiment of the invention.

Those skilled in the art will appreciate that the modules in the devices of the embodiments can be adaptively changed and placed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and further they may be divided into a plurality of sub-modules or sub-units or sub-components. In addition to such features and/or at least some of the processes or units being mutually exclusive, any combination of the features disclosed in the specification, including the accompanying claims, the abstract and the drawings, and any methods so disclosed, or All processes or units of the device are combined. Each feature disclosed in this specification (including the accompanying claims, the abstract and the drawings) may be replaced by alternative features that provide the same, equivalent or similar purpose.

In addition, those skilled in the art will appreciate that, although some embodiments described herein include certain features that are included in other embodiments and not in other features, combinations of features of different embodiments are intended to be within the scope of the present invention. Different embodiments are formed and formed. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

The various component embodiments of the present invention may be implemented in hardware, or in a software module running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used in practice to implement some of some or all of the components of the speech recognition error correction device in an intelligent hardware device in accordance with an embodiment of the present invention. Or all features. The invention can also be implemented as a device or device program (e.g., a computer program and a computer program product) for performing some or all of the methods described herein. Such a program implementing the invention may be stored on a computer readable medium or may be in the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

For example, Figure 3 shows a block diagram of a computing device for performing the method in accordance with the present invention. The computing device conventionally includes a processor 310 and a computer program product or computer readable medium in the form of a memory 320. The memory 320 may be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), an EPROM, a hard disk, or a ROM. The memory 320 has a storage space 330 that stores program code 331 for performing any of the method steps described above. For example, the storage space 330 for storing program code may separately store respective program codes 331 for implementing various steps in the above method. The program code can be read from or written to one or more computer program products to the one or more computer programs In the product. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards or floppy disks. Such computer program products are typically portable or fixed storage units such as those shown in FIG. The storage unit may have storage segments, storage spaces, and the like that are similarly arranged to memory 320 in the computing device of FIG. The program code can be compressed in an appropriate form. In general, the storage unit stores computer readable program code 331' for performing the steps of the method according to the invention, ie program code readable by a processor such as 310, when the program code is run by the computing device, resulting in The computing device performs the various steps in the methods described above.

It is to be noted that the above-described embodiments are illustrative of the invention and are not intended to be limiting, and that the invention may be devised without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as a limitation. The word "comprising" does not exclude the presence of the elements or steps that are not recited in the claims. The word "a" or "an" The invention can be implemented by means of hardware comprising several distinct elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means can be embodied by the same hardware item. The use of the words first, second, and third does not indicate any order. These words can be interpreted as names.

In addition, it should be noted that the language used in the specification has been selected for the purpose of readability and teaching, and is not intended to be construed or limited. Therefore, many modifications and changes will be apparent to those skilled in the art without departing from the scope of the invention. The disclosure of the present invention is intended to be illustrative, and not restrictive, and the scope of the invention is defined by the appended claims.

Claims

A speech recognition error correction method in an intelligent hardware device, wherein the method comprises:

Transmitting the voice signal received by the intelligent hardware device into text information through voice recognition technology;

Extracting keywords from the text information;

Matching the extracted keywords with a vocabulary related to the intelligent hardware service, and selecting one or more candidate words matching the keyword from the vocabulary;

The error correction processing is performed on the keywords in the text information according to the selected one or more candidate words.
The method of claim 1 wherein

The method further includes: presetting one or more fixed sentence patterns associated with the business voice interaction of the smart hardware device; marking a location of the keyword in each fixed sentence pattern;

The extracting the keyword from the text information comprises: matching the text information with the one or more fixed sentence patterns; and from the position of the keywords marked in the matched fixed sentence pattern, from the text The corresponding position of the information is extracted from the keyword.
The method of claim 1 or 2, wherein

The method further includes: marking type information for each keyword in the fixed sentence; determining type information of the vocabulary related to the intelligent hardware service;

The matching the extracted keywords with the vocabulary related to the intelligent hardware service comprises: determining the type information of the extracted keywords according to the type information of the keywords in the matched fixed sentence patterns, according to the extracted keywords The type information matches the extracted keywords with a vocabulary that matches the type.
The method according to any one of claims 1 to 3, wherein the error correction processing of the keywords in the text information according to the selected one or more candidate words comprises:

For each candidate phrase selected to match the keyword, the candidate word is scored according to the matching degree of the extracted keyword and the candidate word;

If the score of the highest score candidate of the keyword is higher than or equal to the first confidence value, correct the keyword with the highest score candidate;

If the score of the highest score candidate of the keyword is higher than the second confidence value but lower than the first confidence value, a further voice dialogue is performed with the user to confirm whether the highest score candidate is needed to correct the Key words;

If the score of the highest score candidate of the keyword is lower than or equal to the second confidence value, no correction is performed positive.
The method according to any one of claims 1 to 4, wherein the scoring the candidate word according to the matching degree of the extracted keyword and the candidate word comprises:

Divide the score from high to low into three ranges of high, medium and low;

If the keyword is the same as the pinyin of the candidate word, but the pitch is different, the score is scored in the high-end range;

If the keyword is the same as the initial or final part of the pinyin of the candidate word, the score is scored in the mid-range range;

If the keyword is not the same as the initial or final in the pinyin of the candidate, the score is scored in the low range.
The method of any of claims 1-5, wherein the method further comprises:

If a plurality of keywords are extracted from the text information, the scores of the highest score candidates of each keyword are multiplied to obtain scores of the plurality of keywords;

If the scores of the plurality of keywords are higher than or equal to the third confidence value, correcting the keywords by using the highest score candidate words of each keyword;

If the score of the plurality of keywords is higher than the fourth confidence value but lower than the third confidence value, a further voice dialogue is performed with the user to confirm whether the highest score candidate of each keyword is needed to correct the keywords ;

If the score of multiple keywords is lower than or equal to the fourth confidence value, no correction is made.
The method of any of claims 1-6, wherein the method further comprises:

Outputting corresponding service services of the smart hardware device according to the result of the correction processing.
A speech recognition error correction device in an intelligent hardware device, wherein the device comprises:

a voice recognition unit, configured to convert a voice signal received by the smart hardware device into text information by using a voice recognition technology;

a keyword extracting unit adapted to extract a keyword from the text information;

a matching unit, configured to match the extracted keyword with a vocabulary related to the intelligent hardware service, and select one or more candidate words matching the keyword from the vocabulary;

The error correction unit is adapted to perform error correction processing on the keywords in the text information according to the selected one or more candidate words.
The apparatus of claim 8, wherein the apparatus further comprises: a configuration unit adapted to pre-set one or more one or more solids associated with the business voice of the smart hardware device Fixed sentence; the location of the keyword in each fixed sentence;

The keyword extracting unit is adapted to match the text information with the one or more fixed sentence patterns; according to the position of the keyword marked in the matched fixed sentence pattern, from the corresponding position of the text information Extract keywords.
The apparatus according to claim 8 or 9, wherein

The configuration unit is further adapted to mark type information for each keyword in the fixed sentence; and determine type information of the vocabulary related to the intelligent hardware service;

The matching unit is adapted to determine type information of the extracted keyword according to type information of the keyword in the matched fixed sentence pattern, and match the extracted keyword with the type according to the type information of the extracted keyword. The vocabulary is matched.
A device according to any one of claims 8 to 10, wherein

The error correcting unit is adapted to score the candidate word according to the matching degree of the extracted keyword and the candidate word for each selected candidate word matching the keyword; if the keyword is the highest If the score of the candidate candidate is higher than or equal to the first confidence value, the keyword is corrected by the highest score candidate; if the score of the highest score candidate of the keyword is higher than the second confidence value but lower than a first confidence value, and then a further voice conversation with the user to confirm whether the keyword needs to be corrected with the highest score candidate; if the score of the highest score candidate of the keyword is lower than or equal to the second confidence Degree value, no correction.
A device according to any of claims 8-11, wherein

The error correction unit is adapted to divide the score from high to low into three ranges of high, medium and low;

If the keyword is the same as the pinyin of the candidate word, but the pitch is different, the score is scored in the high-end range; if the keyword is the same as the initial or final part in the pinyin of the candidate, the score is scored in the mid-range range; If the initials and the finals in the pinyin of the candidate are not the same, the scores are scored in the low range.
A device according to any one of claims 8 to 12, wherein

The error correction unit is further adapted to: when the plurality of keywords are extracted from the text information, multiply the scores of the highest score candidates of each keyword to obtain scores of the plurality of keywords; if multiple If the score of the keyword is higher than or equal to the third confidence value, the keywords are corrected by the highest score candidate of each keyword; if the score of the multiple keywords is higher than the fourth confidence value but lower than the third Confidence value, then a further voice dialogue with the user to confirm whether the highest score candidate for each keyword is needed to correct the key Word; if the score of multiple keywords is lower than or equal to the fourth confidence value, no correction is made.
The apparatus of any of claims 8-13, wherein the apparatus further comprises:

The service service unit is adapted to output a corresponding service service of the intelligent hardware device according to the result of the correction process.
A computer program comprising computer readable code, when the computer readable code is run on a computing device, causing the computing device to perform in an intelligent hardware device according to any one of claims 1-7 Speech recognition error correction method.
A computer readable medium storing the computer program of claim 15.