CN109522392A - Voice-based search method, server and computer readable storage medium - Google Patents
Voice-based search method, server and computer readable storage medium Download PDFInfo
- Publication number
- CN109522392A CN109522392A CN201811182073.1A CN201811182073A CN109522392A CN 109522392 A CN109522392 A CN 109522392A CN 201811182073 A CN201811182073 A CN 201811182073A CN 109522392 A CN109522392 A CN 109522392A
- Authority
- CN
- China
- Prior art keywords
- text
- target
- keyword
- vocabulary
- voice
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012545 processing Methods 0.000 claims abstract description 56
- 238000000605 extraction Methods 0.000 claims abstract description 32
- 230000011218 segmentation Effects 0.000 claims abstract description 23
- 238000005194 fractionation Methods 0.000 claims abstract description 13
- 238000004590 computer program Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 239000000284 extract Substances 0.000 claims description 5
- 238000005070 sampling Methods 0.000 claims description 4
- 238000005516 engineering process Methods 0.000 abstract description 3
- 238000012549 training Methods 0.000 description 21
- 238000012360 testing method Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The present invention is suitable for field of computer technology, provides a kind of voice-based search method, server and computer readable storage medium, wherein method includes: acquisition retrieval request, and the retrieval request includes the voice messaging for being retrieved;Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;Word segmentation processing is carried out to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and extraction is carried out to the fractionation phrase based on preset stop words dictionary and handles to obtain keyword, and the vocabulary that the stop words dictionary is filtered by needs forms.The lists of keywords of the corresponding keyword of voice messaging and text each in text library match more accurately, when there is the target keyword to match with keyword in some lists of keywords, resulting text corresponding to the lists of keywords is exported as search result, so that the accuracy of search result is higher.
Description
Technical field
The invention belongs to field of computer technology more particularly to a kind of voice-based search methods, server and calculating
Machine readable storage medium storing program for executing.
Background technique
With the development of technology, in order to enable the search operaqtion of user is more convenient, voice can be passed through by realizing user
Mode input the method retrieved of retrieval information, server retrieve to the text in database and be obtained by retrieval information
Search result, and search result is ranked up according to relevance size, to generate retrieval list, be supplied to user into
Row is checked.When retrieving for inputting retrieval information by way of voice, server needs to be analyzed to obtain to voice
Then corresponding text sentence is retrieved text sentence directly as retrieval information in the database of server, however
How this mode preferably carries out matched ask with the text data in database due to not fully taking into account text sentence
Topic, it is not accurate enough so as to cause the result of retrieval.
Summary of the invention
It can in view of this, the embodiment of the invention provides a kind of voice-based search method, server and computers
Read storage medium, by solve in the prior art user inputted in a manner of voice retrieval information retrieve when, the result of retrieval
Not accurate enough problem.
The first aspect of the embodiment of the present invention provides a kind of voice-based search method, comprising:
Retrieval request is obtained, the retrieval request includes the voice messaging for being retrieved;
Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;
It carries out word segmentation processing to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and based on preset
Stop words dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary needs are filtered
Vocabulary is formed;
Processing is extracted to all texts in the text library prestored, obtains the corresponding lists of keywords of each text;
If existing and the keyword phase in lists of keywords corresponding to each text in the text library prestored
The target keyword matched then exports resulting text corresponding to the lists of keywords as search result.
The second aspect of the embodiment of the present invention provides a kind of server, including memory, processor and is stored in institute
The computer program that can be run in memory and on the processor is stated, the processor executes real when the computer program
Existing following steps:
Retrieval request is obtained, the retrieval request includes the voice messaging for being retrieved;
Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;
It carries out word segmentation processing to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and based on preset
Stop words dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary needs are filtered
Vocabulary is formed;
Processing is extracted to all texts in the text library prestored, obtains the corresponding lists of keywords of each text;
If existing and the keyword phase in lists of keywords corresponding to each text in the text library prestored
The target keyword matched then exports resulting text corresponding to the lists of keywords as search result.
The third aspect of the embodiment of the present invention provides a kind of computer readable storage medium, the computer-readable storage
Media storage has computer program, and the computer program performs the steps of when being executed by processor
Retrieval request is obtained, the retrieval request includes the voice messaging for being retrieved;
Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;
It carries out word segmentation processing to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and based on preset
Stop words dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary needs are filtered
Vocabulary is formed;
Processing is extracted to all texts in the text library prestored, obtains the corresponding lists of keywords of each text;
If existing and the keyword phase in lists of keywords corresponding to each text in the text library prestored
The target keyword matched then exports resulting text corresponding to the lists of keywords as search result.
A kind of voice-based search method, server and computer readable storage medium tool provided in an embodiment of the present invention
Have it is following the utility model has the advantages that
The embodiment of the present invention obtains retrieval request, and the retrieval request includes the voice messaging for being retrieved;To institute
The voice messaging in retrieval request is stated to be handled to obtain the text sentence for corresponding to the voice messaging.Based on preset participle word
Allusion quotation carries out word segmentation processing to the text sentence and obtains splitting phrase, and based on preset stop words dictionary to the fractionation phrase
It carries out extraction to handle to obtain keyword, to remove some unwanted part vocabulary in the corresponding text sentence of voice messaging;
Processing is extracted to all texts in the text library prestored, obtains arranging with the higher keyword of each text relevant
Table, so that the lists of keywords of each text can be carried out more accurately in the corresponding keyword of voice messaging and text library
Matching;Only when there is the target keyword to match with keyword in some lists of keywords, keyword column are just exported
Resulting text corresponding to table is as search result, so that the accuracy of search result is higher.
Detailed description of the invention
It to describe the technical solutions in the embodiments of the present invention more clearly, below will be to embodiment or description of the prior art
Needed in attached drawing be briefly described, it should be apparent that, the accompanying drawings in the following description is only of the invention some
Embodiment for those of ordinary skill in the art without any creative labor, can also be according to these
Attached drawing obtains other attached drawings.
Fig. 1 is a kind of implementation flow chart for voice-based search method that first embodiment of the invention provides;
Fig. 2 is a kind of implementation process of the step 102 for voice-based search method that first embodiment of the invention provides
Figure;
Fig. 3 is a kind of implementation flow chart for voice-based search method that second embodiment of the invention provides;
Fig. 4 is a kind of schematic diagram for server that third embodiment of the invention provides;
Fig. 5 is a kind of schematic diagram for server that fourth embodiment of the invention provides.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Refering to fig. 1, Fig. 1 is the implementation flow chart of the voice-based search method in first embodiment of the invention.This reality
The executing subject for applying the voice-based search method in example is server.Voice-based search method as described in Figure can be with
The following steps are included:
S101 obtains retrieval request, and the retrieval request includes the voice messaging for being retrieved.
In S101, the server in the present embodiment includes text library, is stored in text library and is largely used to be supplied to
The text information of user's progress retrieval and inquisition.Each text information specifically includes the hair of content of text, the title of text, text
The information such as the issuing time of cloth Internet channel and text.When user needs to obtain some text in text library or certain texts
When relevant information, can include in voice messaging by inputting the retrieval request comprising voice messaging in the terminal to server
User wants some important informations in the text retrieved, and server obtains the retrieval request that user is inputted by terminal.
S102 handles the voice messaging in the retrieval request to obtain the text language for corresponding to the voice messaging
Sentence.
In S102, after server receives retrieval request, need to carry out voice knowledge to the voice messaging in retrieval request
Other places reason, obtains the text sentence for corresponding to the voice messaging, in order to which server carries out retrieval analysis according to text sentence.
Further, referring to Fig.2, Fig. 2 is a kind of voice-based search method that first embodiment of the invention provides
The implementation flow chart of step 102.The step 102 specifically comprises the following steps:
S1021 carries out feature extraction to the voice messaging in the retrieval request and obtains speech feature vector.
In S1021, for the retrieval request comprising voice messaging of acquisition, server believes the voice in retrieval request
Breath carries out the text sentence that voice recognition processing obtains corresponding to the voice messaging, wherein carrying out voice recognition processing to voice messaging
When, it needs first to carry out feature extraction to voice messaging to obtain the corresponding speech feature vector of the voice messaging.
Further, server carries out feature extraction to voice messaging and obtains the corresponding speech feature vector of the voice messaging
When, it needs to carry out sampling processing to the voice messaging in retrieval request according to predeterminated frequency to obtain sampled speech signal, in order to
Subsequent pretreatment is carried out to voice signal, wherein predeterminated frequency can be set to 8000hz or 16000hz.
The sampled speech signal obtained for carrying out sampling processing, in order to improve the high fdrequency component in sampled speech signal
Resolution capability, server also carries out preemphasis to sampled speech signal and handles to obtain the first voice signal, by sampled speech
Signal carry out preemphasis processing, server, which realizes, compensates the high fdrequency component of sampled speech signal, in order to continue into
Row subsequent processing.Wherein the method that preemphasis handles to obtain the first voice signal is carried out to sampled speech signal to be specifically as follows:
It is H (z)=1- α z by transmission function-1Single order FIR high-pass digital filter realize preemphasis, wherein in transmission function
α is pre emphasis factor, and the condition that pre emphasis factor α meets is 0.9 < α < 1.
The first voice signal handled for carrying out preemphasis, server also carry out at adding window the first voice signal
Reason obtains the second voice signal, and the second voice signal obtained after windowing process has short-term stationarity.In addition, in this reality
It applies in example, the voice frame length of the second voice signal obtained after windowing process is generally 10-30 milliseconds.
The second voice signal is obtained for carrying out windowing process, server also needs to carry out endpoint inspection to the second voice signal
Survey handles to obtain targeted voice signal, filters out mute part and noise section in the second voice signal, to will filter out quiet
Voice signal after line point and noise section is as targeted voice signal.And server carries out endpoint to the second voice signal
The purpose of detection processing is, can both reduce calculation amount when carrying out feature extraction to voice signal, save the time, excluding
After the interference of noise, accuracy rate when carrying out feature extraction to voice signal can also be enhanced.To the second voice in the present embodiment
Signal carries out the method that endpoint detection processing obtains targeted voice signal and is specifically as follows using according to the characteristic parameter in time domain
Come the method distinguished to mute part and noise section, can also use according to the characteristic parameter in frequency domain come to mute
The method that part and noise section distinguish, or according to the characteristic parameter in the characteristic parameter and frequency domain in time domain come to quiet
The method that line point and noise section distinguish, is of course not solely limited to this.
Targeted voice signal is obtained for carrying out endpoint detection processing, server also needs to carry out targeted voice signal special
Sign is extracted and obtains speech feature vector, has wherein carrying out the method that feature extraction obtains speech feature vector to targeted voice signal
Body can be for targeted voice signal progress MFCC (mel-frequency cepstrum coefficient) feature extraction.
Wherein during to targeted voice signal MFCC feature extraction, server needs to first pass through fast Fourier change
It changes and targeted voice signal is handled, to realize the power that targeted voice signal is changed into frequency-region signal from time-domain signal
Spectrum;Server is simultaneously filtered the power spectrum for obtaining frequency-region signal by Mel filter group, wherein being filtered
Number of filter be set as M, generally between 20 to 28.The centre frequency of the triangular filter of Mel filter group is set as f
(m), wherein m is 1,2 .., M, and the interval between each center frequency f (m) is broadening with the increase of m value, with m value
Reduction and reduce so that the filter effect of Mel filter group is more preferable.Server also needs to own Mel filter group
The output of filter successively takes logarithm to obtain logarithmic energy spectrum, and right to the M of the output of all filters of Mel filter group
Number energy spectrum executes discrete cosine transform, and to remove the similitude between each dimensional signal, each dimensional signal is reflected in realization
It is mapped to the characteristic parameter that lower dimensional space obtains the p order component of MFCC feature.Wherein the calculation formula of the p order component of MFCC feature isWherein P is the dimension of MFCC feature, and C (p) is indicated
The pth order component of MFCC feature.By the first-order difference parameter of the characteristic parameter of the p order component of MFCC feature and MFCC feature into
Row combination, is determined as speech feature vector corresponding to targeted voice signal.
S1022, by the speech feature vector be input to preset acoustic model carry out processing output obtain character.
In S1022, the speech feature vector obtained for carrying out feature extraction, server inputs speech feature vector
Processing output is carried out to preset acoustic model and obtains corresponding character, and wherein the character is specially single word, i.e. acoustic mode
Type input is the multiple and different words obtained after handling speech feature vector, wherein above-mentioned acoustic model is by instruction
It gets, is stored in server.
It is the sample data obtained for being trained to acoustic model to the method that above-mentioned acoustic model is trained, and
Sample data is divided into training set and test set;Wherein, every sample data in sample data includes speech feature vector
And the corresponding character of speech feature vector;By the speech feature vector and voice that include in every sample data in training set spy
Input of the corresponding character of vector as acoustics training pattern is levied, using the corresponding character of speech feature vector as acoustics training mould
The output of type is trained acoustics training pattern.
For the acoustics training pattern after training, server is needed to also it being needed to test.The method verified
It is specifically as follows and the speech feature vector that every sample data in test set includes is input to the acoustics training mould after training
Type, the character predicted;By the way that verification is compared in the character of prediction character corresponding with speech feature vector;Work as prediction
The percent similarity of character character corresponding with speech feature vector when reaching preset value, then illustrate that verification passes through, at this time then
Acoustics training pattern can be identified as acoustic model.
The character input to preset language model is carried out processing output and obtains text sentence by S1023.
In S1023, processing output is carried out for preset acoustic model and obtains character, server is input characters into pre-
If language model carry out processing output obtain text sentence.Wherein text sentence is text language composed by character information
Sentence, wherein above-mentioned language model is to obtain by training, is stored in database server.
It is to obtain sample data for being trained to language model to the method that language model is trained, and by sample
Notebook data is divided into training set and test set;Wherein, every sample data in sample data includes all words of text sentence
Text sentence corresponding to symbol and all characters;By all words for the text sentence for including in every sample data in training set
Input of the text sentence as speech training model corresponding to symbol and all characters, by text sentence corresponding to all characters
As the output of speech training model, speech training model is trained.
For the speech training model after training, server is needed to also it being needed to test.The method verified
Language after being specifically as follows all character inputs to training for the text sentence for including by every sample data in test set
Training pattern obtains prediction text sentence;By the way that prediction text sentence is compared with text sentence corresponding to all characters
To verification;When the percent similarity for predicting text sentence corresponding to text sentence and all characters reaches preset value, then say
Bright verification passes through, and then speech training model can be identified as language model at this time.
S103 carries out word segmentation processing to the text sentence based on preset dictionary for word segmentation and obtains splitting phrase, and is based on
Preset stop words dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary is to need to carry out
The vocabulary of filtering is formed.
In S103, for the text sentence obtained by processing, server is according to preset dictionary for word segmentation to the text
Sentence carries out word segmentation processing and obtains splitting phrase, wherein preset dictionary for word segmentation is based on the content of text, text in text library
Title, text the information such as delivery network channel as reference, the obtained dictionary comprising a large amount of vocabulary, being used for can be accurately
Word segmentation processing is carried out to a text sentence and obtains multiple fractionation phrases, fractionation phrase is extracted convenient for server and is used
In the keyword retrieved.For obtained fractionation phrase, server be based on preset stop words dictionary to split phrase into
Row extraction handles to obtain keyword, and stop words dictionary is formed by the vocabulary in need being filtered, wherein needing to carry out
The vocabulary of filter may include all modal particle and auxiliary word, can also include some sensitive vocabulary certainly.
S104 extracts processing to all texts in the text library prestored, obtains the corresponding keyword column of each text
Table.
For the text library in server, server extracts processing to all texts in the text library prestored, obtains
The corresponding lists of keywords of each text, which includes the important vocabulary of each text, extraction is obtained every
The lists of keywords of a text, server by lists of keywords keyword corresponding with the voice messaging in above-mentioned retrieval request into
Row matching so that the lists of keywords of the corresponding keyword of voice messaging and text each in text library can be carried out it is more smart
It really matches, so that the accuracy of search result is higher.
Further, described 104, comprising:
According to the ratio of the target vocabulary of target text in the text library and all vocabulary of the target text, and
First weight of the target vocabulary determines and obtains the first score value of the target vocabulary, and the target text is the text
Any one text in this library, the target vocabulary are any one vocabulary in the target text;According to the text
There are the ratio of the amount of text of the target vocabulary and the target words in all amount of text and the text library in library
The second weight converged, determination obtain the second score value of the target vocabulary;According to first score value and the second score value
Product determination obtain the keyword score of the target vocabulary;If the keyword score is greater than preset value, by the mesh
Mark vocabulary is added in the corresponding lists of keywords of the target text.
Specifically, for any one target text in text library, the target any one vocabulary herein all may be used
Using as target vocabulary.All words of number and target text that server is realized by each target vocabulary in target herein
First weight of the ratio of remittance and the target vocabulary, determination obtains the first score value of the target vocabulary, for frequency of occurrence
More vocabulary, corresponding first score value are bigger.It should be noted that the weight size of each target vocabulary not phase
Together, for example, some modal particles " " etc. vocabulary weight be 0, such as text be some technical field technical literature, and for
Some technical words in the field, corresponding weight are then larger.
Server is also according to there are the ratios of the amount of text of target vocabulary in amount of text all in text library and text library
Second weight of value and the target vocabulary, determination obtains the second score value of the target vocabulary, when occurring target in text library
The amount of text of other texts of vocabulary is smaller, and corresponding second score value is bigger.
For each target vocabulary, server is commented according to the first score value of the target vocabulary and the second of the target vocabulary
The product determination of score value obtains the keyword score of the target vocabulary, to obtain the keyword score of the target vocabulary, keyword
It scores bigger, illustrates that the target vocabulary is more important in target text.
For all target vocabularies in target text, server selects the vocabulary that keyword score is greater than preset value
Come, is added in the corresponding lists of keywords of the target text, to obtain the lists of keywords of the target text.
S105, if existing and the keyword in lists of keywords corresponding to each text in the text library prestored
The target keyword to match then exports resulting text corresponding to the lists of keywords as search result.
Exist in S105, in the lists of keywords corresponding to the text each in the text library prestored and keyword phase
The target keyword matched then exports resulting text corresponding to the lists of keywords as search result, only when some key
When there is the target keyword to match with keyword in word list, the work of resulting text corresponding to the lists of keywords is just exported
For search result, so that the accuracy of search result is higher.
Further, the step of resulting text corresponding to the output lists of keywords is as search result, packet
It includes:
The resulting text is ranked up according to the number of matched target keyword to obtain sequence text list, it is described
Resulting text includes two or more texts;Export the sequence text list.
There are when multiple texts in the resulting text exported for needs, server can be according to matched target keyword
Number is ranked up resulting text to obtain sequence text list, wherein the rule to sort is that matched target keyword number is got over
More text alignments is preceding, and the fewer text alignment of matched target keyword number is rear.It should be noted that for matching
Identical two texts of target keyword number, two texts can be ranked up according to issuing time, issuing time
Shorter text alignment is preceding, after the longer text alignment of issuing time.Server exports the end of sequence text list to user
In end, consulted convenient for user by terminal.
Above as can be seen that server obtains retrieval request, the retrieval request includes the voice letter for being retrieved
Breath;Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging.Based on default
Dictionary for word segmentation to the text sentence carry out word segmentation processing obtain split phrase, and based on preset stop words dictionary to described
Fractionation phrase carries out extraction and handles to obtain keyword, to remove some unwanted portions in the corresponding text sentence of voice messaging
Participle converges;Processing is extracted to all texts in the text library prestored, is obtained and the higher pass of each text relevant
Keyword list, so that the lists of keywords of each text can be carried out more in the corresponding keyword of voice messaging and text library
It accurately matches;Only when there is the target keyword to match with keyword in some lists of keywords, the pass is just exported
Resulting text corresponding to keyword list is as search result, so that the accuracy of search result is higher.
It is the implementation flow chart for the voice-based search method that second embodiment of the invention provides refering to Fig. 3, Fig. 3.This
The difference of embodiment and first embodiment is, further includes S206 after S204 in the present embodiment.Wherein S201~S205 with
S101~S105 in first embodiment is identical, referring specifically to the associated description of S101~S105 in first embodiment, herein
It does not repeat.S206 is specific as follows:
S206, if being not present and the keyword phase in lists of keywords corresponding to each text in the text library prestored
Matched target keyword then generates the notice without search result.
It is not present in S206, in the lists of keywords corresponding to the text each in the text library prestored and keyword phase
Matched target keyword illustrates that text library is not present and matches with the extracted keyword of voice messaging in retrieval request
Text, server can generate the notice of no search result, which can also be comprising for prompting user to re-enter voice letter
The prompt information of breath.Server sends the terminal of user by the notice without search result is generated, and checks retrieval to be supplied to user
As a result, convenient for user again in the terminal input include voice messaging retrieval request.
It is a kind of schematic diagram for server that third embodiment of the invention provides refering to Fig. 4, Fig. 4.Server includes each
Unit is used to execute each step in the corresponding embodiment of FIG. 1 to FIG. 3.Referring specifically to the corresponding embodiment of FIG. 1 to FIG. 3
In associated description.For ease of description, only the parts related to this embodiment are shown.Referring to fig. 4, server 4 includes:
Acquiring unit 101, for obtaining retrieval request, the retrieval request includes the voice messaging for being retrieved.
First processing units 102 obtain corresponding institute's predicate for being handled the voice messaging in the retrieval request
The text sentence of message breath.
The second processing unit 103 is obtained for carrying out word segmentation processing to the text sentence based on preset dictionary for word segmentation
Phrase is split, and extraction is carried out to the fractionation phrase based on preset stop words dictionary and handles to obtain keyword, it is described to deactivate
The vocabulary that word dictionary is filtered by needs forms.
It is corresponding to obtain each text for extracting processing to all texts in the text library prestored for extraction unit 104
Lists of keywords.
Output unit 105, if existing in lists of keywords corresponding to each text in the text library for prestoring
The target keyword to match with the keyword then exports resulting text corresponding to the lists of keywords as retrieval knot
Fruit.
Optionally, the server further include:
Generation unit, if in text library for prestoring in lists of keywords corresponding to each text there is no with it is described
The target keyword that keyword matches then generates the notice without search result.
Optionally, the extraction unit, comprising:
First determines subelement, for according to the target vocabulary of target text in the text library and the target text
First weight of the ratio of all vocabulary and the target vocabulary, determination obtain the first score value of the target vocabulary, institute
Stating target text is any one text in the text library, and the target vocabulary is any one in the target text
Vocabulary.
Second determines subelement, for according to existing in amount of text all in the text library and the text library
Second weight of the ratio of the amount of text of target vocabulary and the target vocabulary determines and obtains the of the target vocabulary
Two score values.
Third determines subelement, for obtaining the mesh according to the determination of the product of first score value and the second score value
Mark the keyword score of vocabulary.
Subelement is added, if being greater than preset value for the keyword score, the target vocabulary is added to described
In the corresponding lists of keywords of target text.
Optionally, the first processing units, comprising:
Extract subelement, for in the retrieval request voice messaging carry out feature extraction obtain phonetic feature to
Amount.
First processing subelement, carries out processing output for the speech feature vector to be input to preset acoustic model
Obtain character.
Second processing subelement obtains text for the character input to preset language model to be carried out processing output
Sentence.
Optionally, the extraction subelement is specifically used for:
Sampling processing is carried out to the voice messaging in the retrieval request according to predeterminated frequency and obtains sampled speech signal.
Preemphasis is carried out to the sampled speech signal to handle to obtain the first voice signal.
Windowing process is carried out to first voice signal and obtains the second voice signal.
Endpoint detection processing is carried out to second voice signal and obtains targeted voice signal.
Feature extraction is carried out to the targeted voice signal and obtains the speech feature vector.
Optionally, the output unit, comprising:
Sorting subunit is ranked up the resulting text for the number according to matched target keyword and is arranged
Sequence text list, the resulting text include two or more texts.
Subelement is exported, for exporting the sequence text list.
Fig. 5 is a kind of schematic diagram for server that fourth embodiment of the invention provides.As shown in figure 5, the clothes of the embodiment
Business device 5 includes: processor 50, memory 51 and is stored in the memory 51 and can run on the processor 50
Computer program 52, such as the control program of server.The processor 50 is realized above-mentioned when executing the computer program 52
Step in each voice-based search method embodiment, such as S101 shown in FIG. 1 to S105.Alternatively, the processor
The function of each unit in above-mentioned each Installation practice, such as unit 101 shown in Fig. 4 are realized when the 50 execution computer program 52
To 105 functions.
Illustratively, the computer program 52 can be divided into one or more units, one or more of
Unit is stored in the memory 51, and is executed by the processor 50, to complete the present invention.One or more of lists
Member can be the series of computation machine program instruction section that can complete specific function, and the instruction segment is for describing the computer journey
Implementation procedure of the sequence 52 in the server 5.For example, the computer program 52 can be divided into acquiring unit, first
Processing unit, the second processing unit, extraction unit and output unit, each unit concrete function are as described above.
The server may include, but be not limited only to, processor 50, memory 51.It will be understood by those skilled in the art that
Fig. 5 is only the example of server 5, does not constitute the restriction to server 5, may include than illustrating more or fewer portions
Part perhaps combines certain components or different components, such as the server can also include input and output server, net
Network access server, bus etc..
Alleged processor 50 can be central processing unit (Central Processing Unit, CPU), can also be
Other general processors, digital signal processor (Digital Signal Processor, DSP), specific integrated circuit
(Application Specific Integrated Circuit, ASIC), ready-made programmable gate array (Field-
Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor logic,
Discrete hardware components etc..General processor can be microprocessor or the processor is also possible to any conventional processor
Deng.
The memory 51 can be the internal storage unit of the server 5, such as the hard disk or memory of server 5.
The memory 51 is also possible to the external storage servers of the server 5, such as the plug-in type being equipped on the server 5
Hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card
(Flash Card) etc..Further, the memory 51 can also both include the internal storage unit of the server 5 or wrap
Include external storage servers.The memory 51 is for other journeys needed for storing the computer program and the server
Sequence and data.The memory 51 can be also used for temporarily storing the data that has exported or will export.
Embodiment described above is merely illustrative of the technical solution of the present invention, rather than its limitations;Although referring to aforementioned reality
Applying example, invention is explained in detail, those skilled in the art should understand that: it still can be to aforementioned each
Technical solution documented by embodiment is modified or equivalent replacement of some of the technical features;And these are modified
Or replacement, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution should all
It is included within protection scope of the present invention.
Claims (10)
1. a kind of voice-based search method characterized by comprising
Retrieval request is obtained, the retrieval request includes the voice messaging for being retrieved;
Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;
Word segmentation processing is carried out to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and is deactivated based on preset
Word dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary is the vocabulary for needing to be filtered
It is formed;
Processing is extracted to all texts in the text library prestored, obtains the corresponding lists of keywords of each text;
Match if existing in lists of keywords corresponding to each text in the text library prestored with the keyword
Target keyword then exports resulting text corresponding to the lists of keywords as search result.
2. voice-based search method according to claim 1, which is characterized in that the described pair of text library prestored
In all texts extract processing, after obtaining the corresponding lists of keywords of each text, comprising:
If there is no the mesh to match with the keyword in lists of keywords corresponding to each text in the text library prestored
Keyword is marked, then generates the notice without search result.
3. voice-based search method according to claim 1, which is characterized in that the described pair of text library prestored
In all texts extract processing, obtain the corresponding lists of keywords of each text, comprising:
According to the ratio of the target vocabulary of target text in the text library and all vocabulary of the target text and described
First weight of target vocabulary determines and obtains the first score value of the target vocabulary, and the target text is the text library
In any one text, the target vocabulary be the target text in any one vocabulary;
According to there are the ratios of the amount of text of the target vocabulary in amount of text all in the text library and the text library
Second weight of value and the target vocabulary, determination obtain the second score value of the target vocabulary;
The keyword score of the target vocabulary is obtained according to the determination of the product of first score value and the second score value;
If the keyword score is greater than preset value, the target vocabulary is added to the corresponding keyword of the target text
In list.
4. voice-based search method according to claim 1, which is characterized in that described in the retrieval request
Voice messaging is handled to obtain the text sentence for corresponding to the voice messaging, comprising:
Feature extraction is carried out to the voice messaging in the retrieval request and obtains speech feature vector;
By the speech feature vector be input to preset acoustic model carry out processing output obtain character;
The character input to preset language model is subjected to processing output and obtains text sentence.
5. voice-based search method according to claim 4, which is characterized in that described in the retrieval request
Voice messaging carries out feature extraction and obtains speech feature vector, comprising:
Sampling processing is carried out to the voice messaging in the retrieval request according to predeterminated frequency and obtains sampled speech signal;
Preemphasis is carried out to the sampled speech signal to handle to obtain the first voice signal;
Windowing process is carried out to first voice signal and obtains the second voice signal;
Endpoint detection processing is carried out to second voice signal and obtains targeted voice signal;
Feature extraction is carried out to the targeted voice signal and obtains the speech feature vector.
6. voice-based search method according to claim 1, which is characterized in that the output lists of keywords
Corresponding resulting text is as search result, comprising:
The resulting text is ranked up according to the number of matched target keyword to obtain sequence text list, the result
Text includes two or more texts;
Export the sequence text list.
7. a kind of server, which is characterized in that the server includes memory, processor and stores in the memory
And the computer program that can be run on the processor, the processor realize following step when executing the computer program
It is rapid:
Retrieval request is obtained, the retrieval request includes the voice messaging for being retrieved;
Voice messaging in the retrieval request is handled to obtain the text sentence for corresponding to the voice messaging;
Word segmentation processing is carried out to the text sentence based on preset dictionary for word segmentation to obtain splitting phrase, and is deactivated based on preset
Word dictionary carries out extraction to the fractionation phrase and handles to obtain keyword, and the stop words dictionary is the vocabulary for needing to be filtered
It is formed;
Processing is extracted to all texts in the text library prestored, obtains the corresponding lists of keywords of each text;
Match if existing in lists of keywords corresponding to each text in the text library prestored with the keyword
Target keyword then exports resulting text corresponding to the lists of keywords as search result.
8. server according to claim 7, which is characterized in that in the described pair of text library prestored all texts into
Row extraction process, after obtaining the corresponding lists of keywords of each text, the processor is gone back when executing the computer program
Realize following steps:
If there is no the mesh to match with the keyword in lists of keywords corresponding to each text in the text library prestored
Keyword is marked, then generates the notice without search result.
9. server according to claim 7, which is characterized in that in the described pair of text library prestored all texts into
Row extraction process obtains the corresponding lists of keywords of each text, comprising:
According to the ratio of the target vocabulary of target text in the text library and all vocabulary of the target text and described
First weight of target vocabulary determines and obtains the first score value of the target vocabulary, and the target text is the text library
In any one text, the target vocabulary be the target text in any one vocabulary;
According to there are the ratios of the amount of text of the target vocabulary in amount of text all in the text library and the text library
Second weight of value and the target vocabulary, determination obtain the second score value of the target vocabulary;
The keyword score of the target vocabulary is obtained according to the determination of the product of first score value and the second score value;
If the keyword score is greater than preset value, the target vocabulary is added to the corresponding keyword of the target text
In list.
10. a kind of computer readable storage medium, the computer-readable recording medium storage has computer program, and feature exists
In realization is such as the step of claim 1 to 6 any one the method when the computer program is executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811182073.1A CN109522392A (en) | 2018-10-11 | 2018-10-11 | Voice-based search method, server and computer readable storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811182073.1A CN109522392A (en) | 2018-10-11 | 2018-10-11 | Voice-based search method, server and computer readable storage medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109522392A true CN109522392A (en) | 2019-03-26 |
Family
ID=65770309
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811182073.1A Pending CN109522392A (en) | 2018-10-11 | 2018-10-11 | Voice-based search method, server and computer readable storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109522392A (en) |
Cited By (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110083837A (en) * | 2019-04-26 | 2019-08-02 | 科大讯飞股份有限公司 | A kind of keyword generation method and device |
CN110442855A (en) * | 2019-04-10 | 2019-11-12 | 北京捷通华声科技股份有限公司 | A kind of speech analysis method and system |
CN110442759A (en) * | 2019-07-25 | 2019-11-12 | 深圳供电局有限公司 | Knowledge retrieval method and system, computer equipment and readable storage medium |
CN110798566A (en) * | 2019-11-07 | 2020-02-14 | 浙江同花顺智能科技有限公司 | Call information recording method and device and related equipment |
CN110941743A (en) * | 2019-10-14 | 2020-03-31 | 广西壮族自治区科学技术情报研究所 | Scientific and technological project duplicate checking method for automatically realizing field weight distribution based on deep learning algorithm |
CN111078937A (en) * | 2019-12-27 | 2020-04-28 | 北京世纪好未来教育科技有限公司 | Voice information retrieval method, device, equipment and computer readable storage medium |
CN111291195A (en) * | 2020-01-21 | 2020-06-16 | 腾讯科技(深圳)有限公司 | Data processing method, device, terminal and readable storage medium |
CN111309856A (en) * | 2019-04-19 | 2020-06-19 | 北京戴纳实验科技有限公司 | Information retrieval system for laboratory engineering design |
CN111832851A (en) * | 2019-04-15 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Detection method and device |
CN112100357A (en) * | 2020-09-24 | 2020-12-18 | 腾讯科技(深圳)有限公司 | Method and device for generating guide language, electronic equipment and computer storage medium |
CN112735412A (en) * | 2020-12-25 | 2021-04-30 | 北京博瑞彤芸科技股份有限公司 | Method and system for searching information according to voice instruction |
US11019207B1 (en) | 2019-11-07 | 2021-05-25 | Hithink Royalflush Information Network Co., Ltd. | Systems and methods for smart dialogue communication |
CN113641801A (en) * | 2021-10-19 | 2021-11-12 | 成都中航信虹科技股份有限公司 | Control method and system of voice scheduling system and electronic equipment |
CN113722508A (en) * | 2021-09-02 | 2021-11-30 | 中国农业银行股份有限公司 | Word cloud display method and device, storage medium and electronic equipment |
CN114238667A (en) * | 2021-11-04 | 2022-03-25 | 北京建筑大学 | Address management method and device, electronic equipment and storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104834736A (en) * | 2015-05-19 | 2015-08-12 | 深圳证券信息有限公司 | Method and device for establishing index database and retrieval method, device and system |
CN106055546A (en) * | 2015-10-08 | 2016-10-26 | 北京慧存数据科技有限公司 | Optical disk library full-text retrieval system based on Lucene |
CN107102985A (en) * | 2017-04-23 | 2017-08-29 | 四川用联信息技术有限公司 | Multi-threaded keyword extraction techniques in improved document |
CN107861753A (en) * | 2017-06-26 | 2018-03-30 | 平安普惠企业管理有限公司 | APP generations index, search method and system and readable storage medium storing program for executing |
CN108538286A (en) * | 2017-03-02 | 2018-09-14 | 腾讯科技(深圳)有限公司 | A kind of method and computer of speech recognition |
-
2018
- 2018-10-11 CN CN201811182073.1A patent/CN109522392A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104834736A (en) * | 2015-05-19 | 2015-08-12 | 深圳证券信息有限公司 | Method and device for establishing index database and retrieval method, device and system |
CN106055546A (en) * | 2015-10-08 | 2016-10-26 | 北京慧存数据科技有限公司 | Optical disk library full-text retrieval system based on Lucene |
CN108538286A (en) * | 2017-03-02 | 2018-09-14 | 腾讯科技(深圳)有限公司 | A kind of method and computer of speech recognition |
CN107102985A (en) * | 2017-04-23 | 2017-08-29 | 四川用联信息技术有限公司 | Multi-threaded keyword extraction techniques in improved document |
CN107861753A (en) * | 2017-06-26 | 2018-03-30 | 平安普惠企业管理有限公司 | APP generations index, search method and system and readable storage medium storing program for executing |
Cited By (26)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110442855A (en) * | 2019-04-10 | 2019-11-12 | 北京捷通华声科技股份有限公司 | A kind of speech analysis method and system |
CN110442855B (en) * | 2019-04-10 | 2023-11-07 | 北京捷通华声科技股份有限公司 | Voice analysis method and system |
CN111832851A (en) * | 2019-04-15 | 2020-10-27 | 北京嘀嘀无限科技发展有限公司 | Detection method and device |
CN111832851B (en) * | 2019-04-15 | 2024-03-29 | 北京嘀嘀无限科技发展有限公司 | Detection method and device |
CN111309856B (en) * | 2019-04-19 | 2023-06-13 | 北京戴纳实验科技有限公司 | Information retrieval system for laboratory engineering design |
CN111309856A (en) * | 2019-04-19 | 2020-06-19 | 北京戴纳实验科技有限公司 | Information retrieval system for laboratory engineering design |
CN110083837B (en) * | 2019-04-26 | 2023-11-24 | 科大讯飞股份有限公司 | Keyword generation method and device |
CN110083837A (en) * | 2019-04-26 | 2019-08-02 | 科大讯飞股份有限公司 | A kind of keyword generation method and device |
CN110442759A (en) * | 2019-07-25 | 2019-11-12 | 深圳供电局有限公司 | Knowledge retrieval method and system, computer equipment and readable storage medium |
CN110442759B (en) * | 2019-07-25 | 2022-05-13 | 深圳供电局有限公司 | Knowledge retrieval method and system, computer equipment and readable storage medium |
CN110941743A (en) * | 2019-10-14 | 2020-03-31 | 广西壮族自治区科学技术情报研究所 | Scientific and technological project duplicate checking method for automatically realizing field weight distribution based on deep learning algorithm |
CN110941743B (en) * | 2019-10-14 | 2023-09-15 | 广西壮族自治区科学技术情报研究所 | Scientific and technological project duplicate checking method for automatically realizing field weight distribution based on deep learning algorithm |
US11019207B1 (en) | 2019-11-07 | 2021-05-25 | Hithink Royalflush Information Network Co., Ltd. | Systems and methods for smart dialogue communication |
US11758047B2 (en) | 2019-11-07 | 2023-09-12 | Hithink Royalflush Information Network Co., Ltd. | Systems and methods for smart dialogue communication |
US11323566B2 (en) | 2019-11-07 | 2022-05-03 | Hithink Royalflush Information Network Co., Ltd. | Systems and methods for smart dialogue communication |
CN110798566A (en) * | 2019-11-07 | 2020-02-14 | 浙江同花顺智能科技有限公司 | Call information recording method and device and related equipment |
CN111078937B (en) * | 2019-12-27 | 2021-08-10 | 北京世纪好未来教育科技有限公司 | Voice information retrieval method, device, equipment and computer readable storage medium |
CN111078937A (en) * | 2019-12-27 | 2020-04-28 | 北京世纪好未来教育科技有限公司 | Voice information retrieval method, device, equipment and computer readable storage medium |
CN111291195A (en) * | 2020-01-21 | 2020-06-16 | 腾讯科技(深圳)有限公司 | Data processing method, device, terminal and readable storage medium |
CN112100357A (en) * | 2020-09-24 | 2020-12-18 | 腾讯科技(深圳)有限公司 | Method and device for generating guide language, electronic equipment and computer storage medium |
CN112735412B (en) * | 2020-12-25 | 2022-11-22 | 北京博瑞彤芸科技股份有限公司 | Method and system for searching information according to voice instruction |
CN112735412A (en) * | 2020-12-25 | 2021-04-30 | 北京博瑞彤芸科技股份有限公司 | Method and system for searching information according to voice instruction |
CN113722508A (en) * | 2021-09-02 | 2021-11-30 | 中国农业银行股份有限公司 | Word cloud display method and device, storage medium and electronic equipment |
CN113641801A (en) * | 2021-10-19 | 2021-11-12 | 成都中航信虹科技股份有限公司 | Control method and system of voice scheduling system and electronic equipment |
CN114238667A (en) * | 2021-11-04 | 2022-03-25 | 北京建筑大学 | Address management method and device, electronic equipment and storage medium |
CN114238667B (en) * | 2021-11-04 | 2024-04-02 | 北京建筑大学 | Address management method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109522392A (en) | Voice-based search method, server and computer readable storage medium | |
CN111179975B (en) | Voice endpoint detection method for emotion recognition, electronic device and storage medium | |
CN110457432B (en) | Interview scoring method, interview scoring device, interview scoring equipment and interview scoring storage medium | |
CN107590172B (en) | Core content mining method and device for large-scale voice data | |
CN109686383B (en) | Voice analysis method, device and storage medium | |
CN109740053B (en) | Sensitive word shielding method and device based on NLP technology | |
CN112259106A (en) | Voiceprint recognition method and device, storage medium and computer equipment | |
Patel et al. | Hybrid stemmer for Gujarati | |
CN105654955B (en) | Audio recognition method and device | |
CN104142831B (en) | Application program searching method and device | |
CN111192601A (en) | Music labeling method and device, electronic equipment and medium | |
Hou et al. | Intelligent model for speech recognition based on svm: a case study on English language | |
Koolagudi et al. | Dravidian language classification from speech signal using spectral and prosodic features | |
CN109147146B (en) | Voice number taking method and terminal equipment | |
Tong et al. | A target-oriented phonotactic front-end for spoken language recognition | |
Dua et al. | Gujarati language automatic speech recognition using integrated feature extraction and hybrid acoustic model | |
Malova et al. | Recognition of emotions in verbal messages based on neural networks | |
Zajíc et al. | First insight into the processing of the language consulting center data | |
Chandramouli et al. | Unsupervised paradigm for information extraction from transcripts using BERT | |
Cornaggia-Urrigshardt et al. | Speech recognition lab | |
CN112071304A (en) | Semantic analysis method and device | |
Sadanandam | HMM Based Language Identification from Speech Utterances of Popular Indic Languages Using Spectral and Prosodic Features HMM Based Language Identification from Speech Utterances of Popular Indic Languages Using Spectral and Prosodic Features | |
Narendra et al. | Performance comparison of multitaper techniques for speaker verification with expressive speech | |
KR101620078B1 (en) | System for classifying emotion strengthen to orthographical error and method thereof | |
Bharti et al. | SVM based Voice Activity Detection by fusing a new acoustic feature PLMS with some existing acoustic features of speech |
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 |