A user awareness and keyword extraction system based on intelligent recognition

Abstract

The present invention relates to the field of AI, and in particular to a user consciousness and keyword extraction system based on intelligent identification, including a consciousness understanding module, an automatic reply module, a user interface, and a data feedback optimization module; the user interface is used to receive user input information or output reply information of the automatic reply module, the consciousness understanding module is used to analyze the input information received by the user interface and identify semantics and extract keywords based on the input information, and the automatic reply module is used to generate reply information based on the analysis results of the consciousness understanding module and send the reply information to the user interface. The present invention understands the user's needs through the consciousness understanding module, searches for corresponding data from the database according to the user's needs and automatically replies, and makes various responses based on the user's reply, without manual operation throughout the process, saving labor.

Description

Technical Field

[0001] The present invention relates to the field of AI, and in particular to a user awareness and keyword extraction system based on intelligent recognition. Background Art

[0002] With the rapid development of the Internet and the continuous maturity of intelligent technology, people have generated a large amount of text and voice data in their daily lives and work. However, these massive amounts of data are often difficult to understand and use directly, which poses challenges to information management and intelligent services. In order to better understand and use these data, natural language processing (NLP) technology has emerged and has become an important tool for processing and analyzing text and voice data. At present, intelligent systems based on NLP technology have been widely used in many fields, such as intelligent customer service, intelligent search, public opinion monitoring, intelligent translation, etc.

[0003] For example, the keyword extraction system disclosed in the prior art of CN1198225C and the text retrieval system using the system, the system generates a single word ID of each single word used in the text and a single word occurrence count of the corresponding single word for each text constituting the text library. The system generates a total single word occurrence count and a text inclusion count for each single word used in any text constituting the text library, which represents the number of texts containing the single word. For each single word contained in the selected text, the importance is calculated using the single word occurrence count, the total single word occurrence count and the text inclusion count. The single words contained in the selected text are sorted in order of importance. At least a part of the sorted single words are displayed as selected keywords.

[0004] Another typical prior art disclosed in CN112784021A is a system and method for using keywords extracted from comments, a method and system for generating search results. In some examples, one or more keywords are extracted from one or more stored comments associated with commercial offerings. The first extracted keyword is associated with a stored list of commercial offerings. The first extracted keyword may not be in the stored list. In response to a search query including a first keyword, a set of search results is provided, wherein the set of search results includes a list associated with the first keyword.

[0005] Let's look at the system and method for extracting index keyword data fields disclosed in the prior art such as CN1345426A. The system includes a definition query for extracting index definitions and column definitions from a database. The system also includes a code generator connected to the definition query. The code generator can create a computer code for extracting index keyword data fields based on the index definition and column definition. A table analyzer is connected to the code generator and uses the computer code to extract index data from a data table.

[0006] Existing natural language processing technology often relies on a dictionary library for word segmentation and semantic recognition. When some new words or proper nouns appear, they are often unable to be correctly recognized, causing many inconveniences in use. In order to solve the common problems in this field, the present invention is made. Summary of the invention

[0007] The purpose of the present invention is to propose a power cable for DC charging that is anti-crushing and anti-interference in view of the current deficiencies.

[0008] In order to overcome the shortcomings of the prior art, the present invention adopts the following technical solutions: a user awareness and keyword extraction system based on intelligent identification, comprising a consciousness understanding module, an automatic reply module, a user interface and a data feedback optimization module; the user interface is used to receive user input information or output reply information of the automatic reply module, the consciousness understanding module is used to analyze the input information received by the user interface and identify semantics and extract keywords based on the input information, the automatic reply module is used to generate reply information based on the analysis results of the consciousness understanding module and send the reply information to the user interface, and the data feedback optimization module is used to receive user feedback on the reply information and optimize the consciousness understanding module and the automatic reply module based on the feedback information.

[0009] Furthermore, the consciousness understanding module includes an input information receiving unit, a preprocessing unit, a semantic analysis unit and a keyword extraction unit. The information receiving unit is used to receive input information received by the user interface. The preprocessing unit preprocesses the input information. The preprocessing includes noise reduction and input information conversion. The input information conversion includes converting voice information into text information. The semantic analysis unit is used to analyze the semantics of text information. The keyword extraction unit is used to extract keywords from text information.

[0010] Furthermore, the automatic reply module includes a reply generator and a reply output unit, the reply generator is used to generate corresponding reply information according to the analysis results of the consciousness understanding module, the reply output unit is used to output the reply information to the user interface, the reply generator includes a rule engine, the rule engine is used to match the semantics analyzed by the semantic analysis unit and the keywords extracted by the keyword extraction unit with the reply rules, and output the corresponding reply content according to the matching results.

[0011] Furthermore, the semantic analysis unit includes a word segmentation unit, a part-of-speech tagging unit, a dictionary, a syntactic analysis unit and a semantic analysis unit. The word segmentation unit is used to segment the text information into words according to the dictionary, the part-of-speech tagging unit is used to tag the part of speech of the words according to the dictionary, the syntactic analysis unit is used to analyze the structure and grammatical relationship of the sentence, and the semantic analysis unit is used to analyze the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit.

[0012] Furthermore, the workflow of the user awareness and keyword extraction system includes the following steps:

[0013] S1, the user interface receives information input by the user;

[0014] S2, the consciousness understanding module understands the input information received by the user interface and sends the analysis results to the automatic reply module;

[0015] S3, the automatic reply module generates reply content according to the analysis results;

[0016] S4, the user interface displays the reply content, and the user provides feedback in the user interface;

[0017] S5, the data feedback optimization module receives the user's feedback information and determines whether the reply content meets the user's expectations. If the reply content meets the user's expectations, the user's input information is added to the dictionary, and the analysis results of the consciousness understanding module and the reply content of the automatic reply module are added to the training set of the rule engine; if the reply content does not meet the user's expectations, a re-question request is sent to the user interface.

[0018] Furthermore, the consciousness understanding module understands the input information received by the user interface including the following steps:

[0019] S21, the preprocessing unit preprocesses the received input information to obtain text information;

[0020] S22, the word segmentation unit divides the text information into words according to the dictionary;

[0021] S23, the part-of-speech tagging unit tags the part-of-speech of the word according to the dictionary;

[0022] S24, the syntactic analysis unit analyzes the structure and grammatical relations of the sentence;

[0023] S25, the semantic analysis unit analyzes the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit;

[0024] S26, the keyword extraction unit extracts keywords from the word segmentation results according to the word segmentation results of the word segmentation unit.

[0025] Furthermore, the automatic reply module generates reply content according to the analysis result, including the following steps:

[0026] S31, searching for corresponding data from a database according to the keywords extracted by the keyword extraction unit;

[0027] S32, inputting the received semantic analysis result into the rule engine, and the rule engine outputs the corresponding reply template;

[0028] S33, fill the data searched from the database into the reply template to obtain reply information;

[0029] S34, the reply output unit sends the reply information to the user interface.

[0030] Furthermore, the word segmentation unit segments the text information into words according to the dictionary, including the following steps:

[0031] S221, starting from the beginning of the text information, obtaining m characters as the fields to be segmented, where m is the maximum number of characters included in the words in the dictionary;

[0032] S222, matching the field to be segmented with the words in the dictionary, if successful, executing S224, otherwise, executing S223;

[0033] S223, remove the last character from the field to be segmented, obtain a new field to be segmented, and return to S222;

[0034] S224, removing the field to be segmented from the text information as a segmented field, and returning to S221 until all characters in the text information are segmented into words;

[0035] S225, determining the accuracy index of each segmented word according to the following formula:

[0036] ;

[0037] in, is the accuracy index, X is the number of words before the segmented word, and Y is the number of words after the segmented word. is the probability that the xth word appears before the segmented word in the paragraph where the segmented word appears in the dictionary, is the probability that the yth word appears after the segmented word in the paragraph where the segmented word appears in the dictionary, for The relevance weight of for The relevance weight of .

[0038] The beneficial effects achieved by the present invention are: 1. Understanding the needs of users through the consciousness understanding module, searching for corresponding data from the database and automatically replying according to the needs of users, and making various reactions according to the replies of users. No manual operation is required throughout the process, saving labor.

[0039] 2. The accuracy of each word is evaluated through the accuracy index, which avoids the system forcibly identifying new words or uncommon words when encountering them, and generates corresponding replies based on the results of forced recognition, which is conducive to improving the accuracy of replies and improving user satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The present invention can be further understood from the following description in conjunction with the accompanying drawings. The components in the figures are not necessarily drawn to scale, but the emphasis is placed on illustrating the principles of the embodiments. In different views, the same reference numerals designate the same parts.

[0041] Figure 1 It is a schematic diagram of the structure of the present invention.

[0042] Figure 2 It is the work flow chart of the present invention.

[0043] Figure 3 The present invention is a flow chart of the consciousness understanding module understanding the input information received by the user interface.

[0044] Figure 4 A flow chart of generating reply content for the dynamic reply module of the present invention. DETAILED DESCRIPTION

[0045] The following is an explanation of the embodiments of the present invention through specific embodiments. Those skilled in the art can understand the advantages and effects of the present invention from the contents disclosed in this specification. The present invention can be implemented or applied through other different specific embodiments, and the details in this specification can also be modified and changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. In addition, the drawings of the present invention are only simple schematic illustrations and are not depicted according to actual sizes. It is stated in advance. The following embodiments will further explain the relevant technical contents of the present invention in detail, but the disclosed contents are not intended to limit the scope of protection of the present invention.

[0046] Embodiment 1: According to Figure 1 , Figure 2 , Figure 3 and Figure 4, this embodiment provides 1. a user awareness and keyword extraction system based on intelligent identification, characterized in that it includes a consciousness understanding module, an automatic reply module, a user interface and a data feedback optimization module; the user interface is used to receive user input information or output reply information of the automatic reply module, the consciousness understanding module is used to analyze the input information received by the user interface and identify semantics and extract keywords based on the input information, the automatic reply module is used to generate reply information based on the analysis results of the consciousness understanding module and send the reply information to the user interface, and the data feedback optimization module is used to receive user feedback on the reply information and optimize the consciousness understanding module and the automatic reply module based on the feedback information.

[0047] 2. A user awareness and keyword extraction system based on intelligent recognition, characterized in that the awareness understanding module includes an input information receiving unit, a preprocessing unit, a semantic analysis unit and a keyword extraction unit, the information receiving unit is used to receive input information received by the user interface, the preprocessing unit preprocesses the input information, the preprocessing includes noise reduction and input information conversion, the input information conversion includes converting voice information into text information, the semantic analysis unit is used to analyze the semantics of text information, and the keyword extraction unit is used to extract keywords from text information.

[0048] 3. A user awareness and keyword extraction system based on intelligent identification, characterized in that the automatic reply module includes a reply generator and a reply output unit, the reply generator is used to generate corresponding reply information according to the analysis results of the awareness understanding module, the reply output unit is used to output the reply information to the user interface, the reply generator includes a rule engine, the rule engine is used to match the semantics analyzed by the semantic analysis unit and the keywords extracted by the keyword extraction unit with the reply rules, and output the corresponding reply content according to the matching results.

[0049] 4. A user awareness and keyword extraction system based on intelligent recognition, characterized in that the semantic analysis unit includes a word segmentation unit, a part-of-speech tagging unit, a dictionary, a syntactic analysis unit and a semantic analysis unit. The word segmentation unit is used to segment text information into words according to the dictionary, the part-of-speech tagging unit is used to tag the part of speech of the words according to the dictionary, the syntactic analysis unit is used to analyze the structure and grammatical relationship of the sentence, and the semantic analysis unit is used to analyze the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit.

[0050] Specifically, the dictionary contains a large number of words and their corresponding parts of speech and meanings, as well as a large number of reference articles. Each word in the dictionary corresponds to multiple articles containing the word. The dictionary is set up by technical personnel in this field. Depending on user awareness and the use field of the keyword extraction system, the scope of articles covered by the dictionary varies, thereby ensuring that word segmentation units, etc. can perform their functions more accurately.

[0051] 5. A user awareness and keyword extraction system based on intelligent identification, characterized in that the workflow of the user awareness and keyword extraction system includes the following steps:

[0052] S1, the user interface receives information input by the user;

[0053] S2, the consciousness understanding module understands the input information received by the user interface and sends the analysis results to the automatic reply module;

[0054] S3, the automatic reply module generates reply content according to the analysis results;

[0055] S4, the user interface displays the reply content, and the user provides feedback in the user interface;

[0056] Specifically, the user's feedback content is whether the reply content meets the user's expected reply;

[0057] S5, the data feedback optimization module receives the user's feedback information and determines whether the reply content meets the user's expectations. If the reply content meets the user's expectations, the user's input information is added to the dictionary, and the analysis results of the consciousness understanding module and the reply content of the automatic reply module are added to the training set of the rule engine; if the reply content does not meet the user's expectations, a re-question request is sent to the user interface.

[0058] Specifically, the re-questioning request is used to remind the user to change the questioning method to ask questions.

[0059] 6. A user awareness and keyword extraction system based on intelligent recognition, characterized in that the awareness understanding module understands the input information received by the user interface including the following steps:

[0060] S21, the preprocessing unit preprocesses the received input information to obtain text information;

[0061] Specifically, the preprocessing unit converts the speech information into text information through a converter model, and the converter model is set by a person skilled in the art based on a self-attention mechanism and a position encoding technology, and the self-attention mechanism and the position encoding technology belong to the prior art and will not be described in detail here;

[0062] S22, the word segmentation unit divides the text information into words according to the dictionary;

[0063] S23, the part-of-speech tagging unit tags the part-of-speech of the word according to the dictionary;

[0064] Specifically, the dictionary includes the part of speech of each word. When a word includes multiple parts of speech, the accuracy index of each part of speech is determined according to the following formula, and the part of speech with the larger accuracy index is selected as the part of speech of the word in the text information:

[0065] ;

[0066] in, is the accuracy index of the part of speech, which is used to evaluate the relative correctness of the part of speech when it is substituted into the corresponding words in the text information; the higher the relative correctness, the higher the accuracy of the part of speech after it is substituted into the corresponding words compared with other parts of speech. is the probability that the previous word of the word corresponding to the part of speech is connected to the word corresponding to the part of speech in the dictionary, is the probability that the next word of the word corresponding to the part of speech is connected to the word corresponding to the part of speech in the dictionary. When the word corresponding to the part of speech is the first word in the text information, Set to 0, when the word corresponding to this part of speech is the last word in the text information, Set to 0;

[0067] S24, the syntactic analysis unit analyzes the structure and grammatical relations of the sentence;

[0068] Specifically, the syntactic analysis unit analyzes the sentence including the following steps: analyzing the hierarchical structure of phrases in the sentence according to the segmented words and their corresponding parts of speech by using a recursive descent method or a predictive analysis method, wherein the phrases may include one or more words; analyzing the dependency relationship between words by using a shift-reduce algorithm or an arc marking algorithm, and constructing a dependency tree, wherein the dependency relationship describes the modification relationship between words. The above method belongs to the prior art and will not be elaborated herein.

[0069] S25, the semantic analysis unit analyzes the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit;

[0070] Specifically, the semantic analysis unit analyzes the semantics of text information, including the following steps: semantic role labeling: inputting the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit into the statistical model, identifying the predicates (verbs or nouns) in the sentence, and determining their semantic roles (such as agent, patient, time, place, etc.), the statistical model includes a hidden Markov model (HMM) or a conditional random field (CRF), and the statistical model is trained by a technician in this field; semantic reasoning: semantic reasoning is performed through a rule engine and the results of semantic role labeling, the rule engine includes TensorFlow, PyTorch, NLTK, etc., and the rule engine is trained by a technician in this field; semantic synthesis: converting the results of semantic reasoning into vector form through tools such as a recurrent neural network (RNN), a long short-term memory network (LSTM), and a Transformer.

[0071] S26, the keyword extraction unit extracts keywords from the word segmentation results according to the word segmentation results of the word segmentation unit.

[0072] Specifically, a keyword library is provided, and the keyword extraction unit extracts keywords by matching the word segmentation results with the keywords in the keyword library.

[0073] 7. A user awareness and keyword extraction system based on intelligent identification, characterized in that the automatic reply module generates reply content according to the analysis results, including the following steps:

[0074] S31, searching for corresponding data from a database according to the keywords extracted by the keyword extraction unit;

[0075] Specifically, the keywords may be "yesterday", "weather", "balance", etc. The keywords and corresponding data are set by those skilled in the art. For example, if the extracted keywords are "yesterday" and "weather", the database is searched for data related to yesterday's weather.

[0076] S32, inputting the received semantic analysis result into the rule engine, and the rule engine outputs the corresponding reply template;

[0077] Specifically, the rule engine may be Drools or ILOG, which belongs to the prior art and will not be described in detail here;

[0078] S33, fill the data searched from the database into the reply template to obtain reply information;

[0079] S34, the reply output unit sends the reply information to the user interface.

[0080] 8. A user awareness and keyword extraction system based on intelligent recognition, characterized in that the word segmentation unit divides text information into words according to a dictionary, comprising the following steps:

[0081] S221, starting from the beginning of the text information, obtaining m characters as the fields to be segmented, where m is the maximum number of characters included in the words in the dictionary;

[0082] Specifically, when the value of m is greater than the number of characters in the text information, the m characters are not obtained as the fields to be segmented, but the remaining characters in the text information are obtained as the fields to be segmented;

[0083] S222, matching the field to be segmented with the words in the dictionary, if successful, executing S224, otherwise, executing S223;

[0084] S223, remove the last character from the field to be segmented, obtain a new field to be segmented, and return to S222;

[0085] S224, removing the field to be segmented from the text information as a segmented field, and returning to S221 until all characters in the text information are segmented into words;

[0086] S225, determining the accuracy index of each segmented word according to the following formula:

[0087] ;

[0088] in, is the accuracy index, X is the number of words before the segmented word, and Y is the number of words after the segmented word. is the probability that the xth word appears before the segmented word in the paragraph where the segmented word appears in the dictionary, is the probability that the yth word appears after the segmented word in the paragraph where the segmented word appears in the dictionary, for The relevance weight of for The relevance weight of .

[0089] Specifically, the preceding and following refer to the position of the word in the text information;

[0090] Specifically, when the segmented word is the first word of the text information, let:

[0091] ;

[0092] When the segmented word is the last word of the text information, let:

[0093] ;

[0094] Specifically, we can use the following formula to and To obtain:

[0095] ;

[0096] ;

[0097] in, is the number of characters between the first character of the segmented word and the last character of the x-th word, is the number of characters between the last character of the segmented word and the first character of the yth word;

[0098] Preferably, it can also be carried out according to the following formula and To obtain:

[0099] ;

[0100] ;

[0101] in, is the accuracy index of the part of speech corresponding to the xth word, is the accuracy index of the part of speech corresponding to the yth word, and D is the feedback weight;

[0102] Specifically, the value of the feedback weight is set by a person skilled in the art;

[0103] Preferably, the value of the feedback weight can be , where m is the maximum number of characters included in the words in the dictionary, and M is the number of characters in the text information;

[0104] Preferably, in the working process of the consciousness understanding module, the accuracy index of the part of speech may not be used to calculate the and , and use this as the basis for the initial word segmentation. After the initial word segmentation, recalculate by obtaining the accuracy index of the part of speech after word segmentation and And re-segmentation and re-analysis of parts of speech are carried out, which is conducive to verifying the accuracy of the word segmentation results according to the results of part of speech analysis, thereby improving the accuracy of the word segmentation results;

[0105] Specifically, when the accuracy index of a segmented word is lower than a set threshold, the word is judged as an unregistered word. When the accuracy indexes of multiple consecutive words are all lower than the set threshold, the combination of the consecutive words is used as a new segmentation result and marked as an unregistered word. The threshold is set by a person skilled in the art.

[0106] For example, when a user inputs a popular online term such as "amazing" or a new term "metaverse", if forced word segmentation is performed, the word segmentation result may be "amazing", "amazing", "son", "meta", "verse". Such a word segmentation result is generally difficult to be coherent with its context (usually "amazing" and "amazing" are not connected together, and "meta" and "verse" are not connected together), and the probability of appearance in the dictionary is relatively low. Therefore, the calculated accuracy index will be relatively low. By combining multiple consecutive words with accuracy indexes lower than the set threshold, out-of-vocabulary words can be obtained (combining "meta" and "verse" can obtain "metaverse").

[0107] Beneficial effects of this solution: 1. The demand of the user is understood through the awareness understanding module, corresponding data is searched from the database according to the user's demand and an automatic reply is made, and various reactions are made according to the user's reply. The whole process does not require manual operation, saving labor force.

[0108] 2. The accuracy of each word is evaluated through the accuracy index, avoiding the system from forcibly recognizing new words or rare words when encountering them and generating corresponding replies according to the forcibly recognized results, which is beneficial to improving the accuracy of the reply and the satisfaction of the user.

[0109] Embodiment 2: This embodiment should be understood as including all the features of any one of the foregoing embodiments and further improved thereon. It also lies in including a method for the awareness understanding module to understand text information when out-of-vocabulary words appear in the word segmentation result, including the following steps:

[0110] STEP1, the part-of-speech tagging unit tags the part of speech of the word according to the dictionary;

[0111] Specifically, the part-of-speech tagging unit tags the part of speech of the out-of-vocabulary word according to the following steps:

[0112] Assume that the word before the out-of-vocabulary word in the text information is A and the word after it is B; when the out-of-vocabulary word is the first word in the text information, only B is considered in the following content, and when the out-of-vocabulary word is the last word in the text information, only A is considered in the following content

[0113] Calculate the probability index of each part of speech according to the following formula:

[0114] ;

[0115] Wherein, is the probability index of part of speech a, in the dictionary, is the total number of times of, is the number of times with part of speech a in the words of, For the dictionary, The number of occurrences of For the dictionary, The number of times the part of speech is a in the next word, is the number of occurrences of B in the dictionary, is the number of times the word before B has the same part of speech as a in the dictionary; the part of speech with the largest probability index is selected as the part of speech of the unregistered word;

[0116] STEP2, the syntactic analysis unit analyzes the structure and grammatical relationship of the sentence;

[0117] Specifically, when the syntactic analysis unit is working, it only analyzes the words according to their parts of speech, so its analysis method is the same for registered words and unregistered words;

[0118] STEP3, the semantic analysis unit analyzes the semantics of the text information containing the unregistered word according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit;

[0119] Specifically, the semantic analysis unit analyzes the semantics of text information containing unregistered words, including the following steps: semantic role labeling: inputting the analysis results of the word segmentation unit, the part-of-speech labeling unit and the syntactic analysis unit into the statistical model, identifying the predicates (verbs or nouns) in the sentence, and determining their semantic roles (such as agent, patient, time, place, etc.), which is the same for registered words and unregistered words; semantic reasoning: semantic reasoning is performed through a rule engine and the results of semantic role labeling, and the rule engine is trained by a person skilled in the art; semantic synthesis: converting the results of semantic reasoning into a vector form, which is the same for registered words and unregistered words;

[0120] Preferably, performing semantic reasoning on unregistered words through the rule engine and the result of semantic role labeling includes the following steps:

[0121] STEP 31, obtain the initial word segmentation results, that is, the word segmentation results before the accuracy index of the uncalculated words and the unmarked and unregistered words;

[0122] Specifically, H is an unregistered word, and the initial segmentation result corresponding to H is ,…… , h is the number of words corresponding to the initial word segmentation results contained in the unregistered words;

[0123] Assume that the unknown word is "metaverse" ("H"), and the initial segmentation result corresponding to H is (Yuan), (universe);

[0124] STEP32, obtain according to the method in STEP1 Part of speech;

[0125] STEP 33, obtain arrive The words with the same part of speech as H;

[0126] STEP 34, calculate according to the following formula arrive Semantic weight index of words with the same part of speech as H in:

[0127] ;

[0128] in, for arrive The semantic weight index of the mth word with the same part of speech as H in is the number of occurrences of the mth word in the dictionary, for arrive The number of words with the same part of speech as H in is the probability that A appears in the first h words of the mth word in the dictionary, is the probability that B appears in the last h words of the mth word in the dictionary;

[0129] STEP35, take the word with the largest semantic weight index as the approximate word of the unregistered word and input it into the rule engine instead of the unregistered word to obtain the semantic reasoning result.

[0130] Specifically, if you obtain arrive If the number of words with the same part of speech as H is 0, then the unregistered word is directly input into the rule engine. However, directly inputting the unregistered word is prone to misunderstanding. By using similar words to replace them, the probability of misunderstanding can be reduced.

[0131] The beneficial effects of this embodiment are as follows: by performing approximate recognition on the semantics of unrecognizable words, words with similar meanings to the unrecognizable words can be found as much as possible, and the unrecognizable words can be replaced with words with similar meanings, which is conducive to providing the most accurate response possible to the user's input information when the words cannot be recognized.

[0132] The above disclosed contents are only preferred feasible embodiments of the present invention, and do not limit the protection scope of the present invention. Therefore, all equivalent technical changes made by using the contents of the present invention specification and drawings are included in the protection scope of the present invention. In addition, the elements therein can be updated with the development of technology. The above units are only examples. Those skilled in the art can use corresponding units according to different designs according to actual needs when implementing this solution.

Claims

1. A user awareness and keyword extraction system based on intelligent recognition, characterized in that: It includes a consciousness understanding module, an automatic reply module, a user interface and a data feedback optimization module; the user interface is used to receive user input information or output reply information of the automatic reply module, the consciousness understanding module is used to analyze the input information received by the user interface and identify semantics and extract keywords according to the input information, the automatic reply module is used to generate reply information according to the analysis result of the consciousness understanding module and send the reply information to the user interface, and the data feedback optimization module is used to receive user feedback on the reply information and optimize the consciousness understanding module and the automatic reply module according to the feedback information; The word segmentation unit divides the text information into words according to the dictionary, including the following steps: S221, starting from the beginning of the text information, obtaining m characters as the fields to be segmented, where m is the maximum number of characters included in the words in the dictionary; S222, matching the field to be segmented with the words in the dictionary, if successful, executing S224, otherwise, executing S223; S223, remove the last character from the field to be segmented, obtain a new field to be segmented, and return to S222; S224, removing the field to be segmented from the text information as a segmented field, and returning to S221 until all characters in the text information are segmented into words; S225, determining the accuracy index of each segmented word according to the following formula: ； in, is the accuracy index, X is the number of words before the segmented word, and Y is the number of words after the segmented word. is the probability that the xth word appears before the segmented word in the paragraph where the segmented word appears in the dictionary, is the probability that the yth word appears after the segmented word in the paragraph where the segmented word appears in the dictionary, for The relevance weight of for The relevance weight of .

2. A user awareness and keyword extraction system based on intelligent identification according to claim 1, characterized in that: The consciousness understanding module includes an input information receiving unit, a preprocessing unit, a semantic analysis unit and a keyword extraction unit. The information receiving unit is used to receive input information received by the user interface. The preprocessing unit preprocesses the input information. The preprocessing includes noise reduction and input information conversion. The input information conversion includes converting voice information into text information. The semantic analysis unit is used to analyze the semantics of text information. The keyword extraction unit is used to extract keywords from text information.

3. A user awareness and keyword extraction system based on intelligent identification according to claim 2, characterized in that: The automatic reply module includes a reply generator and a reply output unit. The reply generator is used to generate corresponding reply information according to the analysis results of the consciousness understanding module, and the reply output unit is used to output the reply information to the user interface. The reply generator includes a rule engine. The rule engine is used to match the semantics analyzed by the semantic analysis unit and the keywords extracted by the keyword extraction unit with the reply rules, and output the corresponding reply content according to the matching results.

4. A user awareness and keyword extraction system based on intelligent identification according to claim 3, characterized in that: The semantic analysis unit includes a word segmentation unit, a part-of-speech tagging unit, a dictionary, a syntactic analysis unit and a semantic analysis unit. The word segmentation unit is used to segment text information into words according to the dictionary, the part-of-speech tagging unit is used to tag the part of speech of the words according to the dictionary, the syntactic analysis unit is used to analyze the structure and grammatical relationship of the sentence, and the semantic analysis unit is used to analyze the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit.

5. A user awareness and keyword extraction system based on intelligent identification according to claim 4, characterized in that: The workflow of the user awareness and keyword extraction system includes the following steps: S1, the user interface receives information input by the user; S2, the consciousness understanding module understands the input information received by the user interface and sends the analysis results to the automatic reply module; S3, the automatic reply module generates reply content according to the analysis results; S4, the user interface displays the reply content, and the user provides feedback in the user interface; S5, the data feedback optimization module receives the user's feedback information and determines whether the reply content meets the user's expectations. If the reply content meets the user's expectations, the user's input information is added to the dictionary, and the analysis results of the consciousness understanding module and the reply content of the automatic reply module are added to the training set of the rule engine; if the reply content does not meet the user's expectations, a re-question request is sent to the user interface.

6. A user awareness and keyword extraction system based on intelligent identification according to claim 5, characterized in that: The consciousness understanding module understands the input information received by the user interface, including the following steps: S21, the preprocessing unit preprocesses the received input information to obtain text information; S22, the word segmentation unit divides the text information into words according to the dictionary; S23, the part-of-speech tagging unit tags the part-of-speech of the word according to the dictionary; S24, the syntactic analysis unit analyzes the structure and grammatical relations of the sentence; S25, the semantic analysis unit analyzes the semantics of the text information according to the analysis results of the word segmentation unit, the part-of-speech tagging unit and the syntactic analysis unit; S26, the keyword extraction unit extracts keywords from the word segmentation results according to the word segmentation results of the word segmentation unit.

7. A user awareness and keyword extraction system based on intelligent identification according to claim 6, characterized in that: The automatic reply module generates reply content according to the analysis result, including the following steps: S31, searching for corresponding data from a database according to the keywords extracted by the keyword extraction unit; S32, inputting the received semantic analysis result into the rule engine, and the rule engine outputs the corresponding reply template; S33, fill the data searched from the database into the reply template to obtain reply information; S34, the reply output unit sends the reply information to the user interface.

Images