WO2019148797A1 - Natural language processing method, device, computer apparatus, and storage medium - Google Patents

Abstract

A natural language processing method, comprising: receiving an input natural language segment, and parsing the input natural language segment by means of a pre-determined natural language parsing database, so as to obtain a natural language dependency tree; extracting a backbone structure from the natural language dependency tree; determining whether a particular interrogative word is present in the extracted backbone structure, and if so, acquiring a type of the particular interrogative word; matching the extracted backbone structure against a first standard sentence, the first standard sentence being stored in a knowledge database and corresponding to the type of the particular interrogative word; and if matching succeeds, extracting a portion corresponding to the particular interrogative word from the first standard sentence, replacing the particular interrogative word in the natural language segment with the extracted portion, and outputting the replaced natural language segment.

Description

Natural language processing method, device, computer device and storage medium

Cross-reference to related applications

This application claims the priority of the Chinese Patent Application entitled "Natural Language Processing Method, Apparatus, Computer Equipment, and Storage Media" by the Chinese Patent Office on January 30, 2018, the entire disclosure of which is incorporated by reference. Combined in this application.

Technical field

The present application relates to a natural language processing method, apparatus, computer device, and storage medium.

Background technique

With the development of computer technology, computer natural language generation has emerged, and computer natural language generation belongs to the field of artificial intelligence pattern recognition. Many of the current work is based on keyword matching mode, based on the huge real-world corpus environment library. Existing language sentences.

However, the inventors realized that the current matching method is based on keyword matching. Since keyword matching depends on the accuracy of keyword extraction, when the accuracy of extraction is low, the matching error rate is high. .

Summary of the invention

In accordance with various embodiments disclosed herein, a natural language processing method, apparatus, computer device, and storage medium are provided.

A natural language processing method that includes:

Receiving the natural language of the input, and parsing the natural language of the input through a preset natural language analysis library to obtain a natural language dependency tree;

Extracting a backbone structure in the natural language dependent tree;

Determining whether there is a special question word in the extracted backbone structure, and when present, identifying the type of the special question word;

Matching the extracted backbone structure with a first standard sentence, the first standard sentence being stored in the knowledge base and corresponding to the type of the special question word;

When the matching is successful, the part corresponding to the special question word in the first standard sentence is extracted, and the extracted part is replaced with the special question word in the natural language, and the replaced natural language is output.

A natural language processing device comprising:

a receiving module, configured to receive the input natural language, and parse the input natural language through a preset natural language parsing library to obtain a natural language dependency tree;

An extraction module, configured to extract a backbone structure in the natural language dependent tree;

a first determining module, configured to determine whether a special question word exists in the extracted backbone structure, and when present, identify a type of the special question word;

a first matching module, configured to match the extracted backbone structure with a first standard sentence, where the first standard sentence is stored in the knowledge base and corresponds to a type of the special question word;

An output module, configured to: when the matching is successful, extract a portion of the first standard sentence corresponding to the special question word, and replace the extracted part with a special question word in the natural language, and output the replacement Natural language.

A computer device comprising a memory and one or more processors having stored therein computer readable instructions, the computer readable instructions being executable by the processor to cause the one or more processors to execute The steps of the natural language processing method provided in any of the embodiments of the present application are implemented.

One or more non-transitory computer readable storage mediums storing computer readable instructions, when executed by one or more processors, cause one or more processors to perform any one of the implementations of the present application The steps of the natural language processing method provided in the example.

Details of one or more embodiments of the present application are set forth in the accompanying drawings and description below. Other features and advantages of the present invention will be apparent from the description, drawings and claims.

DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings to be used in the embodiments will be briefly described below. Obviously, the drawings in the following description are only some embodiments of the present application, Those skilled in the art can also obtain other drawings based on these drawings without any creative work.

1 is an application scenario diagram of a natural language processing method in accordance with one or more embodiments.

2 is a flow diagram of a natural language processing method in accordance with one or more embodiments.

3 is a schematic diagram of loading of a natural language parsing library in accordance with one or more embodiments.

4 is a block diagram showing the structure of a natural language dependent tree in accordance with one or more embodiments.

FIG. 5 is a flow diagram of a natural language processing method in accordance with another or more embodiments.

FIG. 6 is a block diagram of a natural language processing device in accordance with one or more embodiments.

FIG. 7 is a block diagram of a computer device in accordance with one or more embodiments.

Detailed ways

In order to make the technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are merely illustrative of the application and are not intended to be limiting.

The natural language processing method provided by the present application can be applied to the application environment as shown in FIG. 1. The user can interact with the terminal by any means such as voice, touch input, keyboard input, remote control input, and the like. Specifically, the user can speak a natural language, the terminal receives the natural language, and parses the input natural language through a preset natural language analysis library to obtain a natural language dependency tree, and extracts a backbone structure in the natural language dependency tree, thereby The processing amount can be reduced, and the key content is retained, and whether the special question word exists in the extracted backbone structure is judged, and when present, the type of the special question word is obtained; the extracted backbone structure and the special question word in the knowledge base are The first standard sentence corresponding to the type is matched; when the matching is successful, the part corresponding to the special question word in the first standard sentence is extracted, and the extracted part is replaced with the special question word in the natural language, and the output is replaced. Natural language, so that the terminal can respond to the natural language input by the user, and replace the special question word with the corresponding part of the first standard sentence in the knowledge base, without changing the structure of the sentence, so that the answer in the intelligent question and answer is related to the problem logic. Improve the accuracy of the answer. The terminal can be, but is not limited to, various personal computers, notebook computers, smart phones, tablets, and portable wearable devices.

In one embodiment, as shown in FIG. 2, a natural language processing method is provided, which is applied to the terminal in FIG. 1 as an example, and includes the following steps:

S202: Receive the natural language of the input, and parse the input natural language through a preset natural language analysis library to obtain a natural language dependency tree.

Specifically, the natural language for receiving the input may be performed by any means such as voice, touch input, keyboard input, remote control input, etc.; for example, when the terminal is installed with the voice recognition device, the natural language spoken by the user may be received, and The natural language is recognized as a voice input; or the user can input through a touch screen, a keyboard, a remote controller, or the like provided by the terminal.

Specifically, the preset natural language parsing library is a natural language parsing library of the University of Stanford. Referring to FIG. 3, the natural language parsing library of the University of Stanford can be preloaded in the terminal or the main control device, including firstly creating a natural language parser parser, and then loading the Chinese natural language training model xinhuaFactoredSegmenting.ser.gz. Optionally, it may also be a natural language training model loaded in other languages, such as English, French, and the like. After the terminal or the master device receives the natural language, the natural language is input into the natural language parsing library to obtain the natural language dependency tree.

The natural language dependent tree is obtained by dividing the natural language into parts and labeling the components of each part after the division; for example, when the natural language spoken or input by the user is "where the tomato comes from", the University of Stanford is passed. The natural language parsing library parses the input natural language to obtain the natural language dependent tree structure. See Figure 4 for details.

S204: Extract a backbone structure in the natural language dependent tree.

Specifically, the backbone structure is a structure including an action word and an object word. Optionally, the backbone structure may include at least one of a subject-predicate structure, a predicate structure, and a mediation structure. After obtaining the natural language dependency tree, the backbone structure of the input natural language is obtained by traversing the natural language dependency tree according to a preset rule. For example, the main predicate structure can be set first, and when the main predicate structure is not extracted, The predicate structure is extracted. When the predicate structure is not extracted, the mediation structure is extracted until the extraction is successful. Still using the above example, the extracted subject-predicate structure is {"ips":[{"s":"[tomato]","v":"[from]","o":"[where ]"}], where "tomato" is the main language, "from" is the predicate, and "where" is the object, then there is no need to continue to extract the object structure and the mediation structure. If the extraction fails during the process of extracting the backbone structure, that is, the subject-predicate structure, the predicate structure, or the mediation structure is not extracted, an error may be output, for example, a preset prompt may be output, for example, “not heard, please repeat” Wait.

S206: Determine whether there is a special question word in the extracted backbone structure, and when present, identify the type of the special question word.

Specifically, the special question words include a plurality of types, for example, the subject and the object related to the What-thing name can be replaced with "what". Who-person related subject and object can be replaced by "who". Where-place related subject and object can be replaced with "where", "what place", "what position" When-time related subject and object can be replaced with "when", "when", "when". Therefore, by traversing whether the special question word exists in the extracted backbone structure, when the feature question word exists in the traversed backbone structure, the type of the special question word is recognized, for example, in the embodiment of “where the tomato comes from” The special question word "where" belongs to the "where" category.

S208: Match the extracted backbone structure with a first standard sentence, where the first standard sentence is stored in the knowledge base and corresponds to the type of the special question word.

Specifically, when there is a special question word in the backbone structure, the extracted backbone structure may be matched with the first standard sentence corresponding to the special question word type in the knowledge base, wherein the knowledge base may be based on the type of the special question word. The classification is stored, so that the first standard sentence corresponding to the special question word type in the extracted backbone structure can be directly obtained, and corresponding matching can be performed. For example, the words other than the special question words in the extracted backbone structure may be matched with the first standard sentence. If the remaining words are successfully matched with the first standard sentence, the matching is considered successful, otherwise, the matching is considered to be failed.

Take the above "where the tomato comes from" as an example. The special question word is "where", and the "standard" sentence corresponding to the "where" type in the knowledge base is matched by "tomato" and "from". In the first standard sentence, there is "tomato from North America", that is, "tomato" and "tomato" match successfully, and "from" and "from" match successfully, and the match is considered successful.

S210: When the matching is successful, the part corresponding to the special question word in the first standard sentence is extracted, and the extracted part is replaced with the special question word in the natural language, and the replaced natural language is output.

Specifically, when the matching is successful, the part corresponding to the special question word in the first standard sentence is extracted, such as “North America” in the above example, and the special question word in the original natural language is replaced by the part, and the replacement is output. After the natural language, replace "where" in the original natural language with "North America" and output "tomato from North America" to complete the whole question and answer process.

The natural language processing method described above obtains a natural language dependency tree by parsing the input natural language through a preset natural language analysis library, and obtains a backbone structure according to the natural language dependency tree, by extracting a sentence backbone structure , removing the useless information; and judging the sentence pattern according to the special question words, and then querying the corresponding knowledge base according to the sentence pattern, and replacing the special question word through the corresponding part of the first standard sentence in the knowledge base, without changing the sentence The structure makes the answers in the intelligent question and answer related to the problem logic, which improves the accuracy of the answer.

In one embodiment, the natural language processing method may further include: when there is no special interrogative word in the extracted backbone structure, determining whether there is a general interrogative word in the backbone structure; when there is a general interrogative word in the backbone structure, Translating the general question words into affirmative words, and matching the transformed backbone structure with the second standard sentence in the knowledge base; when the matching is successful, converting the general question words in the natural language into affirmative words, and outputting The natural language after conversion; when the match fails, the general question words in the natural language are converted into negative words, and the converted natural language is output. In one embodiment, the natural language storage method may further include: storing the backbone structure into the knowledge base when there is neither a special question word nor a general question word in the backbone structure.

Specifically, the natural language backbone structure in this embodiment includes a subject-predicate structure, a mediation structure, and a predicate structure, and the sentence types include: a declarative sentence or an imperative sentence: a sentence without a special question word and a general question word, the sentence of the type Recognized as a knowledge statement class, the backbone of such sentences can be extracted into the knowledge base for storage. Special interrogative sentences with interrogative words: These interrogative sentences have significant subject-predicate structures and interrogative words. This type of sentence is identified as a query class. By extracting the query word query knowledge base to obtain relevant knowledge points, the corresponding natural language generation methods (replacement and negation) are used to generate the query results in the form of natural language sentences and returned to the user. General interrogative sentences: such interrogative sentences also have significant subject-predicate structure and general interrogative structure, mainly in the grammatical analysis of "verb-negative-verb" structure. This type of sentence is a type of decision problem, which is generated by the natural language generation system. After the operation method of the negative operation method, the query statement is obtained, and the knowledge base is searched. If there is such a judgment in the knowledge base, the result can be returned to the user after the operation operation is negated by the natural language generation method.

The natural language generation methods include substitution operation, negative operation and compound operation. Among them, the replacement operation is different for the different attributes of the subject and the object, and is replaced by the question without the interrogative words, and the interrogative sentences with different semantics are obtained, and the corresponding special question words can also be used. Replace the corresponding result words. The negative operation is mainly for the object part. Adding the appropriate negative word prefix or removing the specific negative word prefix can get the backbone structure of the negative semantics, or get the backbone structure of the affirmative semantics. The negative word prefix can include “no”, “no”. , "No", "No", etc. For the compound operation, the replacement operation and the negative budget are superimposed, that is, one structure can be obtained first, and the second operation can be continued on the obtained structure to obtain another structure, and so on.

Specifically, referring to FIG. 5, FIG. 5 is a flowchart of a natural language processing method in another embodiment. The terminal first receives the natural language input by the user, and then parses the grammatic dependency tree through the preset natural language parser, and extracts the corresponding sentence backbone structure from the grammatical dependency tree, and performs the extracted sentence backbone structure. Judgment can get three kinds of sentence types, and deal with different sentence types accordingly. When there are special interrogative words in the backbone of the sentence, the corresponding knowledge base is obtained according to the type of the special interrogative word, and the other words in the backbone component of the sentence are matched with the second standard sentence in the knowledge base, and the matching is successful. The second standard sentence is extracted corresponding to the part of the special question word, and the extracted word is replaced with the special question word in the natural language sentence and output to realize the intelligent question and answer. When there is no special interrogative word in the backbone structure of the sentence, it is judged whether there is a structure of "verb-negative-verb" in the sentence. If it exists, it is regarded as a general interrogative sentence. First, the sentence is negated to form a declarative sentence, and then according to The other backbone components in the sentence query the knowledge base. If the content in the knowledge base matches successfully, the statement is output. If the match is not successful, the sentence is again negated and the negative statement is output. When there is no special interrogative word in the backbone structure and there is no "verb-negative-verb" structure, the sentence is considered as a declarative sentence, and the extracted backbone structure is stored in the knowledge base to lay the foundation for other intelligent questions and answers.

Specifically, assuming that the natural language input by the user is "where the tomato comes from", the backbone structure "tomato", "from" and "where" are extracted, wherein "where" is a special question word, and is "where" type, then The second standard sentence related to the query location in the knowledge base, and the qualification condition is "tomato" in the backbone. If the response from the knowledge base is "North America", then the replacement operation in the natural language generation method described above is utilized. The operation replaces the question word "where" in the original input sentence with "North America", and finally serializes the synthetic backbone structure into a natural language sentence, and obtains the expression point of "tomato from North America" and returns it to the user.

Assuming that the natural language input by the user is "Tomato is from North America", then extract the backbone structure "tomato", "is not" and "from" "North America", then determine whether there is a "verb-negative word" in the backbone of the sentence. The structure of the verb "if present, the negative operation of the extracted backbone structure is first obtained to obtain an affirmative sentence, and then the sentence structure in the affirmative sentence is matched with the content in the knowledge base in turn, for example, the subject is first matched, and then Predicate, and finally, for the object, if the match is successful, the affirmative sentence is directly output. If the match fails, the affirmative sentence is again subjected to a negative operation and then output. In the above example, the negative structure of the backbone structure is first obtained, and the affirmative sentence of “tomato”, “yes”, “from” and “North America” is obtained, and then the knowledge base is queried whether there is such a sentence statement. If the query is true, then The structure can be serialized directly to the user according to the natural language generation method for "tomato is from North America"; if it is false, then the negative operation in the natural language generation method of this case is used to transform the query structure into "tomato is not From North America" returned to the user. Specifically, when the query is true, the "verb-negative-verb" in the original natural language is modified to "verb", otherwise it is modified to "negative-verb", and the replaced word is output.

Suppose the user inputs "tomato from North America", then extracts the backbone structure "tomato" "from" "North America", which does not contain special question words, nor does it contain general question words, ie does not contain "verb-negative words" - the structure of the verb, it is considered that the sentence does not state a sentence or imperative sentence, when the sentence is a declarative sentence or an imperative sentence, the sentence is directly saved to the knowledge base to expand the knowledge base, and in order to enhance the interest, It is possible to randomly output a certain part of the backbone of the sentence in the statement with a special question word, or perform a composite operation and output to achieve fun. For example, you can output "where the tomato comes from", which can improve the fun. To avoid duplication, you can first query whether the backbone structure exists in the knowledge base. If it exists, no operation is performed. Only when the backbone structure does not exist in the knowledge base, the backbone structure is stored in the knowledge base.

In the above embodiment, the sentence type is divided into a special question sentence, a general question sentence, and a statement sentence. In the intelligent question and answer, the sentence backbone component is first extracted, the useless information is removed, and the sentence pattern is determined according to the special question word, and then according to the sentence pattern. Query the corresponding knowledge base, replace the special question words by the replacement operation, replace the "verb-negative-verb" structure in the backbone structure by negation or replacement operation, and store the statement directly in the knowledge base, without changing The structure of the sentence makes the answer in the intelligent question and answer related to the problem logic, which improves the accuracy of the answer.

In one of the embodiments, in order to improve the efficiency of the matching, a fuzzy matching manner is introduced, wherein the components in the backbone structure may be standardized before the fuzzy matching, or after the matching fails, the manual intervention step is introduced. The manual intervention step establishes a mapping relationship between the failed backbone structure and the standard sentence, so that when the backbone structure is received subsequently, the corresponding standard sentence can be directly obtained from the knowledge base, thereby implementing the knowledge base not only through the declarative sentence. Expansion, the expansion of the knowledge base can also be achieved through manual intervention. The steps that may exist for the fuzzy matching in the above embodiment include the step of matching the converted backbone structure with the standard sentence in the knowledge base and/or the criterion for matching the extracted backbone structure with the type of the special question word in the knowledge base. The step in which the sentence is matched.

In an embodiment, the step of matching the converted backbone structure with the second standard sentence in the knowledge base may include: performing fuzzy matching on the converted backbone structure and the second standard sentence in the knowledge base; Receiving a first mapping instruction for the converted backbone structure when the converted backbone structure fails to match the second standard sentence in the knowledge base; and establishing the converted backbone structure and the first target sentence according to the first mapping instruction Match the relationship and store the first target sentence in the knowledge base.

Specifically, when there is a general interrogative word in the backbone structure, the negative structure of the backbone structure is first performed, and then the converted backbone structure is fuzzyly matched with the second standard sentence in the knowledge base, including the fuzzy matching of each part in the backbone structure. For example, when the backbone structure is the main-predicate structure, the subject, the object, and the predicate are all fuzzy-matched. For example, when the extracted backbone structure is “the tomato is from North America,” the first one is matched with “tomato”. The contents of the knowledge base are “small tomato” and “tomato”. Since the matching rate of “tomato” is 100%, which is greater than the preset value and greater than the matching rate of small tomato by 66.6%, select “tomato” as the final match. As a result, similarly, North America also performs the same match. Optionally, before the start of the fuzzy matching, the extracted backbone structure may be pre-processed, that is, standardized processing. For example, when the extracted backbone component is “tomato”, the “tomato” is first converted into “tomato”. Then, matching is performed in accordance with the above embodiment.

If the matching fails, the first mapping instruction for the converted backbone structure may be received; the matching relationship between the converted backbone structure and the first target sentence is established according to the first mapping instruction, and the first target sentence is stored to In the knowledge base, for example, when the matching fails, the content such as “Don't know” may be output, and the user may perform manual intervention to input the first target sentence “Tomato from North America”, so that after receiving the instruction, the terminal receives the instruction. The first target sentence can be stored in the knowledge base to implement the expansion of the knowledge base.

In an embodiment, the step of matching the extracted backbone structure with the first standard sentence may include: performing fuzzy matching on the extracted backbone structure with the first standard sentence; when extracting the backbone structure and the first standard sentence When the fuzzy matching fails, the second mapping instruction for the extracted backbone structure is received; the matching relationship between the extracted backbone structure and the second target sentence is established according to the second mapping instruction, and the second target sentence is stored in the knowledge base.

Specifically, when there is a special interrogative word in the backbone structure, the extracted backbone structure is fuzzyly matched with the first standard sentence in the knowledge base, including fuzzy matching of each part in the backbone structure, for example, when the backbone structure is the main-predicate structure In addition, the subject, object and predicate of the special question part should be fuzzy matched. For example, when the extracted backbone structure is “where the tomato comes from”, the content in the knowledge base matched with “tomato” is first “ "Small tomato" and "tomato", because the matching rate of "tomato" is 100%, which is greater than the preset value, and is greater than the matching rate of small tomato by 66.6%, so "tomato" is selected as the final matching result, and similarly, "from "The same match is also made. Optionally, before the start of the fuzzy matching, the extracted backbone structure may be pre-processed, that is, standardized processing. For example, when the extracted backbone component is “tomato”, the “tomato” is first converted into “tomato”. Then, matching is performed in accordance with the above embodiment.

Wherein, when the matching fails, the second mapping instruction for the converted backbone structure may be received; the matching relationship between the converted backbone structure and the second target sentence is established according to the second mapping instruction, and the second target sentence is stored to In the knowledge base, for example, when the matching fails, the content such as “not knowing” may be output. At this time, the user may manually intervene and input the second target sentence “Tomato from North America”, so that after receiving the instruction, the terminal receives the instruction. The second target sentence can be stored in the knowledge base to implement the expansion of the knowledge base.

In the above embodiment, the fuzzy matching method can improve the matching efficiency, and in the case of matching failure, manual intervention is introduced, thereby realizing the expansion of the knowledge base and improving the matching efficiency of the next time.

It should be understood that although the various steps in the flowcharts of FIGS. 2 and 5 are sequentially displayed as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Except as explicitly stated herein, the execution of these steps is not strictly limited, and the steps may be performed in other orders. Moreover, at least some of the steps in FIGS. 2 and 5 may include a plurality of sub-steps or stages, which are not necessarily performed at the same time, but may be performed at different times, or The order of execution of the stages is also not necessarily sequential, but may be performed alternately or alternately with at least a portion of the sub-steps or stages of other steps or other steps.

In an embodiment, as shown in FIG. 6, a natural language processing apparatus is provided, including: a receiving module 100, an extracting module 200, a first determining module 300, a first matching module 400, and an output module 500, wherein:

The receiving module 100 is configured to receive the input natural language, and parse the input natural language through a preset natural language analysis library to obtain a natural language dependency tree.

The extraction module 200 is configured to extract a backbone structure in the natural language dependency tree.

The first determining module 300 is configured to determine whether there is a special question word in the extracted backbone structure, and when present, identify the type of the special question word.

The first matching module 400 is configured to match the extracted backbone structure with the first standard sentence, where the first standard sentence is stored in the knowledge base and corresponds to the type of the special question word.

The output module 500 is configured to: when the matching succeeds, extract a part corresponding to the special question word in the first standard sentence, and replace the extracted part with the special question word in the natural language, and output the replaced natural language.

In one embodiment, the apparatus may further include:

The second judging module is configured to determine whether there is a general interrogative word in the backbone structure when there is no special interrogative word in the extracted backbone structure.

The second matching module is configured to convert the general question word into an affirmative word when there is a general question word in the backbone structure, and match the converted backbone structure with the second standard sentence in the knowledge base.

The output module 500 is further configured to: when the matching is successful, convert the general interrogative word in the natural language into an affirmative word, and output the converted natural language; when the matching fails, convert the general interrogative word in the natural language to a negative After the word, the converted natural language is output.

In one embodiment, the apparatus may further include:

The storage module is configured to store the backbone structure into the knowledge base when there is no special question word in the backbone structure and there is no general question word.

In one embodiment, the second matching module can include:

The first fuzzy matching unit is configured to perform fuzzy matching on the converted backbone structure with the second standard sentence in the knowledge base.

The first mapping instruction receiving unit is configured to receive a first mapping instruction for the converted backbone structure when the converted backbone structure fails to match the second standard sentence in the knowledge base.

The first mapping relationship storage unit is configured to establish a matching relationship between the converted backbone structure and the first target sentence according to the first mapping instruction, and store the first target sentence into the knowledge base.

In one embodiment, the first matching module may include:

The second fuzzy matching unit is configured to perform fuzzy matching on the extracted backbone structure with the first standard sentence.

The second mapping instruction receiving unit is configured to receive a second mapping instruction for the extracted backbone structure when the extracted backbone structure fails to match the first standard sentence.

The second mapping relationship storage unit is configured to establish a matching relationship between the extracted backbone structure and the second target sentence according to the second mapping instruction, and store the second target sentence in the knowledge base.

In one embodiment, the backbone structure may include at least one of a subject-predicate structure, a predicate structure, and a mediation structure.

For specific definitions of the natural language processing device, reference may be made to the above definition of the natural language processing method, and details are not described herein again. The various modules in the above-described natural language processing device may be implemented in whole or in part by software, hardware, and combinations thereof. Each of the above modules may be embedded in or independent of the processor in the computer device, or may be stored in a memory in the computer device in a software form, so that the processor invokes the operations corresponding to the above modules.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in FIG. The computer device includes a processor, memory, network interface, display screen, and input device connected by a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores operating systems and computer readable instructions. The internal memory provides an environment for operation of an operating system and computer readable instructions in a non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. The computer readable instructions are executed by a processor to implement a natural language processing method. The display screen of the computer device may be a liquid crystal display or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or may be a button, a trackball or a touchpad provided on the computer device casing. It can also be an external keyboard, trackpad or mouse.

It will be understood by those skilled in the art that the structure shown in FIG. 7 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation of the computer device to which the solution of the present application is applied. The specific computer device may It includes more or fewer components than those shown in the figures, or some components are combined, or have different component arrangements.

A computer device comprising a memory and one or more processors having stored therein computer readable instructions, the computer readable instructions being executed by the processor such that the one or more processors perform the step of: receiving the input natural language The natural language dependency tree is obtained by parsing the input natural language through a preset natural language analysis library; extracting the backbone structure in the natural language dependency tree; determining whether there is a special question word in the extracted backbone structure, when present, Identifying the type of the special question word; matching the extracted backbone structure with the first standard sentence, the first standard sentence is stored in the knowledge base and corresponding to the type of the special question word; and when the match is successful, the first is extracted The part of the standard sentence that corresponds to the special question word, and replaces the extracted part with the special question word in the natural language, and outputs the replaced natural language.

In one embodiment, when the processor executes the computer readable instructions, the following steps are further implemented: when there is no special interrogative word in the extracted backbone structure, it is determined whether there is a general question word in the backbone structure; when there is a general question in the backbone structure When the word is used, the general question word is converted into a positive word, and the transformed backbone structure is matched with the second standard sentence in the knowledge base; when the match is successful, the general question word in the natural language is converted into a positive word. After that, the converted natural language is output; and when the matching fails, the general question words in the natural language are converted into negative words, and the converted natural language is output.

In one embodiment, the processor, when executing the computer readable instructions, further implements the step of storing the backbone structure in the knowledge base when there are no special interrogative words in the backbone structure and no general interrogative words exist.

In an embodiment, the step of matching the converted backbone structure with the second standard sentence in the knowledge base implemented by the processor when executing the computer readable instructions may include: converting the converted backbone structure into a knowledge base The second standard sentence performs fuzzy matching; when the converted backbone structure fails to match the second standard sentence in the knowledge base, the first mapping instruction for the converted backbone structure is received; and the conversion is established according to the first mapping instruction The matching relationship between the back backbone structure and the first target sentence, and storing the first target sentence into the knowledge base.

In one embodiment, the step of matching the extracted backbone structure with the first standard sentence implemented by the processor when executing the computer readable instructions may include: performing fuzzy matching on the extracted backbone structure with the first standard sentence. Receiving a second mapping instruction for the extracted backbone structure when the extracted backbone structure fails to match the first standard sentence; and establishing a matching relationship between the extracted backbone structure and the second target sentence according to the second mapping instruction, And store the second target sentence in the knowledge base.

In one embodiment, the backbone structure in the steps implemented by the processor when executing the computer readable instructions comprises at least one of a subject-predicate structure, a predicate structure, and a mediation structure.

One or more non-transitory computer readable storage mediums storing computer readable instructions, when executed by one or more processors, cause one or more processors to perform the steps of: receiving input in nature The language, through the preset natural language analysis library, parses the input natural language to obtain a natural language dependency tree; extracts the backbone structure in the natural language dependency tree; determines whether there is a special interrogative word in the extracted backbone structure, when present, Identifying the type of the special question word; matching the extracted backbone structure with the first standard sentence, the first standard sentence is stored in the knowledge base and corresponding to the type of the special question word; and when the match is successful, the first A part of a standard sentence that corresponds to a special question word, and replaces the extracted part with a special question word in the natural language, and outputs the replaced natural language.

In one embodiment, when the computer readable instructions are executed by the processor, the following steps are further implemented: when there is no special interrogative word in the extracted backbone structure, it is determined whether there is a general interrogative word in the backbone structure; when there is a general structure in the backbone structure In the case of interrogative words, the general interrogative words are converted into affirmative words, and the transformed backbone structure is matched with the second standard sentence in the knowledge base; when the matching is successful, the general interrogative words in the natural language are converted into affirmative After the word, the converted natural language is output; and when the matching fails, the general question word in the natural language is converted into a negative word, and the converted natural language is output.

In one embodiment, when the computer readable instructions are executed by the processor, the following steps are further implemented: when there are no special interrogative words in the backbone structure and no general interrogative words exist, the backbone structure is stored in the knowledge base.

In one embodiment, the step of matching the converted backbone structure with the second standard sentence in the knowledge base implemented by the processor when the computer readable instructions are executed may include: transforming the backbone structure and the knowledge base The second standard sentence in the fuzzy matching is performed; when the converted backbone structure fails to match the second standard sentence in the knowledge base, the first mapping instruction for the converted backbone structure is received; and the first mapping instruction is established according to the first mapping instruction The matching relationship between the converted backbone structure and the first target sentence, and storing the first target sentence into the knowledge base.

In one embodiment, the step of matching the extracted backbone structure with the first standard sentence implemented by the processor when the computer readable instructions are executed may include: performing the extracted backbone structure with the first two standard sentences Fuzzy matching; receiving a second mapping instruction for the extracted backbone structure when the extracted backbone structure fails to match the first standard sentence; and establishing a matching of the extracted backbone structure with the second target sentence according to the second mapping instruction Relationship and store the second target sentence in the knowledge base.

In one embodiment, the backbone structure in the steps implemented when the computer readable instructions are executed by the processor comprises at least one of a subject-predicate structure, a predicate structure, and a mediation structure.

One of ordinary skill in the art can understand that all or part of the process of implementing the above embodiments can be completed by computer readable instructions, which can be stored in a non-volatile computer. The readable storage medium, which when executed, may include the flow of an embodiment of the methods as described above. Any reference to a memory, storage, database or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. Non-volatile memory can include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of formats, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronization chain. Synchlink DRAM (SLDRAM), Memory Bus (Rambus) Direct RAM (RDRAM), Direct Memory Bus Dynamic RAM (DRDRAM), and Memory Bus Dynamic RAM (RDRAM).

The technical features of the above embodiments may be arbitrarily combined. For the sake of brevity of description, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, It is considered to be the range described in this specification.

The above-mentioned embodiments are merely illustrative of several embodiments of the present application, and the description thereof is more specific and detailed, but is not to be construed as limiting the scope of the invention. It should be noted that a number of variations and modifications may be made by those skilled in the art without departing from the spirit and scope of the present application. Therefore, the scope of the invention should be determined by the appended claims.

Claims (20)

1. A natural language processing method that includes:

   Receiving the natural language of the input, and parsing the natural language of the input through a preset natural language analysis library to obtain a natural language dependency tree;

   Extracting a backbone structure in the natural language dependent tree;

   Determining whether there is a special question word in the extracted backbone structure, and when present, identifying the type of the special question word;

   Matching the extracted backbone structure with a first standard sentence, the first standard sentence being stored in the knowledge base and corresponding to the type of the special question word;

   When the matching is successful, the part corresponding to the special question word in the first standard sentence is extracted, and the extracted part is replaced with the special question word in the natural language, and the replaced natural language is output.
2. The method of claim 1 further comprising:

   When there is no special interrogative word in the extracted backbone structure, it is determined whether there is a general question word in the backbone structure;

   When there is a general question word in the backbone structure, the general question word is converted into an affirmative word, and the converted backbone structure is matched with the second standard sentence in the knowledge base;

   When the matching is successful, the general question words in the natural language are converted into positive words, and the converted natural language is output;

   When the matching fails, the general question words in the natural language are converted into negative words, and the converted natural language is output.
3. The method of claim 2, wherein the method further comprises:

   When there is neither a special question word nor a general question word in the backbone structure, the backbone structure is stored in the knowledge base.
4. The method according to claim 2, wherein the matching the converted backbone structure with the second standard sentence in the knowledge base comprises:

   Fuzzy matching the converted backbone structure with the second standard sentence in the knowledge base;

   Receiving a first mapping instruction for the converted backbone structure when the converted backbone structure fails to match the second standard sentence in the knowledge base;

   And establishing, according to the first mapping instruction, a matching relationship between the converted backbone structure and the first target sentence, and storing the first target sentence into the knowledge base.
5. The method according to any one of claims 1 to 3, wherein the matching the extracted backbone structure with the first standard sentence comprises:

   Fuzzyly matching the extracted backbone structure with the first standard sentence;

   Receiving, when the extracted backbone structure fails to match the first standard sentence, a second mapping instruction for the extracted backbone structure; and

   And establishing, according to the second mapping instruction, a matching relationship between the extracted backbone structure and the second target sentence, and storing the second target sentence into the knowledge base.
6. The method according to any one of claims 1 to 3, wherein the backbone structure comprises at least one of a subject-predicate structure, a predicate structure, and a mediation structure.
7. A natural language processing device comprising:

   a receiving module, configured to receive the input natural language, and parse the input natural language through a preset natural language parsing library to obtain a natural language dependency tree;

   An extraction module, configured to extract a backbone structure in the natural language dependent tree;

   a first determining module, configured to determine whether a special question word exists in the extracted backbone structure, and when present, identify a type of the special question word;

   a first matching module, configured to match the extracted backbone structure with a first standard sentence, where the first standard sentence is stored in the knowledge base and corresponds to a type of the special question word;

   An output module, configured to: when the matching is successful, extract a portion of the first standard sentence corresponding to the special question word, and replace the extracted part with a special question word in the natural language, and output the replacement Natural language.
8. The device according to claim 7, wherein the device further comprises:

   a second determining module, configured to determine whether a general question word exists in the backbone structure when there is no special question word in the extracted backbone structure;

   a second matching module, configured to convert the general interrogative word into an affirmative word when the general interrogative word exists in the backbone structure, and perform the converted backbone structure and the second standard sentence in the knowledge base Match; and

   The output module is further configured to: when the matching is successful, convert the general question words in the natural language into a positive words, and output the converted natural language; when the matching fails, the general language is After the interrogative word is converted into a negative word, the converted natural language is output.
9. A computer device comprising a memory and one or more processors having stored therein computer readable instructions, the computer readable instructions being executed by the one or more processors to cause the one or more The processor performs the following steps: receiving the natural language of the input, parsing the input natural language through a preset natural language parsing library to obtain a natural language dependency tree; extracting the backbone structure in the natural language dependent tree; and determining the extracted Whether there is a special interrogative word in the backbone structure, when present, identifying the type of the special interrogative word; matching the extracted backbone structure with a first standard sentence, the first standard sentence being stored in the knowledge base And corresponding to the type of the special question word; and when the matching is successful, extracting a portion of the first standard sentence corresponding to the special question word, and replacing the extracted part in the natural language After the special question word, the natural language after the replacement is output.
10. The computer device according to claim 9, wherein the processor, when executing the computer readable instructions, further performs the step of determining the backbone structure when there is no special interrogative word in the extracted backbone structure Whether there is a general question word in the middle; when there is a general question word in the backbone structure, the general question word is converted into an affirmative word, and the converted backbone structure and the second standard sentence in the knowledge base are performed Matching; when the matching is successful, the general question word in the natural language is converted into a positive word, and the converted natural language is output; and when the matching fails, the general question word in the natural language is converted into After the negative word, the converted natural language is output.
11. The computer apparatus according to claim 10, wherein said processor, when said computer readable instructions are executed, further performs the step of: when there is neither a special question word nor a general question word in said backbone structure The backbone structure is then stored in the knowledge base.
12. The computer device according to claim 10, wherein the processor implements the computer readable instructions to match the converted backbone structure with a second standard sentence in the knowledge base, including The fuzzy matching between the converted backbone structure and the second standard sentence in the knowledge base; when the transformed backbone structure fails to match the second standard sentence in the knowledge base, receiving the first for the converted backbone structure Mapping the instruction; establishing a matching relationship between the converted backbone structure and the first target sentence according to the first mapping instruction, and storing the first target sentence into the knowledge base.
13. The computer device according to any one of claims 9 to 11, wherein the processor performs the computer readable instruction to match the extracted backbone structure with a first standard sentence, The method includes: performing fuzzy matching on the extracted backbone structure with a first standard sentence; receiving a second mapping instruction for the extracted backbone structure when the extracted backbone structure fails to match the first standard sentence; and The second mapping instruction establishes a matching relationship between the extracted backbone structure and the second target sentence, and stores the second target sentence into the knowledge base.
14. The computer device according to any one of claims 9 to 11, wherein the backbone structure involved in the execution of the computer readable instructions by the processor comprises a subject-predicate structure, a predicate structure, and a mediator. At least one of the structures.
15. One or more non-transitory computer readable storage mediums storing computer readable instructions, when executed by one or more processors, cause the one or more processors to perform the following steps: Receiving the input natural language, parsing the input natural language through a preset natural language analysis library to obtain a natural language dependency tree; extracting the backbone structure in the natural language dependent tree; determining whether there is a special question in the extracted backbone structure a word, when present, identifying a type of the special question word; matching the extracted backbone structure with a first standard sentence, the first standard sentence being stored in the knowledge base, and the special question Corresponding to the type of the word; and when the matching is successful, extracting the part of the first standard sentence corresponding to the special question word, and replacing the extracted part with the special question word in the natural language, and outputting the replacement After the natural language.
16. A storage medium according to claim 15, wherein said computer readable instructions, when executed by said processor, further perform the step of: determining said backbone when there is no special interrogative word in said extracted backbone structure Whether there is a general question word in the structure; when there is a general question word in the backbone structure, the general question word is converted into an affirmative word, and the converted backbone structure and the second standard sentence in the knowledge base are Matching; when the matching is successful, converting the general interrogative words in the natural language into affirmative words, outputting the converted natural language; and when the matching fails, converting the general interrogative words in the natural language After the negative word, the converted natural language is output.
17. A storage medium according to claim 16, wherein said computer readable instructions, when executed by said processor, further perform the step of: when there is neither a special interrogative word in said backbone structure nor a general question When the word is used, the backbone structure is stored in the knowledge base.
18. The storage medium according to claim 16, wherein said computer readable instructions are matched by said processor to perform said converted backbone structure with a second standard sentence in a knowledge base, The method comprises: performing fuzzy matching on the converted backbone structure with a second standard sentence in the knowledge base; and receiving, when the converted backbone structure fails to match the second standard sentence in the knowledge base, receiving the converted backbone structure a mapping instruction; establishing a matching relationship between the converted backbone structure and the first target sentence according to the first mapping instruction, and storing the first target sentence into the knowledge base.
19. A storage medium according to any one of claims 15 to 17, wherein said computer readable instructions are matched by said processor to perform said matching of said extracted backbone structure with a first standard sentence The method includes: performing fuzzy matching on the extracted backbone structure with the first standard sentence; receiving a second mapping instruction for the extracted backbone structure when the extracted backbone structure fails to match the first standard sentence; and The second mapping instruction establishes a matching relationship between the extracted backbone structure and the second target sentence, and stores the second target sentence into the knowledge base.
20. The storage medium according to any one of claims 15 to 17, wherein said backbone structure involved in execution of said computer readable instructions by said processor comprises a subject-predicate structure, a predicate structure, and a mediation At least one of the guest structures.

Images