CN114610842A - Associated searching method and system based on intention identification - Google Patents
Associated searching method and system based on intention identification Download PDFInfo
- Publication number
- CN114610842A CN114610842A CN202210082469.9A CN202210082469A CN114610842A CN 114610842 A CN114610842 A CN 114610842A CN 202210082469 A CN202210082469 A CN 202210082469A CN 114610842 A CN114610842 A CN 114610842A
- Authority
- CN
- China
- Prior art keywords
- intention
- search
- retrieval
- data source
- field
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses an associated search method and system based on intention recognition, wherein the method comprises the following steps: acquiring a retrieval content text; performing intention identification according to the retrieval content text to obtain an intention retrieval field; obtaining a target data source according to the intention retrieval field association; and searching the target data source to obtain a search result. The intention retrieval field is obtained by firstly identifying the intention according to the retrieval content text, so that the intention retrieval field is associated with the target data source, accurate search is performed without traversing all data sources, the efficiency of the search process is improved, and the search result can meet the search intention of the user.
Description
Technical Field
The present application relates to the field of information retrieval technologies, and in particular, to an associated search method and system based on intention recognition.
Background
At present, search engines for enterprise users and individual users generally perform content search on different data sources according to the search intention of the enterprise users or the individual users.
However, for indiscriminate search, the search engine needs to query all data sources one by one and then perform mixed ranking on the searched results, which results in low efficiency of the search process and also fails to ensure that the search results satisfy the search intention of the user.
Disclosure of Invention
In order to improve the efficiency of a search process and enable a search result to meet the search intention of a user, the application provides an associated search method and system based on intention identification.
In a first aspect, the present application provides an association search method based on intent recognition, which adopts the following technical solutions:
an intent recognition based associative search method, comprising:
acquiring a retrieval content text;
performing intention identification according to the retrieval content text to obtain an intention retrieval field;
obtaining a target data source according to the intention retrieval field association;
and searching the target data source to obtain a search result.
By adopting the technical scheme, before searching, intention recognition is carried out according to the retrieval content text to obtain the intention retrieval field, so that the intention retrieval field is associated with the target data source, accurate searching is carried out without traversing all data sources for searching, the efficiency of the searching process is improved, and the searching result can meet the searching intention of a user.
Optionally, the performing intent recognition according to the search content text to obtain an intent search field includes:
matching the search content text with a preset keyword library through keywords;
and if the target keywords are successfully matched, performing intention identification based on the target keywords to obtain an intention retrieval field.
Optionally, the performing intent recognition according to the search content text to obtain an intent search field includes:
carrying out hot word matching on the search content text and a preset hot word library;
and if the target hot words are successfully matched, performing intention identification based on the target hot words to obtain the intention retrieval field.
Optionally, the method further includes:
if the target keywords and/or the target hot words are not successfully matched, vectorizing the retrieval content text;
inputting the search content text subjected to vectorization processing into a pre-trained convolutional neural intention recognition model to obtain a model output result, and taking the model output result as the intention search field.
Optionally, before the performing the intention recognition according to the search content text and obtaining the intention search field, the method further includes:
acquiring important words in data of all data sources, and taking the important words as first hot words;
acquiring a user search history text, and performing non-noun cleaning on the user search history text to obtain a second hot word;
forming a hot word bank according to the first hot word and the second hot word;
and receiving the hot repair words input by the operation and maintenance personnel, and adding the hot repair words into the hot word library as new hot words.
Optionally, before the performing the intention recognition according to the search content text and obtaining the intention search field, the method further includes:
acquiring service marking data and data in a data source;
according to the type of a preset label, giving a type label to the service marking data and the data in the data source;
and after vectorization processing is carried out on the data of the class labels, training is carried out on a text convolution neural network model, and a convolution neural intention recognition model is obtained.
Optionally, obtaining the target data source according to the association of the intention retrieval field includes:
determining a number of sub-domains of the intent retrieval domain;
when one sub-domain exists, determining a target data source corresponding to the sub-domain;
when a plurality of sub-fields exist, setting the priority of each sub-field according to the historical search heat of the sub-fields;
and determining a target data source corresponding to each sub-domain according to a principle that the priority is from high to low.
Optionally, the searching the target data source to obtain a search result includes:
when only one target data source corresponding to one sub-field exists, searching the target data source to obtain a search result;
when a plurality of target data sources corresponding to the sub-fields exist, the corresponding target data sources are sequentially searched according to the priority of each sub-field and the principle that the priority is from high to low, and the search result corresponding to each target data source is obtained.
In a second aspect, the present application provides an association search system based on intent recognition, which adopts the following technical solutions:
the acquisition module is used for acquiring a retrieval content text;
the intention identification module is used for identifying the intention according to the retrieval content text to obtain an intention retrieval field;
the data source correlation module is used for obtaining a target data source according to the intention retrieval field correlation;
and the searching module is used for searching the target data source to obtain a searching result.
In summary, the present application includes the following beneficial technical effects:
before searching, intention recognition is carried out according to the search content text to obtain an intention search field, so that the intention search field is associated with a target data source to carry out accurate search, all data sources do not need to be traversed for searching, the efficiency of a search process is improved, and a search result can meet the search intention of a user.
Drawings
Fig. 1 is a schematic flow chart of an association search method based on intention recognition according to the present application.
Fig. 2 is a flowchart illustrating a process of obtaining an intention search field by performing intention recognition in a keyword manner according to the present application.
Fig. 3 is a flowchart illustrating a process of obtaining an intention search field by performing intention recognition in a hotword manner according to the present application.
Fig. 4 is a flowchart illustrating a process of obtaining an intention retrieval area by performing intention recognition by means of a convolutional neural intention recognition model according to the present application.
FIG. 5 is a schematic flow chart illustrating search results obtained by searching according to the intention of the domain-associated target data source.
FIG. 6 is a schematic structural diagram of the intent-based associative search system of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
The embodiment of the application discloses an associated search method based on intention identification.
Referring to fig. 1, the method includes:
101, obtaining the search content text.
In the present situation, when an application with a search function is oriented to an enterprise user or a personal user, since the service and service range that can be provided are not single, the user needs to input search content during searching and searching, the search content may be an option of clicking an application, a segment of text may be input, or after voice input, voice information is converted into a text, and the final result is a search content text, which may generally be a short word, a sentence or a long segment of text. Specifically, the embodiment of the present application has an application of a search function, and the main functions include policy search, patent search, trademark search, copyright search, and research and development cooperation search.
And 102, identifying the intention according to the search content text, and determining an intention search field.
For example, an intellectual property service platform may relate to policy retrieval, patent retrieval, trademark retrieval, copyright retrieval and research and study cooperation retrieval, and there are 5 data sources, namely, a trademark database, a copyright database, a patent database, a policy database and a research and study cooperation database, each of which needs to be searched separately, and consumes a lot of time and resources to perform traversal search. The mode of carrying out intention identification to obtain the intention retrieval field specifically comprises 3, wherein the first mode is a keyword, the second mode is a hotword, and the third mode is a convolution nerve intention identification model. For example, the input search content text is H04N, H04N is one subclass of the H large class in the international patent classification number IPC, the classification number H04N indicates that the patent classification belongs to image communication, and it is obvious that the user inputs H04N, all patents belonging to H04N need to be searched from the patent database, and the intention search field is the patent database.
103, obtaining the target data source according to the intention retrieval domain association.
The only corresponding associated target data source is searched through the determined intention retrieval field, accurate searching is carried out, and all data sources do not need to be traversed for searching.
And 104, searching the target data source to generate a search result.
The application has the implementation principle that: before searching, intention recognition is carried out according to the search content text to obtain an intention search field, so that the intention search field is associated with a target data source, accurate searching is carried out without traversing all data sources for searching, the efficiency of a search process is improved, and a search result can meet the search intention of a user.
The method for obtaining the intention search area by performing intention recognition in step 102 of the embodiment shown in fig. 1 includes 3 specific ways, the first is a keyword, the second is a hotword, and the third is a convolutional neural intention recognition model, which will be described below by embodiments. This segment may be placed after step 102.
As shown in fig. 2, the process of obtaining the intention retrieval field by performing intention recognition in a keyword manner includes the following specific steps:
and 201, performing keyword matching on the search content text and a preset keyword library.
The preset keyword library is extracted according to the specific keywords of the data in the data source, for example, the keywords contained in the preset keyword library related to the patent database include all patent classification numbers, which are possessed by each patent and not possessed by a trademark, which are possessed by the trademark classification numbers. If the retrieval content text comprises the patent classification number, the retrieval intention is obviously a retrieval patent, and the corresponding intention retrieval field is certainly a patent field, so that keywords with obvious intentions can be extracted to form a preset keyword library, keyword matching is carried out on the retrieval content text and the preset keyword library, and whether words in the retrieval content text are the same as the keywords in the preset keyword library or not is identified.
And 202, if the target keywords are successfully matched, performing intention identification based on the target keywords to obtain an intention retrieval field.
If the related key words are the same, the target key words are successfully matched, intention recognition is carried out based on the target key words, for example, the target key words are 'trademarks', the trademarks are wanted to be searched, and the corresponding intention searching field is the trademark field; the target keyword is 'copyright', namely the intention to retrieve the copyright, and the corresponding intention retrieval field is the copyright field. In this embodiment, patents, copyrights, and trademarks are used as examples, and in practical applications, many keywords are used for identification purposes, which cannot be listed.
It should be noted that when the target keyword is not successfully matched, it indicates that the intention recognition by the keyword mode fails, and in order to guarantee the user experience, the intention recognition by the hotword or the convolutional neural intention recognition model may be turned to.
The implementation principle of the embodiment is as follows: the method comprises the steps of carrying out keyword matching on a retrieval content text and a preset keyword library, and identifying the user intention according to the matched target keyword, so that the intention identification field is determined, the intention identification efficiency is improved, and the accuracy of the identified intention identification field is high.
As shown in fig. 3, the process of obtaining the intention search field by performing intention recognition in a hotword manner includes the following specific steps:
301, obtaining important words in the data of all the data sources, and taking the important words as first hot words.
The important words are extracted from the data of all data sources, and the extracted important words can be used as first hot words according to the occurrence times of the important technical words, or the short names of listed companies or the short names of famous institutions and the like.
302, obtaining a user search history text, and performing non-noun cleaning on the user search history text to obtain a second hot word.
The method can also include obtaining a user search history text, namely aggregating the user search history, selecting a sentence with the highest occurrence frequency from the user search history text, and then cleaning a non-noun for the sentence, for example, the sentence is a verb, preposition, conjunctions and the like except the noun, namely, only the words of "Hunan province", "high-new enterprises" and "declaration conditions" are reserved to obtain a second hotword.
And 303, forming a hot word library according to the first hot word and the second hot word.
And summarizing the extracted first hot words and the extracted second hot words to form a hot word bank.
And 304, receiving the input hot repair words, and adding the hot repair words as new hot words to the hot word library.
And 305, performing hot word matching on the search content text and a preset hot word library.
And carrying out hot word matching on the search content text and a preset hot word bank, and identifying whether words in the search content text are the same as hot words in the preset hot word bank.
And 306, if the target hot words are successfully matched, performing intention identification based on the target hot words to obtain an intention retrieval field.
If the hot words are the same, the target hot word is successfully matched, and the intention is identified based on the hot words to obtain the intention retrieval field.
It should be noted that when the target hotword is not successfully matched, it indicates that the intention recognition by the hotword method fails, and in order to guarantee the user experience, the intention recognition by the keyword or the convolutional neural intention recognition model may be turned to.
The implementation principle of the embodiment is as follows: the method comprises the steps of carrying out hot word matching on a search content text and a preset hot word library, and identifying the user intention according to the matched target hot words, so that the intention identification field is determined, the intention identification efficiency can be improved, and the identified intention identification field can be more accurate.
As shown in fig. 4, the process of obtaining the intention retrieval domain by performing intention recognition in a convolution neural intention recognition model includes the following specific steps:
401, obtaining service marking data and data in a data source.
The service marking data is obtained by marking service data which has been subjected to retrieval service historically, and the data in the data source mainly refers to a real-time Search Engine (ES) data source.
And 402, according to the preset label type, giving a type label to the service labeling data and the data in the data source.
According to the preset label type, a type label is given to the service labeling data and the data in the data source, specifically, the type label can be marked on the sentence.
403, after vectorization processing is performed on the data of the category labels, training is performed on the text convolutional neural network model to obtain a convolutional neural intention recognition model.
The method comprises the steps of conducting vectorization processing on data of category labels, namely conducting vectorization processing on sentences, inputting the vectorized sentences into a text convolutional neural network (TextCNN) model, calculating categories through a softmax logistic regression mode, using cross entropy as a loss function in a training process, using the cross entropy as the loss function which is commonly used in a classification problem and used for representing the difference between a model prediction result and real result distribution, improving the model prediction index through continuously optimizing the cross entropy until the model index converges, and finally conducting training to obtain the convolutional neural intention recognition model.
And 404, if the target keywords and/or the target hot words are not successfully matched, vectorizing the retrieval content text.
In the embodiments of fig. 2 and fig. 3, since both keyword matching and hotword matching can only be performed on repeatedly appearing words and important search sentences, and a large number of long search sentences cannot be covered, in the process of matching target keywords and target hotwords, if one of the words is not matched or both words are not matched, vectorization processing needs to be performed on the search content text.
And 405, inputting the search content text subjected to vectorization into a pre-trained convolutional neural intention recognition model to obtain a model output result, and taking the model output result as an intention search field.
The implementation principle of the embodiment is as follows: the intention identification mode of the keywords and the hot words only aims at the repeated words and important search sentences, and in the keyword matching and the hot word matching process, as long as one is not matched or both are not matched, in order to ensure the completeness of the search,
the user intention is identified through the pre-trained convolutional neural intention identification model to obtain the intention retrieval field, and the integrity of the identification intention identification field is improved.
In the above embodiment shown in fig. 1, in the process of associating the target data source according to the intention retrieval area in step 103 and step 104 and obtaining the search result, because the intention retrieval area is not only one, in order to increase the speed of the search, the association and the search of the data source may be performed in a priority manner, specifically as shown in fig. 5, the steps include:
501, the number of sub-domains of the intended search domain is determined.
There may be a plurality of intention retrieval fields, for example, in an intellectual property retrieval platform, there may be enterprise retrieval, trademark retrieval, copyright retrieval, local policy retrieval and patent retrieval for different services, and if the intention retrieval field identified in retrieving the content text includes both the trademark field and the patent field, that is, there are two included sub-fields.
502, when a sub-domain exists, a target data source corresponding to the sub-domain is determined.
If only one sub-field, namely the trademark field, exists, the target data source corresponding to the trademark field can be determined.
When there are a plurality of sub-domains, priority is set for each sub-domain according to the search heat of the history of the sub-domains 503.
Wherein, if there are a plurality of sub-fields, if two are respectively the policy field and the patent field, the priority of the sub-field patent field is higher than that of the policy field according to the historical search heat of the sub-field, and the search times of the patent field is more than that of the policy field.
And 504, determining a target data source corresponding to each sub-domain according to a principle that the priority is from high to low.
505, when only one target data source corresponding to the sub-field exists, searching the target data source to obtain a search result.
And 506, when a plurality of target data sources corresponding to the sub-fields exist, sequentially searching the corresponding target data sources according to the priority of each sub-field and the principle that the priority is from high to low to obtain the search result corresponding to each target data source.
When searching, because the target data sources are determined according to the priorities of the sub-fields from high to low, the corresponding target data sources can be sequentially searched according to the principle that the priorities are from high to low, and the search result corresponding to each target data source is obtained.
The implementation principle of the embodiment is as follows: the intention retrieval fields can be one or more, when only one is provided, only one corresponding target data source is provided, and the purpose of direct searching is achieved.
It should be noted that, in steps 505 and 506 of the embodiment of fig. 5, a plurality of target data sources corresponding to one sub-field may be provided, for example, taking a patent database as an example, each country has a patent database, and databases of various languages are also divided in the global scope, so that the target database corresponding to the "patent field" is a plurality of sub-databases including a plurality of countries, regions, and languages, and for the plurality of sub-databases, especially the chinese patent database and the english patent database, the number is very large, the time spent on searching is relatively large, and in order to further improve the search efficiency, the following steps are performed before searching the target database in steps 505 and 506:
when the target database comprises a plurality of sub-databases, acquiring language characteristics and region characteristics of each sub-database, wherein the language characteristics represent the used languages of the sub-databases, and the region characteristics represent the countries or regions of data sources of the sub-databases;
extracting language information and region information according to the retrieval content text;
and matching the target sub-database from the plurality of sub-databases according to the language information, the region information, and the language characteristics and the region characteristics of the sub-databases.
In the above embodiment, the method for searching for associations based on intention identification is explained in detail, and in the following, an association searching system based on intention identification to which the method is applied is explained by an embodiment, as shown in fig. 6, the application provides an association searching system based on intention identification, which includes:
an obtaining module 601, configured to obtain a search content text;
an intention identification module 602, configured to perform intention identification according to the retrieved content text to obtain an intention retrieval field;
a data source association module 603, configured to obtain a target data source according to the intention retrieval domain association;
the searching module 604 is configured to search the target data source to obtain a search result.
The implementation principle of the application is as follows: before the search module 604 performs search, the intention identification module 602 performs intention identification on the search content text acquired by the acquisition module 601 to obtain an intention search field, and the data source association module 603 associates with the target data source, so that the search module 604 performs accurate search without traversing all data sources for search, the efficiency of the search process is improved, and the search result can meet the search intention of the user.
The foregoing is a preferred embodiment of the present application and is not intended to limit the scope of the application in any way, and any features disclosed in this specification (including the abstract and drawings) may be replaced by alternative features serving equivalent or similar purposes, unless expressly stated otherwise. That is, unless expressly stated otherwise, each feature is only an example of a generic series of equivalent or similar features.
Claims (9)
1. An intent recognition-based association search method, comprising:
acquiring a retrieval content text;
performing intention identification according to the retrieval content text to obtain an intention retrieval field;
obtaining a target data source according to the intention retrieval field association;
and searching the target data source to obtain a search result.
2. The associated search method based on intention recognition according to claim 1, wherein the intention recognition is performed according to the retrieved content text to obtain an intention retrieval field, comprising:
matching the search content text with a preset keyword library by keywords;
and if the target keywords are successfully matched, performing intention identification based on the target keywords to obtain an intention retrieval field.
3. The associated search method based on intention recognition according to claim 1, wherein the intention recognition is performed according to the retrieved content text to obtain an intention retrieval field, comprising:
carrying out hot word matching on the search content text and a preset hot word library;
and if the target hot words are successfully matched, performing intention identification based on the target hot words to obtain the intention retrieval field.
4. The intent recognition based association search method according to claims 2 and 3, further comprising:
if the target keywords and/or the target hot words are not successfully matched, vectorizing the retrieval content text;
inputting the search content text subjected to vectorization processing into a pre-trained convolutional neural intention recognition model to obtain a model output result, and taking the model output result as the intention search field.
5. The associated search method based on intention recognition according to claim 4, wherein before the intention recognition is performed according to the retrieved content text to obtain an intention retrieval area, the method further comprises:
acquiring important words in data of all data sources, and taking the important words as first hot words;
acquiring a user search history text, and performing non-noun cleaning on the user search history text to obtain a second hot word;
forming a hot word bank according to the first hot word and the second hot word;
and receiving an input hot repair word, and adding the hot repair word as a new hot word into the hot word library.
6. The associated search method based on intention recognition according to claim 4, wherein before the intention recognition is performed according to the retrieved content text to obtain an intention retrieval area, the method further comprises:
acquiring service marking data and data in a data source;
according to the type of a preset label, giving a type label to the service marking data and the data in the data source;
and after vectorization processing is carried out on the data of the class labels, training is carried out on a text convolution neural network model, and a convolution neural intention recognition model is obtained.
7. The method for searching for association based on intention identification according to claim 1, wherein the obtaining of the target data source according to the intention retrieval domain association comprises:
determining a number of sub-domains of the intent retrieval domain;
when one sub-domain exists, determining a target data source corresponding to the sub-domain;
when a plurality of sub-domains exist, setting the priority of each sub-domain according to the historical searching heat of the sub-domains;
and determining a target data source corresponding to each sub-field according to a principle that the priority is from high to low.
8. The intent recognition based association search method of claim 7, wherein searching the target data source for a search result comprises:
when only one target data source corresponding to one sub-field exists, searching the target data source to obtain a search result;
when a plurality of target data sources corresponding to the sub-fields exist, the corresponding target data sources are sequentially searched according to the priority of each sub-field and the principle that the priority is from high to low, and the search result corresponding to each target data source is obtained.
9. An intent recognition based associative search system, comprising:
the acquisition module is used for acquiring a retrieval content text;
the intention identification module is used for identifying the intention according to the retrieval content text to obtain an intention retrieval field;
the data source correlation module is used for obtaining a target data source according to the intention retrieval field correlation;
and the searching module is used for searching the target data source to obtain a searching result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210082469.9A CN114610842A (en) | 2022-01-24 | 2022-01-24 | Associated searching method and system based on intention identification |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210082469.9A CN114610842A (en) | 2022-01-24 | 2022-01-24 | Associated searching method and system based on intention identification |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114610842A true CN114610842A (en) | 2022-06-10 |
Family
ID=81858067
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210082469.9A Pending CN114610842A (en) | 2022-01-24 | 2022-01-24 | Associated searching method and system based on intention identification |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114610842A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743606A (en) * | 2024-02-21 | 2024-03-22 | 天云融创数据科技（北京）有限公司 | Intelligent retrieval method and system based on big data |
-
2022
- 2022-01-24 CN CN202210082469.9A patent/CN114610842A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117743606A (en) * | 2024-02-21 | 2024-03-22 | 天云融创数据科技（北京）有限公司 | Intelligent retrieval method and system based on big data |
| CN117743606B (en) * | 2024-02-21 | 2024-04-30 | 天云融创数据科技（北京）有限公司 | Intelligent retrieval method and system based on big data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108804521B (en) | Knowledge graph-based question-answering method and agricultural encyclopedia question-answering system | |
| CN110765257A (en) | Intelligent consulting system of law of knowledge map driving type | |
| US20040249808A1 (en) | Query expansion using query logs | |
| CN111967761B (en) | Knowledge graph-based monitoring and early warning method and device and electronic equipment | |
| CN103678576A (en) | Full-text retrieval system based on dynamic semantic analysis | |
| CN109308321A (en) | A kind of knowledge question answering method, knowledge Q-A system and computer readable storage medium | |
| CN108509521B (en) | Image retrieval method for automatically generating text index | |
| CN113076411B (en) | Medical query expansion method based on knowledge graph | |
| CN111666425B (en) | Automobile accessory searching method based on semantic knowledge | |
| CN112463944B (en) | Search type intelligent question-answering method and device based on multi-model fusion | |
| CN112307182B (en) | Question-answering system-based pseudo-correlation feedback extended query method | |
| CN112328800A (en) | System and method for automatically generating programming specification question answers | |
| CN110659357B (en) | Geographic knowledge question-answering system based on ontology semantic similarity | |
| CN115840812A (en) | Method and system for intelligently matching enterprises according to policy text | |
| CN113761208A (en) | Scientific and technological innovation information classification method and storage device based on knowledge graph | |
| CN111625621A (en) | Document retrieval method and device, electronic equipment and storage medium | |
| CN111159381B (en) | Data searching method and device | |
| Sasikumar et al. | A survey of natural language question answering system | |
| Kamdi et al. | Keywords based closed domain question answering system for indian penal code sections and indian amendment laws | |
| CN114610842A (en) | Associated searching method and system based on intention identification | |
| CN108491407B (en) | Code retrieval-oriented query expansion method | |
| RU2345416C1 (en) | Method of synthesis of self-trained analytical question-answer system with extraction of knowledge from texts | |
| CN111209375B (en) | Universal clause and document matching method | |
| CN110309258A (en) | A kind of input checking method, server and computer readable storage medium | |
| Chen et al. | FAQ system in specific domain based on concept hierarchy and question type |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 518051 2201, block D, building 1, bid section 1, Chuangzhi Yuncheng, Liuxian Avenue, Xili community, Xili street, Nanshan District, Shenzhen, Guangdong Applicant after: Qizhi Technology Co.,Ltd. Address before: 518051 2201, block D, building 1, bid section 1, Chuangzhi Yuncheng, Liuxian Avenue, Xili community, Xili street, Nanshan District, Shenzhen, Guangdong Applicant before: Qizhi Network Technology Co.,Ltd. |