WO2016082575A1

WO2016082575A1 - Information mining method and apparatus, and storage medium

Info

Publication number: WO2016082575A1
Application number: PCT/CN2015/086095
Authority: WO
Inventors: 刘松; 孙凯; 陶明远
Original assignee: 百度在线网络技术（北京）有限公司
Priority date: 2014-11-27
Filing date: 2015-08-05
Publication date: 2016-06-02
Also published as: CN104346480A; CN104346480B

Abstract

Provided are an information mining method and apparatus, and a storage medium. The method includes: monitoring a message issued in an instant communication software application (110); parsing the monitored message to obtain message content (120); matching the message content with a keyword in a pre-established feature identification dictionary (130); and when the matching succeeds, capturing the message content or the message content and related content of the message content as feature description information, and saving the feature description information (140). The message issued in the instant communication software application has high category definition and high information specialism, such that the feature description information of a specific object can be automatically captured by matching the parsed message content with the keyword in the feature identification dictionary and capturing the successfully matched message content or capturing the successfully matched message content and the related content of the message content, thereby reducing the labour costs, and improving the specialism and accuracy of the obtained feature description information of the specific object.

Description

Information mining method, device and storage medium

This patent application claims the priority of the Chinese patent application filed on November 27, 2014, the application number is 201410710424.7, the applicant is Baidu Online Network Technology (Beijing) Co., Ltd., and the invention name is "Information Mining Method and Apparatus". The full text of the application is hereby incorporated by reference.

Technical field

The embodiments of the present invention relate to the field of information technologies, and in particular, to an information mining method, apparatus, and storage medium.

Background technique

In the prior art, information related to objects such as products or services, such as product defect description information that is helpful for product improvement, is usually manually captured in a forum or webpage in a related field, which is inefficient and accurate. not tall.

Summary of the invention

The embodiment of the invention provides an information mining method, device and storage medium, so as to automatically capture feature information of a specific object, save labor cost, and improve the accuracy of the captured feature information of the specific object.

In a first aspect, an embodiment of the present invention provides an information mining method, including:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

Matching the message content with keywords in a pre-established feature recognition dictionary;

When the matching is successful, the content of the message that matches the success is captured, or the content of the message that matches the success and the related content of the message content that is successfully matched are captured as the feature description information, and the feature description information is saved.

In a second aspect, an embodiment of the present invention further provides an information mining apparatus, including:

a message monitoring module for monitoring messages published in an instant messaging software application;

a message parsing module, configured to parse the intercepted message to obtain a message content;

a matching module, configured to match the content of the message with a keyword in a pre-established feature recognition dictionary;

a feature description information processing module, configured to: when the matching is successful, capture the content of the message that matches the success, or capture the content of the message that matches the success and the related content of the message content that is successfully matched as the feature description information, and the feature is Describe the information for saving.

In a third aspect, an embodiment of the present invention further provides a non-volatile computer storage medium, where the computer storage medium stores one or more modules, when the one or more modules are used by a device that performs an information mining method. When executed, the device is caused to perform the following operations:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

The information mining method, device and storage medium provided by the embodiments of the present invention monitor and parse messages published in the instant messaging software application, because the message published in the instant messaging software application not only has high category definition, but also has high professionalism, so Matching the parsed message content with the keywords in the pre-established feature recognition dictionary, and capturing the successfully matched message content, or capturing the successfully matched message content and the related content of the message content, can automatically capture the specific object The feature description information saves the labor cost, and improves the professionalism and accuracy of the obtained feature description information of the specific object, and is beneficial to improve the specific object according to the feature description information.

DRAWINGS

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following description will be made on the embodiments. The drawings to be used are briefly described. Of course, the drawings in the following description are only some embodiments of the present invention, and those skilled in the art can also attach these without any creative work. The diagram is modified and replaced.

FIG. 1 is a flowchart of an information mining method according to Embodiment 1 of the present invention;

2 is a flowchart of an information mining method according to Embodiment 2 of the present invention;

3a is a flowchart of an information mining method according to Embodiment 3 of the present invention;

FIG. 3b is a flowchart of another information mining method according to Embodiment 3 of the present invention;

3c is a flowchart of still another information mining method according to Embodiment 3 of the present invention;

4 is a schematic structural diagram of an information mining apparatus according to Embodiment 4 of the present invention;

FIG. 5 is a schematic structural diagram of hardware of an apparatus for performing an information mining method according to Embodiment 6 of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings. Rather than limiting the invention to these specific embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

Embodiment 1

FIG. 1 is a flowchart of an information mining method according to Embodiment 1 of the present invention. The method of an embodiment of the present invention may be performed by an information mining device configured to be implemented in hardware and/or software, which is typically configured in a server capable of providing data mining services.

The method includes operations 110 through 140.

110. Monitor messages published in the instant messaging software application.

Usually, each enterprise has an instant messaging software application related to the enterprise product or department, so as to facilitate the release of messages in the enterprise responsible for each product development group or responsible for operation and maintenance.

For example, Baidu Hi launched by Baidu is a set of text messages, voice and video calls and file transfers. For the instant messaging software application of the loss function, you can establish a group corresponding to the product "Baidu Map" or the product "Baidu Translation" in Baidu Hi, so as to facilitate the release of the staff responsible for product development or operation and maintenance in Baidu. Message.

There are a plurality of ways for publishing a message, which may be advertised in the form of text, or may be advertised in other forms such as voice, video, or picture. This embodiment does not limit this, as long as the instant messaging software application support is available.

The operation may specifically be to listen to a text message published in a group related to the enterprise product or a group related to the enterprise department in the instant messaging software application. If the message is published in other forms, it is necessary to identify the content through a related identification technology (for example, voice recognition technology or picture recognition technology) to obtain a corresponding text message.

120. Parse the monitored message to obtain the message content.

In this operation, specifically, according to the communication protocol of the instant messaging software application, the intercepted message is translated, and the original data corresponding to the monitored message is correctly restored, that is, the readable readable string is restored.

130. Match the content of the message with a keyword in a pre-established feature recognition dictionary.

The operation specifically uses a keyword matching technology to determine whether the keyword in the feature recognition dictionary is included in the message content according to a pre-established feature recognition dictionary.

It should be noted that the group publishing messages corresponding to each object in the enterprise are different, and the content of the parsed message is different. The group has the characteristics of high definition, high information professionality and obvious language features (for example, each group contains group members that are a category or a group of people who have the same product or similar professional background) Therefore, messages published by different groups can reflect enterprise object information.

Among them, the object can be a specific product, or it can be a macro object such as enterprise management.

For example, the group corresponding to the "Baidu Map" product is a group that Baidu Company is responsible for the development or operation maintenance of "Baidu Map". The messages published by the members of the group usually contain the advantages and disadvantages of the product, or the follow-up of the product. Improve the information.

For example, the message posted by the group members in the debugging group corresponding to the "Baidu Browser" product usually contains bugs or suspected problems during the debugging process of the product.

Therefore, a corresponding feature recognition dictionary can be established for a group corresponding to different objects of the enterprise, thereby Obtain feature description information corresponding to different objects (such as different products, or enterprise management) (such as the advantages and disadvantages of different products, or problems in enterprise management); for different groups of the same object of the enterprise, it is preferred to establish corresponding features. The dictionary is identified to obtain different levels of feature description information related to the same object.

For example, the R&D group in the "Baidu Map" product establishes a feature recognition dictionary related to R&D, and the keywords in the dictionary may include "R&D", "Progress", "Trend", "Cost", and "Entity"; A feature recognition dictionary related to debugging is established for the debugging group in the "Baidu Map" product, and the keywords in the dictionary may include "debug error", "debug cycle", "bug", "vulnerability" and "defect", etc.; The feature recognition dictionary related to the release group in the "Baidu Map" product is established, and the keywords in the dictionary may include "release", "release conference", "release schedule", and "release date".

140. When the matching is successful, the content of the message that matches the success is captured, or the content of the message that matches the success and the related content of the message content that is successfully matched are captured as the feature description information, and the feature description information is saved.

In this operation, there may be two implementation manners, one is to capture the message content of the matching success as the feature description information and save the feature description information when the matching is successful; the other is that when the matching is successful, The content of the matching successful message and the related content of the successfully matched message content are captured as feature description information, and the feature description information is saved.

Preferably, the content of the message matching the success and the related content of the message content that is successfully matched are captured as the feature description information, and the feature description information is saved, compared to only capturing the message content that is successfully matched. The preferred manner facilitates obtaining complete characterization information for the object.

You can set the crawl interval and/or the number of crawls to capture the relevant content of the message content that matches the success. For example, set the crawl interval to 15s and the number of grabs to 5.

Further, the related content of the successfully matched message content may include: the context message matching the successful message content; and/or establishing a session with the user who issues the message content and sending the message content to the user The supplementary content returned by the user after the request is supplemented.

Example 1

Take the object as a "Baidu Browser" product as an example. The release of a group of the product The content contains a lot of comments and problem discussions about the product, for example: a designer of the product posted a message in the development group "When logging into Baidu browser, there is a problem with the login permissions", then another designer of the product is The development group publishes the message "Really, the reason is A". After the matching operation, the "release login has a problem when logging in to the Baidu browser" message matches the keyword "problem" in the feature recognition dictionary. Succeeded, by grabbing the message content "When logging in to Baidu browser, there is a problem with the login permission", the feature description information corresponding to the defect of the product can be obtained, and the context message of the content of the message is "acquired, the reason is A", The feature description information corresponding to the cause of the defect in the product can be obtained, thereby enriching the feature description information of the product.

It should be noted that the feature description information corresponding to the defect corresponding to the defect of the product and the feature description information corresponding to the cause of the defect are described as an example, in addition to the feature description information corresponding to the cause of the defect, Other feature description information such as the solution corresponding to the defect may be captured as the complete information of the product defect, and formatted (for example, [product name, defect content, cause of occurrence]) is stored, which is not limited in this embodiment.

Example 2

After establishing a session with the user who publishes the content of the message, and taking a heuristic question to send a message content supplement request to the user to request a complete description of the product defect, and at this time, the session may be crawled. That is, for a case where the defect description dimension is large (for example, the defect type, the cause of the defect, etc.), a long crawl time (for example, one minute) is set, and during this time, the supplementary content returned by the user is captured. If there is still no supplementary description within this time, only the basic information is recorded, or the failure is returned because the necessary information is incomplete.

The technical solution of the embodiment, by monitoring and parsing the message published in the instant messaging software application, because the message published in the instant messaging software application is not only high in class definition, but also high in information professionality, so the content of the message to be parsed is The keywords in the established feature recognition dictionary are matched, and the content of the message matching the success is captured, or the content of the message matching the success and the related content of the message content are captured, and the feature description information of the specific object can be automatically captured, thereby saving manpower. Cost, and improve the professionalism and accuracy of the obtained feature description information of a specific object, and facilitate improvement of a specific object according to the feature description information.

In this embodiment, the establishing the feature recognition dictionary may specifically include:

Receiving keywords in a manually configured feature recognition dictionary; or,

A typical sentence of the manual inclusion is searched in the chat history of the instant messaging software, and keywords corresponding to the corresponding features are mined and added in the feature recognition dictionary according to the context co-occurrence relationship of the typical sentence.

In other words, each keyword in the feature recognition dictionary can be manually configured, for example, keywords such as "problem", "defect" or "improvement" are configured in the feature recognition dictionary.

It is also possible to manually include some typical sentences, and according to the context co-occurrence relationship of the typical sentences in the chat history, the words expressing the features in the typical sentences satisfying a certain co-occurrence frequency are used as keywords and added to the feature recognition dictionary; Or dig out a semantic template that expresses features.

For example, in Baidu Hi's "Baidu Browser" product research and development group, one person said "search type = xxx, the picture is wrong, who sees", another person replies "yes, it is a problem, has been recorded Defect", if there are multiple pairs of "map error" and "recorded defect" in the group message, it is considered that there is a co-occurrence relationship between the two sentences, indicating that this is a defect that needs to be recorded, based on In this way, the semantic template "[arbitrary word] map error" expressing the defect can be unearthed.

The information mining method provided in this embodiment can be applied to various scenarios. For example, according to the feature recognition dictionary corresponding to the established product defect, the object is obtained as defect description information of the product; for example, according to the established product debugging corresponding feature recognition. The dictionary obtains the description information of the debugging problem of the product. For example, according to the feature recognition dictionary corresponding to the established enterprise management, the description information such as the management opinion collection of the enterprise management event is obtained, which is not limited in this embodiment.

Specifically, when the defect description information for capturing an object is a product, the keyword in the feature recognition dictionary includes a keyword reflecting a defect of the product, and the feature description information is information describing a defect of the product. The present embodiment provides a fully automated implementation method from product mining to defect-related content capture and finally to a designated space, which can cover an important product group of all product lines of the enterprise.

Embodiment 2

FIG. 2 is a flowchart of an information mining method according to Embodiment 2 of the present invention. This embodiment provides a preferred solution before listening to messages published in an instant messaging software application on the basis of the above embodiments. The preferred method includes operations 210 through 220.

210. After obtaining access rights of the server corresponding to the instant messaging software application, establish a connection with the server.

For example, the access right of the server corresponding to the instant messaging software application "Baidu Hi" is obtained, and a connection is established with the server.

220. Send a request for joining a group account or a personal user account in the instant messaging software application to the server.

For example, the server corresponding to the instant messaging software application “Baidu Hi” sends a join request of the group account “Baidu Browser-R&D Group”, so that the newly joined group members can publish the product “Baidu Browser” in the group. "Related news.

For example, the server corresponding to the instant messaging software application “Baidu Hi” sends a personal user account joining request, and the newly added personal account can chat with the same personal account that has joined the application to form a published message; newly added A personal account can apply to join a group account that has joined the application, so that the newly joined group member posts a message in the group.

In the technical solution of the embodiment, before monitoring the message published in the instant messaging software application, establishing a connection with the server corresponding to the instant messaging software application, and interacting with the account joining request, so that the account added to the instant messaging software application can be The app posts the message.

It should be noted that after the request for the group account or the personal user account in the instant messaging software application is sent to the server, the message posted in the instant messaging software application may be monitored, which may include: receiving the After the response message returned by the server agrees to join, the user in the joined group or the message posted by the joined individual user is monitored.

Embodiment 3

Please refer to FIG. 3a, which is a flowchart of an information mining method according to Embodiment 3 of the present invention. On the basis of the above embodiments, the present embodiment provides a message content that captures a successful match, or captures the content of the message that is successfully matched and the related content of the message content that is successfully matched as the feature description information. The preferred description of the feature description information before saving.

The preferred method includes operations 310 through 360.

310. Monitor messages published in the instant messaging software application.

320. Parse the intercepted message to obtain the content of the message.

330. Match the content of the message with a keyword in a pre-established feature recognition dictionary.

340. When the matching is successful, the content of the message that matches the success is captured, or the content of the message that matches the success and the related content of the message content that is successfully matched are captured as the feature description information.

350. Match the feature description information with a keyword in a pre-established category recognition dictionary, and determine a category corresponding to the feature description information according to the matching result.

As described above, the information mining method provided by the embodiment of the present invention can be applied to various scenarios, and therefore, a category recognition dictionary including a plurality of application requirements can be established according to actual application requirements.

Keywords in the category recognition dictionary can be manually configured. The keyword in the category recognition dictionary may include: Baidu map development defect, Baidu browser debugging defect, and Baidu translation development improvement, etc., which is not limited in this embodiment.

360. Associate the determined category with the feature description information.

The technical solution of the embodiment, by monitoring and parsing the message published in the instant messaging software application, because the message published in the instant messaging software application is not only high in class definition, but also high in information professionality, so the content of the message to be parsed is The keywords in the established feature recognition dictionary are matched, and the content of the message matching the success is captured, or the content of the message matching the success and the related content of the message content are captured, and the feature description information of the specific object can be automatically captured, thereby saving manpower. Cost, and improve the professionalism and accuracy of the obtained feature description information of the specific object, and facilitate improvement of the specific object according to the feature description information; after grasping the feature description information of the object, by determining the feature Describe the category corresponding to the information, and associate the determined category with the feature description information, which is beneficial to the responsible group corresponding to the binding category, thereby enabling the corresponding responsible group to know the object in time according to the professional feature description information of the specific object. Valuable feedback.

It should be noted that the operation 350 is only one of the implementation manners for determining the category corresponding to the feature description information, and determining the category corresponding to the feature description information may also be: determining the feature by a Natural Language Processing (NLP) model. The category corresponding to the description information (operation 351 shown in Figure 3b).

Specifically, the semantic similarity algorithm model and/or the click similarity algorithm model may be used to determine the category corresponding to the feature description information.

Among them, the semantic similarity uses the supervised method training model of natural language processing cloud background training to analyze the similarity of two pieces of text. The larger the value, the more similar. Networking of semantic similarities provides the ability to compute similarities. For example, input "laptop", "notebook" semantic similarity is 2.08478.

Among them, the click similarity can be used when the semantic similarity cannot reach the threshold (such as 1.8), analyze the click similarity of the two texts (such as the title in the search and search results), and calculate the cosine similarity using the trained embedding vector. Degree value, the range of values [-1, 1], the greater the value, the stronger the click similarity. For example, the click similarity between "Baidu Hello" and "Zhou Hongyi Hello" is -0.121407. The click similarity between "Baidu Hello" and "Li Yanhong Hello" is 0.218664; the latter clicks the similarity ratio The former is high.

In actual use, the feature description information is preferentially compared with the preset plurality of categories, and the semantic similarity reaches the threshold and the highest category is returned. If the semantic similarity between the feature description information and the preset category does not reach the threshold, then The feature description information is further determined by clicking similarity with the preset category, and if the click similarity reaches the threshold, the corresponding category is returned, and if the click similarity does not reach the threshold, the default category (eg, other) is returned. Thresholds are constantly fitted to historical data to maintain a higher degree of accuracy.

It should be further noted that determining the category corresponding to the feature description information may further be: determining, by using a probability model trained according to the feature description text of the labeled category information in advance, a category corresponding to the feature description information, where the input of the probability model is The characterization text is output as a probability value belonging to the set category (operation 352 as shown in Figure 3c). Specifically, the probability model is pre-trained according to the feature description text of the labeled category information, and the feature description information is input into the probability model, and the category A corresponding to the feature description information output by the probability model and the corresponding category A are obtained. The probability value, if the probability value satisfies a certain threshold, determines that the category corresponding to the feature description information is category A. For example, a probabilistic model of P (type | feature description information) can be trained through manual classification annotation and corresponding description text in the chat record, and the training method can be flexibly selected according to the characteristics of the service domain of the system, and a typical naive Bayesian method is used. . In the application, if the probability that the user problem description belongs to a certain problem classification meets a certain threshold, it can be considered as belonging to the classification.

On the basis of the embodiment, after determining the category corresponding to the feature description information, the following operations may also be included:

Determining information of a recipient of the feature description information according to the category;

And transmitting the feature description information to the receiver according to the information of the receiver.

The information of the receiving party may be an address of setting a website, setting a short message number of the receiving user, an email address, or setting an instant messaging software account of the receiving user.

In this embodiment, after the feature description information of the object is captured and the category corresponding to the feature description information is determined, the receiver is notified of the feature description information of the object, and the receiver is regarded as the responsible group corresponding to the category. And the professional feature description information of the object interacting with the responsible group, so that the corresponding responsible group can timely obtain the valuable feedback of the object according to the professional feature description information of the specific object.

Embodiment 4

FIG. 4 is a schematic structural diagram of an information mining apparatus according to Embodiment 4 of the present invention. The device includes a message monitoring module 410, a message parsing module 420, a matching module 430, and a feature description information processing module 440.

The message monitoring module 410 is configured to listen to the message published in the instant messaging software application; the message parsing module 420 is configured to parse the monitored message to obtain the message content; the matching module 430 is configured to use the message content and the pre-established The keyword in the feature recognition dictionary is matched; the feature description information processing module 440 is configured to capture the message content that is successfully matched when the matching is successful, or capture the content of the message that matches the success and the related content of the message content that is successfully matched. The feature description information is saved as the feature description information.

In the above solution, the apparatus may further include: a connection establishment module and a request transmission module.

The connection establishing module is configured to establish a connection with the server after obtaining the access right of the server corresponding to the instant messaging software application before listening to the message published in the instant messaging software application; and requesting the sending module to use the The server sends a request to join the group account or the personal user account in the instant messaging software application; the message monitoring module 410 is specifically configured to: after receiving the response message returned by the server and agree to join, Messages posted by users in the group or joined by individual users.

In the above solution, the device may further include a feature recognition dictionary establishing module, configured to receive a keyword in the manually configured feature recognition dictionary; or

And searching for a typical sentence of the manual inclusion in the chat history of the instant messaging software, and excerpting the keyword expressing the corresponding feature according to the context co-occurrence relationship of the typical sentence and adding it to the feature recognition dictionary.

In the above solution, the apparatus may further include: a first category determining module, or a second category determining module, or a third category determining module.

The first category determining module is configured to: after capturing the successfully matched message content, or after capturing the successfully matched message content and the related content of the successfully matched message content as the feature description information, performing the feature description information. Before the saving, the feature description information is matched with the keywords in the pre-established category recognition dictionary, and the category corresponding to the feature description information is determined according to the matching result; the second category determining module is configured to capture the message that the matching is successful. Determining the feature description by a natural language processing (NLP) model after the content, or the content of the successfully matched message content and the content of the matching successful message content are captured as feature description information, and the feature description information is saved. a category corresponding to the information; the third category determining module is configured to: after the content of the message that matches the success is captured, or the content of the message that matches the successful match and the related content of the message content that is successfully matched are used as the feature description information, the feature is Before the description is saved, the pre-according is used Feature description text information is not trained to determine the probability model corresponding to the category feature descriptor; the feature of the information processing module 440 may be specifically configured to: determine a category of the features described in the associated information stored.

The second category determining module is specifically configured to: determine a category corresponding to the feature description information by using a semantic similarity algorithm model and/or a click similarity algorithm model.

Further, the apparatus may further include: a receiver information determining module and a feature description information sending module.

The receiver information determining module is configured to determine information of the receiver of the feature description information according to the category after determining the category corresponding to the feature description information; and the feature description information sending module is configured to be used according to the receiver The information transmits the feature description information to the recipient.

The related content of the successfully matched message content may include: the context message matching the successful message content; and/or, after establishing a session with the user who issues the message content and sending a message content supplement request to the user , the supplemental content returned by the user.

In the above solution, the keyword in the feature recognition dictionary may include a keyword reflecting a defect of the product, and correspondingly, the feature description information may be information describing a defect of the product.

The information mining device provided by the embodiment of the present invention can execute the information mining method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

Embodiment 5

The embodiment provides a non-volatile computer storage medium storing one or more modules, when the one or more modules are executed by a device performing an information mining method, causing the device Do the following:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

When the module stored in the foregoing storage medium is executed by the device, before monitoring the message published in the instant messaging software application, the method may further include:

After obtaining the access right of the server corresponding to the instant messaging software application, and the service Establish a connection;

Sending, to the server, a join request for a group account or a personal user account in the instant messaging software application;

The message that is sent in the monitoring instant messaging software application may specifically include:

After receiving the response message of the consent to join returned by the server, the user in the joined group or the message posted by the joined individual user is monitored.

When the module stored in the foregoing storage medium is executed by the device, the feature recognition dictionary may be established, which may specifically include:

Receiving keywords in a manually configured feature recognition dictionary; or,

When the module stored in the storage medium is executed by the device, after capturing the content of the message that is successfully matched, or capturing the content of the message that is successfully matched and the content of the message that is successfully matched as the feature description information, Before the feature description information is saved, the method may further include:

Matching the feature description information with a keyword in a pre-established category recognition dictionary, and determining a category corresponding to the feature description information according to the matching result; or determining a category corresponding to the feature description information by using a natural language processing NLP model Or determining a category corresponding to the feature description information by using a probability model trained in advance according to the feature description text of the tagged category information;

Saving the feature description information may include: associating the determined category with the feature description information.

When the module stored in the foregoing storage medium is executed by the device, determining a category corresponding to the feature description information by using a natural language processing NLP model may specifically include:

The semantic similarity algorithm model and/or the click similarity algorithm model are used to determine the category corresponding to the feature description information.

When the module stored in the foregoing storage medium is executed by the device, after determining the category corresponding to the feature description information, the method may further include:

When the module stored in the storage medium is executed by the device, the information of the receiver may be an address of setting a website, setting a short message number of the receiving user, an email address, or setting an instant messaging software account of the receiving user.

When the module stored in the storage medium is executed by the device, the related content of the message content that is successfully matched may include: the context message that matches the successful message content; and/or, in the content of the message content The supplemental content returned by the user after the user establishes a session and sends a message content supplement request to the user.

When the module stored in the storage medium is executed by the device, the keyword in the feature recognition dictionary may include a keyword reflecting a defect of the product, and the feature description information is information describing a defect of the product.

Embodiment 6

The device includes:

One or more processors 510, one processor 510 is taken as an example in FIG. 5;

Memory 520; and one or more modules.

The device may also include an input device 530 and an output device 540. The processor 510, the memory 520, the input device 530, and the output device 540 in the device may be connected by a bus or other means, and the bus connection is taken as an example in FIG.

The memory 520 is used as a computer readable storage medium for storing software programs, computer executable programs, and modules, such as the program instructions/modules corresponding to the information mining method in the embodiment of the present invention (for example, the information shown in FIG. 4). The message monitoring module 410 in the mining device, the message parsing module 420, the matching module 430, and the feature description information processing module 440). The processor 510 executes various functional applications and data processing of the server by executing software programs, instructions, and modules stored in the memory 520, that is, implementing the information mining method in the foregoing method embodiments.

The memory 520 can include a storage program area and a storage data area, wherein the storage program area can store operations The system, at least one function required application; the storage data area can store data created according to the use of the terminal device, and the like. Further, the memory 520 may include a high speed random access memory, and may also include a nonvolatile memory such as at least one magnetic disk storage device, flash memory device, or other nonvolatile solid state storage device. In some examples, memory 520 can further include memory remotely located relative to processor 510, which can be connected to the terminal device over a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

Input device 530 can be used to receive input digital or character information and to generate key signal inputs related to user settings and function control of the terminal. The output device 540 can include a display device such as a display screen.

The one or more modules are stored in the memory 520, and when executed by the one or more processors 510, perform the following operations:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

Further, before listening to the message published in the instant messaging software application, the method may further include:

After obtaining access rights of the server corresponding to the instant messaging software application, establishing a connection with the server;

The message that is sent in the monitoring instant messaging software application specifically includes:

Further, the establishing the feature recognition dictionary may specifically include:

Receiving keywords in a manually configured feature recognition dictionary; or,

Finding a typical sentence of the manual inclusion in the chat history record of the instant communication software, and excavating a keyword expressing the corresponding feature according to the context co-occurrence relationship of the typical sentence and adding the feature recognition In the dictionary.

Further, after the matching of the successfully matched message content or the matching of the successfully matched message content and the related content of the successfully matched message content as the feature description information, before the feature description information is saved, the method may further include :

Further, determining the category corresponding to the feature description information by using the natural language processing NLP model may specifically include:

Further, after determining the category corresponding to the feature description information, the method further includes:

Further, the information of the receiving party may be an address of setting a website, setting a short message number of the receiving user, an email address, or setting an instant messaging software account of the receiving user.

Further, the keywords in the feature recognition dictionary may include keywords reflecting product defects, and the feature description information is information describing product defects.

Through the above description of the embodiments, those skilled in the art can clearly understand that the present invention can be implemented by software and necessary general hardware, and can also be implemented by hardware, but in many cases, the former is a better implementation. . Based on such understanding, the technical solution of the present invention, which is essential or contributes to the prior art, can be embodied in the form of a software product. The software product can be stored in a computer readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory (FLASH), hard disk or optical disk, etc. A number of instructions are included to cause a computer device (which may be a personal computer, server, or network device, etc.) to perform the methods described in various embodiments of the present invention.

It should be noted that, in the embodiment of the information mining apparatus, the units and modules included in the information mining apparatus are only divided according to functional logic, but are not limited to the foregoing division, as long as the corresponding functions can be implemented; The specific names of the functional units are also for convenience of distinguishing from each other and are not intended to limit the scope of the present invention.

The above is only a specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present disclosure. All should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the appended claims.

Claims

An information mining method, comprising:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

Matching the message content with keywords in a pre-established feature recognition dictionary;

When the matching is successful, the content of the message that matches the success is captured, or the content of the message that matches the success and the related content of the message content that is successfully matched are captured as the feature description information, and the feature description information is saved.
The method of claim 1 further comprising: before listening to the message published in the instant messaging software application, further comprising:

After obtaining access rights of the server corresponding to the instant messaging software application, establishing a connection with the server;

Sending, to the server, a join request for a group account or a personal user account in the instant messaging software application;

The message that is sent in the monitoring instant messaging software application specifically includes:

After receiving the response message of the consent to join returned by the server, the user in the joined group or the message posted by the joined individual user is monitored.
The method according to claim 1 or 2, wherein the establishing the feature recognition dictionary comprises:

Receiving keywords in a manually configured feature recognition dictionary; or,

A typical sentence of the manual inclusion is searched in the chat history of the instant messaging software, and keywords corresponding to the corresponding features are mined and added in the feature recognition dictionary according to the context co-occurrence relationship of the typical sentence.
The method according to any one of claims 1 to 3, characterized in that, after the content of the message matching the success is captured, or the content of the message matching the success and the content of the message content of the matching success are captured as the feature description information Before saving the feature description information, the method further includes:

Matching the feature description information with a keyword in a pre-established category recognition dictionary, and determining a category corresponding to the feature description information according to the matching result; or determining a category corresponding to the feature description information by using a natural language processing NLP model Or, using a pre-specified information based on the category information The probability model trained by the characterization text determines a category corresponding to the characterization information;

Saving the feature description information includes: associating the determined category with the feature description information.
The method according to claim 4, wherein the determining the category corresponding to the feature description information by the natural language processing NLP model comprises:

The semantic similarity algorithm model and/or the click similarity algorithm model are used to determine the category corresponding to the feature description information.
The method according to claim 4 or 5, further comprising: after determining the category corresponding to the feature description information, further comprising:

Determining information of a recipient of the feature description information according to the category;

And transmitting the feature description information to the receiver according to the information of the receiver.
The method according to claim 6, wherein the information of the receiving party is an address of setting a website, setting a short message number of the receiving user, an email address, or setting an instant messaging software account of the receiving user.
The method according to any one of claims 1 to 7, wherein the related content of the successfully matched message content comprises: the context message matching the successful message content; and/or, in publishing the message The supplemental content returned by the user after the user of the content establishes a session and sends a message content supplement request to the user.
The method according to any one of claims 1 to 8, wherein the keywords in the feature recognition dictionary contain keywords reflecting product defects, and the feature description information is information describing product defects.
An information mining device, comprising:

a message monitoring module for monitoring messages published in an instant messaging software application;

a message parsing module, configured to parse the intercepted message to obtain a message content;

a matching module, configured to match the content of the message with a keyword in a pre-established feature recognition dictionary;

a feature description information processing module, configured to: when the matching is successful, capture the content of the message that matches the success, or capture the content of the message that is successfully matched and the related content of the message content that is successfully matched as a feature Describe the information and save the feature description information.
The device of claim 10, wherein the device further comprises:

a connection establishing module, configured to establish a connection with the server after acquiring the access right of the server corresponding to the instant messaging software application before listening to the message published in the instant messaging software application;

a request sending module, configured to send, to the server, a join request for a group account or a personal user account in the instant messaging software application;

The message monitoring module is specifically configured to: after receiving the response message that the server returns to join, to listen to the message posted by the user in the joined group or the joined individual user.
The device according to claim 10 or 11, wherein the device further comprises a feature recognition dictionary establishing module, configured to receive keywords in the manually configured feature recognition dictionary; or

And searching for a typical sentence of the manual inclusion in the chat history of the instant messaging software, and excerpting the keyword expressing the corresponding feature according to the context co-occurrence relationship of the typical sentence and adding it to the feature recognition dictionary.
The device of any of claims 10-12, wherein the device further comprises:

a first category determining module, configured to save the feature description information after capturing the successfully matched message content, or after capturing the successfully matched message content and the related content of the successfully matched message content as the feature description information Having previously matched the feature description information with a keyword in a pre-established category recognition dictionary, and determining a category corresponding to the feature description information according to the matching result; or

a second category determining module, configured to save the feature description information after the content of the matching successful message content or the content of the successfully matched message content and the content of the successfully matched message content are captured as the feature description information Previously, determining a category corresponding to the feature description information by a natural language processing NLP model; or

a third category determining module, configured to save the feature description information after the content of the successfully matched message content is captured, or the content of the successfully matched message content and the content of the successfully matched message content are captured as the feature description information Previously, determining a category corresponding to the feature description information by using a probability model trained in advance according to the feature description text of the tagged category information;

The feature description information processing module is specifically configured to: associate the determined category with the feature description information.
The apparatus according to claim 13, wherein the second category determining module is specifically configured to: determine a category corresponding to the feature description information by using a semantic similarity algorithm model and/or a click similarity algorithm model.
The device according to claim 13 or 14, wherein the device further comprises:

a receiver information determining module, configured to determine information of a recipient of the feature description information according to the category after determining a category corresponding to the feature description information;

The feature description information sending module is configured to send the feature description information to the receiver according to the information of the receiver.
The device according to claim 15, wherein the information of the receiving party is an address of setting a website, setting a short message number of the receiving user, an email address, or setting an instant messaging software account of the receiving user.
The device according to any one of claims 10-16, wherein the related content of the successfully matched message content comprises: the context message matching the successful message content; and/or, in publishing the message The supplemental content returned by the user after the user of the content establishes a session and sends a message content supplement request to the user.
The apparatus according to any one of claims 10-17, wherein the keyword in the feature recognition dictionary contains a keyword reflecting a defect of the product, and the feature description information is information describing a defect of the product.
A non-volatile computer storage medium storing one or more modules, wherein when the one or more modules are executed by a device performing an information mining method, the device is caused Do the following:

Listening for messages posted in instant messaging software applications;

Parsing the intercepted message to get the message content;

Matching the message content with keywords in a pre-established feature recognition dictionary;

When the matching is successful, the content of the message that matches the success is captured, or the content of the message that matches the success and the related content of the message content that is successfully matched are captured as feature description information, and the feature description is described. The information is saved.