TW201918901A - Topic providing apparatus and could file prompting method thereof - Google Patents

Abstract

A topic providing apparatus and a cloud file prompting method thereof are provided. In the method, an interactive text content is obtained, the interactive text content is converted from multimedia conversation content of at least two users. Key vocabulary is obtained from the interactive text content. Multiple topics recorded in a file database is screened according to the key vocabulary. Then, content (such as cloud file, teaching material, etc.) the screened topic can be provided to the user. Accordingly, multiple field can be implemented, and the users in discussion can obtain extra reference data.

Description

Theme providing device and cloud storage file prompting method thereof

The invention relates to the field of semantic analysis of machine learning and artificial intelligence technology, and in particular to a theme providing device and a cloud storage file prompting method thereof.

In recent years, one of the most popular technologies in the field of artificial intelligence, all major electronics companies have invested a lot of manpower and funds to research and development. Among them, natural language processing is an important research project in artificial intelligence technology, which mainly discusses how to deal with and use natural language, and let computers understand human language. In the past, people might need to enter a preset number of keywords through an input device such as a keyboard or a mouse to provide a response. With natural language processing technology, computers can understand the normal conversations of human beings, enabling the public to interact with smart products (eg, smart phones, robots, smart butlers, etc.) in a simpler and more convenient way. In addition, another popular research project in artificial intelligence technology is machine learning, which mainly enables computers to learn automatically from input materials. Machine learning has been widely used in data exploration, natural language processing, biometrics, search engines, etc.

On the other hand, with the rapid development of the Internet, it is almost impossible for modern people to get out of the online world. However, the functions of electronic products used by general users are limited, and there may even be problems such as insufficient performance. In order to solve the above problems, many operators will provide a cloud processing service, and the part that requires arithmetic processing is executed by the server, and the user only needs to issue a request through a browser or an application.

In view of this, the present invention provides a theme providing apparatus and a cloud storage file prompting method thereof, which provide an archive of a suitable theme, a teaching material, and the like according to an interactive conversation content.

The cloud storage file prompting method of the present invention comprises the following steps. Acquire interactive text content, which is converted from the multimedia conversation content of at least two users. Get keyword collections from this interactive text content. According to these keywords, several topics recorded in the archive database are filtered. Provide content for filtering topics.

The subject matter of the present invention provides an apparatus including a communication unit, a storage unit, and a processing unit. The communication unit transmits or receives data. The storage unit records a number of modules and archives, which store the contents of several topics. The processing unit is coupled to the communication unit and the storage unit, and accesses and loads those modules recorded by the storage unit. Those modules include a message exchange module, an interactive vocabulary extraction module, a topic analysis module, and a theme providing module. The message exchange module obtains multimedia conversation content of at least two users through the communication unit, and converts the multimedia conversation content of these users into interactive text content. The interactive vocabulary extraction module obtains key words from the interactive text content. The topic analysis module filters those topics based on these keywords. The theme provides modules to provide content for filtering topics through the communication unit.

Based on the above, the embodiment of the present invention is based on a machine operation method, and the subject of the online interactive text content (including the voice-to-text voice discussion) of two or more people is marked with the attention word to automatically prompt or recommend the corresponding topic in the cloud storage space. Archives or textbooks to achieve intelligentization of the file system, and to enhance the use of cloud storage file systems, and to enhance knowledge learning among people and stimulate economic development.

The above described features and advantages of the invention will be apparent from the following description.

1 is a block diagram of components of a subject providing apparatus 1 in accordance with an embodiment of the present invention. The theme providing device 1 can be an electronic device such as a server, a computer host, a workstation, and the like, and includes at least but not limited to the communication unit 110, the storage unit 120, and the processing unit 130.

The communication unit 110 can support a wired network technology such as an optical fiber or an Ethernet network, or a communication transceiver of a wireless network such as a Wi-Fi, a mobile communication network, or a WiMAX, which can receive other user equipment (for example, a computer). , smart phones, tablets, etc.) messages, chat content, multimedia content and other files or materials, and can send content of various topics (for example, files, textbooks, etc.) to the corresponding user equipment.

The storage unit 120 can be any type of fixed or removable random access memory (RAM), read only memory (ROM), flash memory, hard disk drive, solid state hard A solid-state drive or the like or a combination of the above components, and used to record the message exchange module 121, the interactive vocabulary extraction module 122, the archive database 123, the database vocabulary extraction module 124, and the topic analysis module 125 And the theme provides software programs such as module 126, theme model, stop word, keyword collection, database vocabulary and other related information. The foregoing modules, databases, files and materials will be described in detail in the following embodiments.

The processing unit 130 is connected to the storage unit 120 and the communication unit 110, and may be a central processing unit (CPU), or other programmable general purpose or special purpose microprocessor (Microprocessor), digital signal processor (DSP). , a programmable controller, an application specific integrated circuit (ASIC) or other similar component or a combination of the above. In the embodiment of the present invention, the processing unit 130 is configured to execute all the operations of the theme providing apparatus 1 and can access and execute the modules recorded in the storage unit 120.

In order to facilitate the understanding of the operation flow of the embodiment of the present invention, the operation flow of the theme providing apparatus 1 in the embodiment of the present invention will be described in detail below. 2 is a flow chart illustrating a cloud storage file prompting method according to an embodiment of the invention. Referring to FIG. 2, the method of the present embodiment is applicable to each device in the subject providing device 1 of FIG. Hereinafter, the methods described in the embodiments of the present invention will be described with reference to various components and modules of the subject providing apparatus 1. The various processes of the method can be adjusted accordingly according to the implementation situation, and are not limited thereto.

First, the theme providing device 1 obtains any type (for example, textbooks, gallery, news, etc.) by capturing a network packet, uploading a user, or using an external or internal storage medium (for example, a flash drive, a compact disc, an external hard drive, etc.). ) file-related content such as file content, file annotation content, file data attributes, or multimedia files such as audio/video files, and convert these files into text, voice, and other information content and store them in the file database 123. As a sample of training and learning.

On the other hand, the message exchange module 121 obtains multimedia conversation content of at least two users through the communication unit, and converts the multimedia conversation content of the users into interactive text content, thereby obtaining interactive text content (step S210). . Specifically, the message exchange module 121 can operate a chat room, a discussion area or other information exchange platform or an external message exchange platform (such as the chat room 310 in FIG. 3) to collect users (for example, teachers, students, customers, and operators). Etc.) The content of multimedia conversations such as plain text messages and video recordings.

Next, the database vocabulary extraction module 124 performs vocabulary extraction on the information content recorded by the archive database 123 and filters out Stop words to generate a database vocabulary. Specifically, the database vocabulary extraction module 124 analyzes the file content, the archive annotation content, the archive material attribute, the bookmark classification name, and the search keyword text content to capture the vocabulary of the user's attention. Since the file content, the archive data attribute and the file annotation content are mostly general sentences, the database vocabulary extraction module 124 can extract important words in the sentence through a suffix array or a PAT-Tree method. The vocabulary extracted by the word-extracting method first performs preliminary filtering using the predetermined rule conditions to delete the non-word-forming vocabulary. Then, the database vocabulary extraction module 124 uses the pre-collected list of Chinese and English stop words to further filter the extracted vocabulary and separate the extracted vocabulary by spaces, and the final vocabulary is used as the database vocabulary.

In addition, the interactive vocabulary extraction module 122 also obtains a keyword sink from the interactive text content (step S220). Similarly, the interactive vocabulary extraction module 122 performs vocabulary extraction on the interactive text content and filters out the stop words to generate a keyword sink. For detailed steps, reference may be made to the foregoing process of obtaining the vocabulary of the database, and details are not described herein again.

The topic analysis module 125 learns the database vocabulary through the topic model of the natural language processing to generate the theme implied by the information content. Specifically, the potential Dirichlet allocation (LDA, Latent Dirichlet Allocation), LSA (Latent Semantic Analysis), and PLSA (Probabilistic Latent Semantic Analysis) in the subject model domain are common model methods for natural language processing.

Please refer to Figure 3 for a schematic diagram of the conceptual mathematical function of the potential Dirichlet distribution. The mathematical meaning of the label is: : Dirichi distribution of the subject to which each document (ie, information content) belongs : Dirichi distribution of words within each subject : The subject probability distribution of document m : The probability distribution of the word k : The subject of the nth word of the document m : the nth word of the document m : total number of topics : total number of documents : Total number of words in the document The potential Dirichlet distribution model document generation steps are: From Dirichi distribution Mid-sampling generates the subject distribution of the document m . Polynomial distribution from the topic Sampling to generate the subject of the nth word of the document m . Distribution from Dirichi Sampling generation theme Word distribution . Polynomial distribution from words Mid-sampling final generated words .

For the training sample D (ie, information content), the data content similarity (likelihood) can be expressed as: (1) The subject analysis module 125 may be configured by a Maximum Likelihood Estimate training condition and a corresponding training method, such as Gibbs Sampling or Variational Inference. The model parameters α, β, θ, φ are obtained.

Please refer to FIG. 5 for an example of a document generation process. For an article, the topic analysis module 125 can use the theme probability distribution, for example, the distribution of the three themes represented by different background color histograms on the right side of the drawing and the vocabulary distribution of each theme, for example, the first theme includes a trigonometric function. Words such as "sinusoidal" and "cosine" have a probability of 0.05, 0.02, 0.02, etc., and so on, and then generate samples of different subject sequences according to the theme distribution, such as the first topic, the first topic, Two themes, the first theme, the second theme, the third theme, the first theme, and the like. Finally, through the topic vocabulary distribution, generate corresponding vocabulary (arrows pointing and labeling as shown in the figure). Through the training of the model, the topic analysis module 125 obtains the topic distribution, vocabulary distribution and subject sample order that best match the training materials.

Then, the topic analysis module 125 can filter the topics recorded by the archive database 123 according to the keyword of the interactive text content (step S230). In this embodiment, the topic analysis module 125 determines the similarity of the keywords to those topics, as a basis for screening. Specifically, the topic analysis module 125 analyzes the probability of the database vocabulary and keyword aggregation for different topics through the aforementioned topic model (eg, potential Dirichlet distribution) as a vector of probability values. In addition, since the interactive text content and the information content have been converted into the same subject space, the topic analysis module 125 can calculate the database vocabulary and the cosine similarity calculation formula (2) (assuming that A and B respectively represent two vectors). The similarity of the vectors of the two probability values corresponding to the keyword sink. …(2)

Taking Figure 6 as an example, the teacher-student interaction content (ie, interactive text content) and the cloud textbook (ie, information content) are input to the Dirichi distribution theme model, and the theme model is used to analyze the two input materials for the theme 1 to the theme n ( The probability vector of n is a positive integer). Finally, the topic analysis module 125 calculates the similarity for different topics through the formula (2) for the two sets of vectors.

Finally, the theme providing module 126 can provide the content of the screening theme to the corresponding user equipment through the communication unit 110 (step S240). The topic providing module 126 can select topics with similarities above a certain threshold (eg, 0.5, 0.8, 0.9, etc.), and obtain corresponding content of the selected topic from the archive database 123 (eg, textbooks, pictures, documents, List, etc.). As shown in FIG. 7, the screen 700 presented by the display of the user equipment is a teaching material provided by the theme providing apparatus 1.

It should be noted that the embodiment of the present invention can be applied in a relatively large number of scenarios. In addition to the provision of cloud files, information search, image information providing, educational learning, etc., the presentity providing device 1 can even serve as an expert system to provide users after the conversation. Additional auxiliary content is also available. For example, after the teacher and the student shown in FIG. 3 are talking in the chat room 310 or during the process, the theme providing device 1 can provide the cloud teaching material 350 of the appropriate theme to the student's device.

In summary, the embodiments of the present invention are mainly performed by the following processes: collecting information related to cloud electronic files (ie, information content). Collect interactive text content between multiple users. The cloud electronic file related information, vocabulary extraction and filter out stop words. The text will be discussed interactively, vocabulary extraction will be performed and the stop words will be filtered out. Use the theme model to learn the topics implied by the content. Calculate the subject to match similarity based on the probability of topic distribution. Finally, based on the similarity, the information content that meets the relevant topic is recommended to the user. This information content may also be applied to cloud textbooks in education, or to other legal documents, news chapters, pictures and other fields of information.

Although the present invention has been disclosed in the above embodiments, it is not intended to limit the present invention, and any one of ordinary skill in the art can make some changes and refinements without departing from the spirit and scope of the present invention. The scope of the invention is defined by the scope of the appended claims.

1‧‧‧ theme providing device

110‧‧‧Communication unit

120‧‧‧ storage unit

121‧‧‧Message Exchange Module

122‧‧‧Interactive vocabulary extraction module

123‧‧‧Archive database

124‧‧‧Database vocabulary extraction module

125‧‧‧Thematic Analysis Module

126‧‧‧ theme providing module

S210~S240‧‧‧Steps

310‧‧‧ Chat Rooms

350‧‧‧Cloud textbook

700‧‧‧presentation

α‧‧‧Dirichley distribution of the subject of each document

β‧‧‧Dirichley distribution of words in each subject

Θ‧‧‧ theme probability distribution

Φ‧‧‧ word probability distribution

The subject of Z‧‧‧

W‧‧‧ words

K‧‧‧ total number of topics

M‧‧ ‧ total number of documents

Total number of words in the N‧‧‧ document

1 is a block diagram of components of a subject providing apparatus in accordance with an embodiment of the present invention. 2 is a flow chart of a cloud storage file prompting method according to an embodiment of the invention. FIG. 3 is an illustration of the provision of a chat room and subject matter. Figure 4 is a schematic diagram of the conceptual mathematical function of the Latent Dirichlet Allocation (LDA). Figure 5 is an illustration of a document generation process. FIG. 6 is an example of a similarity operation flow. Figure 7 is an illustration of the subject matter provided.

Claims (11)

A cloud storage file prompting method, comprising: obtaining an interactive text content, wherein the interactive text content is obtained by converting at least two user multimedia conversation contents; obtaining at least one keyword sink from the interactive text content; The keywords are used to filter a plurality of topics recorded in an archive database; and to provide content for screening topics. The method for prompting a cloud storage file according to claim 1, wherein the step of obtaining the at least one keyword from the interactive text content comprises: performing vocabulary extraction on the interactive text content and filtering out the stop word, Generating the at least one keyword sink. The method for prompting the cloud storage file according to claim 1, wherein the step of filtering the topics recorded by the archive database according to the at least one keyword includes: determining that the at least one keyword meets the topics Similarity, as a basis for screening. The method for prompting the cloud storage file according to claim 3, wherein before the step of determining the similarity between the at least one keyword and the topics, the method further comprises: performing vocabulary extraction on the information content recorded in the archive database And deleting the stop words to generate at least one database vocabulary; and learning the at least one database vocabulary through the natural language processing related topic model to generate the topics. The method for prompting the cloud storage file according to claim 4, wherein the step of determining the similarity between the at least one keyword and the topics comprises: analyzing the at least one database vocabulary and the at least one keyword The probability of different topics, as a vector of multiple probability values; and calculating the similarity of the at least one database vocabulary and the vectors of the probability values of the at least one keyword sink. For example, the cloud storage file prompting method described in claim 4, wherein the information content recorded by the archive database is a cloud textbook. A subject providing device, comprising: a communication unit for transmitting or receiving data; a storage unit for recording a plurality of modules and a file database, wherein the file database stores contents of a plurality of topics; and a processing unit coupled to the The communication unit and the storage unit access and load the modules recorded by the storage unit, and the modules include: a message exchange module, through which the multimedia conversation content of at least two users is obtained And converting the multimedia conversation content of the at least two users into an interactive text content; an interactive vocabulary extraction module, obtaining at least one keyword sink from the interactive text content; a theme analysis module, filtering according to the at least one keyword sink Themes; and a theme providing a module through which the content of the screening theme is provided. The theme providing device of claim 7, wherein the interactive vocabulary extracting module performs vocabulary extraction on the interactive text content and filters out the stop words to generate the at least one keyword sink. The subject providing device of claim 7, wherein the subject analysis module determines the similarity between the at least one keyword and the topics as a screening basis. The theme providing device of claim 9, wherein the file database further stores information content, and the modules further comprise: a database vocabulary extraction module, and the information content recorded in the file database is The vocabulary extracts and filters out the stop words to generate at least one database vocabulary; and the topic analysis module learns the at least one database vocabulary through a natural language processing related topic model to generate the topics. The theme providing apparatus according to claim 9, wherein the subject analysis module analyzes the probability of the at least one database vocabulary and the at least one keyword to different topics, as a vector of a plurality of probability values, and the The subject analysis module calculates a similarity between the at least one database vocabulary and the vectors of the probability values of the at least one keyword sink.