TWI453684B - An Evaluation System and Method of Intelligent Mobile Service Commodity Application Information Retrieval Technology - Google Patents

Description

Intelligent mobile service product evaluation system and method using information retrieval technology

The system and method, which combines the real-time commodity management system, the commodity review suggestion system and the commodity price comparison recommendation system to provide instant commodity evaluation services to users to recommend related products, and to learn each by using self-organizing neural networks. Product features, and effective classification, combined with multi-file automatic summary technology to help users to judge and obtain important purchase decision reference information in a short time.

In recent years, with the development and growth of Duhui's economy, consumer purchasing behavior has undergone significant changes in recent years. This change is mainly reflected in the decreasing ratio of consumers' planned purchases, rather than planned (sensible) purchases. The behavioral ratio has risen sharply. This kind of perceptual factor drives the purchase behavior called “Impulse Buying”. Due to the tight work and less amateur time, the shopping process of consumers may be extremely hasty, plus In the environment, such as personnel recommendation, product promotion, etc., it allows consumers to generate shopping behavior in an impulsive situation. According to the survey, more than 80% of "young women" are impulsive buyers, and "cosmetics" and "book magazines" are the main impulse purchases. Another study pointed out that 75% of consumers in the supermarket's shopping decision is decided within 15 seconds. Impulse-purchasing consumers have a very fast shopping decision-making process: however, buying in such a situation often results in higher money to purchase goods, and according to statistics, many have negative reactions after purchase, such as Complaining, regretting, etc.

Due to the popularity and convenience of the Internet, consumers are increasingly demanding to use the Internet to search for product reviews or comments. In the future, the application of product comparison information and recommendations will be more important as the mobile network expands.

The invention adopts Image Identification System (IIS), Self-Organizing Maps (SOM), Intelligent Agent (IA), and Multi-Document Automatic Summary Technology (Multiple Documents). Summarization (MDS) is integrated with the information system to provide instant commodity evaluation services, and proposes a Mobile Merchandise Evaluate Service Platform (MMESP), which includes a 3-Tiers architecture. Real Time Merchandise Identifying System (RMIS), Merchandise Evaluation System (MES) and Merchandise Recommendation System (MRS) help consumers judge and gain important results in a short time. Purchase decision reference information.

The main object of the present invention is to provide a smart mobile service product evaluation system and method using information retrieval technology, which comprises a Real-Time Merchandise Identifying System (RMIS) and a commodity review suggestion system (Merchandise Evaluating). System, MES), and Merchandise Recommandation System (MRS) to provide Merchandise Evaluate Service (MES) to recommend related products and to use self-organizing neural networks (Self-Organizing) Maps, SOM) learns the characteristics of various products and effectively classifies them, combined with Multiple Document Summarization (MDS) to help consumers judge and obtain important purchase decision reference information in a short time.

In order to further clarify the present invention, it will be helpful to review the review by the reviewer, the description of the drawings, and the detailed description of the invention.

Please refer to FIG. 1 . FIG. 1 is a structural diagram of a smart mobile service product evaluation system using the information retrieval technology of the present invention. FIG. 1 is a smart mobile service product evaluation using the information retrieval technology of the present invention. The system 01 includes at least an instant product identification system 02, a product review suggestion system 03, and a product price comparison recommendation system 04. Details are as follows.

Real Time Merchandise Identification System (RMIS), which is located at the front end of the Internet, mainly provides multimedia processing and image recognition related services, including instant multimedia transmission technology and real-time image recognition technology. When consumers have an impulse to buy, they can instantly take photos of the camera and transmit them to the server via J2ME (MIDP2.0) technology via mobile communication (3G or IEEE 802.11b). After receiving the image image, the instant product identification system combines the Region Adjacency Graph (RAG) and the self-organizing neural network (SOM) to process and classify the image for product identification. The process is shown in Figure 2. As shown in FIG. 2, the instant product identification system 01 includes an instant multimedia transmission module 22 and an instant image recognition module 23. The multimedia data transmission module 22 can transmit multimedia materials including text and pictures (Picture). The user can use the camera on the 3G mobile phone or PDA to instantly capture the product to be purchased, and immediately transmit the product image information back to the instant product identification and weaving system for image recognition processing, and provide the user with instant Convenient and fast product evaluation service. Real-time image recognition module 23 (RMIS) combines real-time block neighbor map (RAG) and self-organizing neural network (SOM) to perform real-time multimedia image processing, which firstly displays product images in real-time block neighboring. The picture (RAG) operation acquires the image features of each product, and then puts it into the self-group network (SOM) for training, so that the neural network can learn various product features. When an instant image is obtained, the image can be sorted in real time, and the product information of the image is recognized and obtained for subsequent comments and recommendations. Using the technology of Region Adjacency Graph (RAG), the image recognition search can be retrieved without being affected by rotation, enlargement or reduction, and partial subgraphs, and the desired similar images can be retrieved. Its main processes include:

(1) Convert to RGB (Red Green Blue) Matrix: Convert the input real-time image into RGB vector matrix representation for subsequent operations.

(2) Blurring: Blurring the concept of the graph using the Low Pass Filter.

(3) Color reduction: The blurred image is reduced to the RGB three primary colors respectively, for example, from 256 steps divided by 64 to 4 steps, so that the original 256 ³ = 16,777,216 kinds are reduced to 4 ³ = 64 kinds. The number of colors.

(4) Histogram: The number of pixels accumulated in the 64 colors after the reduction is recorded in the form of a bar graph to check whether there is a problem or error in the processing of the actual graphics.

(5) Block and adjacent relationship: Mark the same color and adjacent points in the graph as the same block.

(6) Block merging: In order not to let too many broken blocks affect the speed of program execution, the fragmentation block below a certain proportion in the graph is merged with the larger block around it.

(7) Region Adjacency Graph (RAG): After the above steps, the first K blocks with the largest area ratio are taken out, and the color, area ratio, shape, and number of adjacent blocks of the K blocks are stored. Form a two-dimensional array and store the neighboring relationships of the K blocks into a block neighbor map (RAG).

(8) Self-Organizing Maps (SOM): The different RAG features produced by each commodity are put into the self-organizing neural network (SOM) for training, through the operation of the SOM algorithm. You can learn the characteristics of each product and can effectively classify it.

(9) Similarity judgment: When a new image comes in, the similarity judgment can be performed with the classification after the SOM operation. The present invention calculates the similarity by the cosine theorem, and then finds similar products to provide subsequent evaluation services. .

The self-organizing neural network (SOM) was proposed by the Finnish Kohonen in 1981. It mimics the self-organizing feature mapping function of the human cranial nervous system and maps the input n-dimensional spatial data to a lower dimension. SOM is a kind of neural network based on "competitive learning" in which the neurons of the output layer compete with each other and are activated to the active state (Active State) or by the "lateral suppression coupling" between the output layer neurons. Inhibiting the Inactive State, the mobile service product evaluation platform of the present invention uses the SOM algorithm to perform image recognition learning, and the detailed steps of the calculation are the scope of the prior art, and will not be described herein.

As can be seen from FIG. 1 , the smart mobile service product evaluation system using the information retrieval technology of the present invention further includes a product review suggestion system, please refer to FIG. 3 , which can be seen from FIG. 3 that the product review suggestion system includes a blog content. The agent module 032 and a multi-file automatic summary module 033 are taken. The main functions of the blog content retrieval agent module 032 include Fuzzy Search, HTML Crawler, and HTML Parser, and the function descriptions are as follows; (i) Fuzzy search: the system provides fuzzy operations and Judging, the search keyword database is searched, and the keyword font content is actively searched for Google Blog Search. (ii) HTML Crawler: It crawls the results of the Google Blog Search search, such as the returned blog content, and traces the relevant links to the webpage and temporarily stores the HTML content. (iii) HTML Parser: It interprets the HTML of the HTML obtained by the HTML Crawler to obtain the main information, and effectively removes related special characters (such as single quotes and double quotes) and avoids attacks such as SQL Injection, and builds Blog Corpus. Follow-up to the automatic summary processing of multiple files.

The multi-file automatic summary module 033 automatically summarizes the comments of related products in each blog, effectively reduces the amount of information, extracts a summary of key comments, and allows the user to quickly browse the opinions of consumers who purchase the goods. Experience, multi-file automatic summary technology mainly uses the multi-file automatic summary module for system implementation, input the relevant product reviews in Blog Corpus into the multi-file automatic summary module, via (i) Preprocess, (ii) Feature Selection, ( Iii) Classifier, (iv) Reranker, (v)Summary, and (vi)Evaluation and other steps for automatic summary extraction, please refer to Figure 4, the detailed steps are as follows, step 41), Preprocess: the blog content is captured by the agent model The group retrieved the blog content (HTML) for word segmentation and clause processing to facilitate subsequent word weight calculation; step 42), Feature Selection: design different features for the weight calculation of the word clause, mainly Three features of Centrality, Sentence Length, and Position are used; Step 43), Classifier: Weight is used to calculate the score of each sentence; Step 44), Reranker: Since the Classifier only calculates and sorts the scores according to the similarity of the sentences, there may be a problem that the similarity is too high between the obtained sentences. Especially in the case of automatic summarization of multiple files, the multi-file automatic summary module has a design Reranker mechanism. Its main function is to recalculate the similarity between sentences and sentences, and set the threshold value for filtering, taking out important and Sentences whose similarity between them is not too high, and then extract according to the set compression ratio; Step 45), Summary: take the words in the original file according to the sentence order sorted by Reranker, and recombine the output summary; And step 46), Evaluation: measure the system's text summarization system performance, including the effect of the output results, and the user's satisfaction.

As described above, the intelligent mobile service product evaluation system using the information retrieval technology of the present invention further includes a commodity price comparison recommendation system. Referring to FIG. 5, it can be seen from FIG. 5 that the commodity price comparison recommendation system 051 of the present invention includes a commodity price information acquisition agent module 052 and a price comparison recommendation module 053, which are respectively described later.

After the identification of the real-time product identification system is completed, the product information will be transmitted to the commodity price recommendation system, and then the commodity price information acquisition agent module 052 searches for the commodity price related information to the comparison network and the related network. Look for the price information of the relevant products in the e-commerce website, and store the Crawl and Parse as the commodity price Corpus. Finally, the price information in the relevant commodity price Corpus is extracted through the price comparison recommendation module 053, and the comparison and the product recommendation are performed to effectively let the user know the price range of the commodity. The processing process of the commodity price information retrieval agent module in the commodity price comparison recommendation system includes (i) fuzzy search; fuzzy calculation and judgment, establishing a search-related keyword font library, and actively searching for the keyword database content and the related website. Search. (ii) HTML Crawl: The result of searching the comparison network and related websites, such as Crawl in each e-commerce website returned, tracking the related links and temporarily storing the HTML content. (iii) HTML Parser: HTML parsing of HTML obtained by HTML Crawl, obtaining main information, effectively removing relevant special characters (such as single quotes and double quotes) and avoiding attacks such as SQL Injection, establishing commodity price Corpus for profit Subsequent price comparison module processing.

The price comparison recommendation module compares the prices of related products in various e-commerce silk stations through the commodity price information acquisition agent module, and provides a Price-Oriented Recommendation method, so that consumers can quickly understand the relevant price range and obtain the lowest price. The equivalent of the relevant product information to decide whether to purchase the product.

The invention and the technical contents of the present invention will be further described in an actual embodiment. Please refer to Figure 6 first. As shown in Figure 6, when the user walks on the road and sees the purchase of the impulsive item 62, he can pick up the mobile phone 61 and take it with him; then, link to the "Mobile Service Product Evaluation Platform". 63. Upload the photo to the evaluation system (as shown in Figure 7). As shown in Figure 8, after receiving the image, the instant product identification system (RMIS) will automatically identify the product and immediately combine the adjacent blocks. Region (Region Adjacency Graph, RAG) and Self-Organizing Maps (SOM) technologies for image processing and classification, and product identification to find matching products, and to list relevant information of this product for users' reference. As shown in Figure 9, the Product Comment Recommendation System (MES) 91 automatically retrieves relevant comments and product usage information from Blog Corpus 92. The multi-file automatic summary technology 93 extracts the summary of each review information and presents it to the mobile phone. On the 94th, the commodity review suggestion system (MES) 101 automatically retrieves the commodity price information from each e-commerce website 102, and the price comparison recommendation mechanism 103 lists the price comparison information at different comparison networks. Kind of merchandise do not necessarily have the same price, the user can understand the price range and check prices through this information to measure whether the purchase of goods.

It can be seen from the above embodiments that the user can immediately transmit the product photo to the mobile service product evaluation platform of the present invention, and after the instant product identification system (RMIS) identification is completed, the product information will be transmitted to the product price comparison recommendation system (MRS). Then, through the commodity price information, the agent searches the comparison network and the relevant network for the price information of the commodity to search for the relevant commodity price information in each e-commerce website, and saves the Crawl and Parse as the commodity price Corpus. Finally, through the price comparison recommendation mechanism, the price information in the relevant commodity price Corpus is extracted, and the comparison and the product recommendation are performed to effectively let the user know the relevant price range of the commodity.

In summary, the structural features and embodiments of the present invention have been disclosed in detail, and the present invention can fully demonstrate the novelty and advancement of the invention in terms of purpose and efficacy, and is extremely valuable for industrial use. And for the unprecedented use on the market, according to the spirit of the patent law, the present invention fully meets the requirements of the invention patent. The above is only the preferred embodiment of the present invention, and the scope of the present invention is not limited thereto, that is, the equivalent variations and modifications made by the scope of the present invention should still belong to the present invention. Within the scope of the patent, I would like to ask your review committee to give a clear understanding and pray for it. It is the prayer.

01. . . Intelligent mobile service product evaluation system using information retrieval technology

02. . . Instant product identification system

03. . . Product review suggestion system

04. . . Commodity price comparison system

022. . . Instant multimedia transmission module

023. . . Instant image recognition module

032. . . Blog content capture agent module

033. . . Multi-file automatic summary module

041, 042, 043, 044, 045, 046. . . Process step

051. . . Commodity price comparison system

052. . . Commodity price information capture agent module

053. . . Price recommendation module

061. . . Mobile phone

062. . . commodity

063. . . Mobile service product evaluation platform

091. . . Product Review Advice System (MES)

092. . . Blog Corpus

093. . . Multi-file automatic summary technology

094. . . Mobile phone

101. . . Product review suggestion system

102. . . E-commerce website

103. . . Price recommendation mechanism

FIG. 1 is a schematic diagram of the architecture of a smart mobile service product evaluation system using the information retrieval technology of the present invention.

Figure 2 is a schematic diagram of the instant commodity identification system of the present invention.

Figure 3 is a schematic diagram showing the processing steps of the commodity review suggestion system of the present invention.

FIG. 4 is a schematic diagram showing the steps of the execution process of the multi-file automatic summary module of the present invention.

Figure 5 is a schematic diagram of the processing steps of the commodity price comparison recommendation system of the present invention.

FIG. 6 is a schematic diagram of uploading and recognizing a product after photographing according to an embodiment of the present invention.

FIG. 7 is another schematic diagram of uploading and recognizing a product after taking a photograph according to an embodiment of the present invention.

Figure 8 is a schematic diagram of product related information in an embodiment of the present invention.

Figure 9 is a schematic diagram of a product review suggestion in an embodiment of the present invention.

Figure 10 is a schematic diagram of a product price comparison recommendation in an embodiment of the present invention.

01. . . Intelligent mobile service product evaluation system using information retrieval technology

02. . . Instant product identification system

03. . . Product review suggestion system

04. . . Commodity price comparison system

Claims (11)

A smart mobile service product evaluation system, the system at least comprising: a mobile user (MUs), the user can use various terminal devices to communicate with the smart mobile device through a mobile communication network or the Internet The service product evaluation system is connected to obtain the product evaluation information released by the smart mobile service product evaluation system; a Real-time Merchandise Identifying System (RMIS), which is located on the Internet The front end is used to provide multimedia information processing and image recognition related services. The service includes instant multimedia transmission technology and instant image recognition technology. The software component used in the instant product identification system includes at least one instant multimedia transmission module and an instant image. Identifying a module, and the execution step of the instant multimedia transmission module includes at least; and a Merchandise Evaluation System (MES) for providing a blog content retrieval agent service and a multi-file automatic summary service. When the real-time product identification system (RMIS) completes the identification of a commodity After that, the information of the product is transmitted to the product comment suggestion system, and the blog content related information is searched for the blog related information through the blog content retrieval agent service, and the related comment information of the product is searched for, and the product related comment information is obtained. Crawl and Parse are stored and stored as Blog Corpus, and the multi-file automatic summary service is used to retrieve the product reviews in the relevant Blog Corpus, and is made into a summary type, which is provided to the user for purchasing decision reference; and a commodity price recommendation system (Merchandise Recommendation System) , MRS), for providing commodity price information to obtain agent service and price comparison recommendation service, when the instant product identification system (RMIS) completes identification of a product, The information of the product is transmitted to the product price comparison recommendation system (MRS), and the commodity price information is used to search for the commodity price related information to the comparison network and the related network, and the related product price information is searched in each e-commerce website. And the product price related information is Crawl and Parse and stored as the commodity price Corpus, and the price recommendation system is used to retrieve the price information in the commodity price Corpus for comparison and product recommendation; wherein, the instant multimedia transmission mode The execution step of the group includes at least: (a) converting the input real-time image into an RGB vector matrix representation for subsequent operations; (b) obscuring the image using a Low Pass Filter; (c) blurring the image The image is processed by the RGB three primary colors respectively; (d) the pixel number of the 64 colors accumulated after the image is reduced is recorded as a bar graph to check whether the actual graphics processing has problems or errors; e) mark the same color and adjacent points in the image as the same block; (f) avoid too many broken blocks to affect the speed of program execution Degree, the action of merging the fragmentation block below a certain proportion in the image with the larger block around the image; (g) taking out the first K blocks having the largest area ratio in the image, and the K blocks are The color, the area ratio, the shape, and the number of adjacent blocks are stored in a two-dimensional array, and the adjacent relationship between the K blocks is stored as a block adjacent map/Region Adjacency Graph (RAG); (h) The different RAG features produced by each commodity are put into a Self-Organizing Maps (SOM) for training, and the characteristics of each commodity are learned and classified through the operation of the SOM algorithm; (i) When a new image enters the system, a similarity judgment can be made with the classification after the SOM operation to find a similar item. For example, in the smart mobile service product evaluation system described in claim 1, the terminal device of the mobile user is a desktop computer. For example, the smart mobile service product evaluation system described in claim 1 wherein the mobile device terminal device can be one of the following devices: a notebook computer, a Tablet PC, a PDA, a Smartphone, a 3G mobile phone, and a handheld device. Mobile devices, embedded system devices and satellite transmission and reception devices. For example, the smart mobile service product evaluation system described in claim 1 of the patent scope, wherein the service commodity evaluation system can be used for a sightseeing medical service recommendation system, and when used in a sightseeing medical service recommendation system, the mobile user terminal (Mobile Users, MUs) devices include at least: a Notebook, Tablet PC, PDA, Smartphone, or 3G mobile phone providing the user for use and display; a satellite transmission receiving device providing identification and positioning services; and an Internet connection device providing network Road connection and transmission; an internet browser provides web browsing and operation functions; a microphone and earphone provide sound input and output; and a display device that transmits (receives) the captured and received data. The smart mobile service product evaluation system according to claim 1, wherein the instant product identification system comprises at least: a personal computer or a server providing system construction and display; and an internet connection device providing Network connection and transmission; a database server function to provide data inventory retrieval service; a microphone and earphone to provide sound input and output; and a display device that transmits (receives) the captured and received data. The smart mobile service product evaluation system described in claim 1, wherein the product review suggestion system comprises at least: a personal computer or a server providing system construction and display; and an internet connection device providing Network connection and transmission; a database server function to provide data inventory retrieval service; a microphone and earphone to provide sound input and output; and a display device that transmits (receives) the captured and received data. The smart mobile service product evaluation system according to claim 1, wherein the product price comparison recommendation system comprises at least: a personal computer or a server providing system construction and display; and an internet connection device providing Network connection and transmission; a database server function to provide data inventory retrieval service; a microphone and earphone to provide sound input and output; and a display device that transmits (receives) the captured and received data. For example, the smart mobile service product evaluation system described in claim 1 , wherein the software component used in the product price comparison recommendation system comprises at least one commodity price information acquisition agent module and a price comparison recommendation module. The smart mobile service product evaluation system described in claim 8 wherein the software component used by the product price information capture agent module comprises at least a fuzzy search module, an HTML Crawler module, and One HTML Parser module. A smart mobile service product evaluation system, the system at least comprising: a mobile user (MUs), the user can use various terminal devices to communicate with the smart mobile device through a mobile communication network or the Internet The service product evaluation system is connected to obtain the product evaluation information released by the smart mobile service product evaluation system; a Real-time Merchandise Identifying System (RMIS), which is located on the Internet The front end is used to provide multimedia information processing and image recognition related services, including instant multimedia transmission technology and instant image recognition technology; and a Merchandise Evaluation System (MES) to provide blog content retrieval The agent service and the multi-file automatic summary service, when the instant product identification system (RMIS) completes the identification of a product, transmits the information of the product to the product review suggestion system, and then obtains the agent service through the blog content. Blog related information to Google Blog Search to search for relevant information about this product The Crawl and Parse are reviewed and stored as Blog Corpus, and the multi-file automatic summary service is used to retrieve the product reviews in the relevant Blog Corpus and make a summary type, which is provided to the user for purchase decision reference. The software component used in the product review suggestion system includes at least a blog content capture agent module and a multi-file automatic summary module; and a Merchandise Recommendation System (MRS) for providing commodity price information. Taking the agent service and the price comparison recommendation service, when the instant product identification system (RMIS) completes the identification of a product, The information of the product is transmitted to the product price comparison recommendation system (MRS), and the commodity price information is used to search for the commodity price related information to the comparison network and the related network, and the related product price information is searched in each e-commerce website. And the relevant price information of the products is Crawl and Parse and stored as the commodity price Corpus, and the price recommendation system is used to retrieve the price information in the commodity price Corpus for comparison and product recommendation; wherein, the multi-file automatic summary The execution steps of the module include at least: (a), Preprocess: the blog content captured by the blog content capture agent module is used for word segmentation and clause processing, so as to facilitate subsequent word weight operation. (b), Feature Selection: design different features for the weight calculation of the word clause, the main Feature includes Centerity, Sentence Length, and Position (Po), Classifier : Calculate the score of each sentence using weight, and calculate and sort the score according to the similarity of sentences; (d), Rera Nker: recalculate the similarity between the sentence and the sentence, and set a threshold to filter, take out sentences that are important and not too similar to each other, and then extract according to the set compression ratio; (e) Summary: According to the sentence sequence sorted by Reranker, the words are taken out from the original file, and the summary of the output is recombined; and (f), Evaluation: measure the performance of the system's text summarization, including the effect of the output, and the user's satisfaction level. For example, the smart mobile service product evaluation system described in claim 10, wherein the blog content captures the software element used by the agent module The component includes at least one fuzzy search module, an HTML Crawler module, and an HTML Parser module.