JP2003044756A - On-line shopping method, shopping site, voice recognizing device and voice recognition supporting device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an on-line shopping method capable of performing on-line shopping with verbal communication similar to over-the-counter sales by performing connection processing between a consumer browser terminal and an electronic commercial site server for providing merchandise. SOLUTION: A procedure word subjected to voice recognition by a user terminal is received via a communication network with respect to contents spoken by the user, a merchandise database is retrieved on the basis of the procedure word, merchandise retrieval results coinciding with this retrieval are referred to with the user terminal, merchandise subjected to voice recognition by the user terminal and designated among the merchandise retrieval results by the user is stored in a cart, and selling and buying are established on the basis of purchase intention display of the merchandise stored in the cart spoken by the user.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an online shopping method suitable for online shopping on the Internet using voice recognition. The present invention also relates to a shopping site using voice recognition,
Particularly, it relates to a shopping site that is easy to use from the viewpoint of the user. Furthermore, the present invention relates to a voice recognition device suitable for use in a user terminal that connects to a shopping site that uses voice recognition and a shopping site that receives voice from the user as voice information. The present invention also relates to a voice recognition support device suitable for use in a shopping site using voice recognition.

[0002]

2. Description of the Related Art Conventionally, when carrying out online shopping using the Internet, first, from a general consumer's house, a personal computer having a browser function is used.
Get access to an e-commerce site. Next, when the product display on the web (Web) page of the e-commerce site is clicked with the mouse, the product display hyperlink function causes a transition to the product purchase application screen or the product search screen for selecting products or It becomes possible to perform the work required for purchase. In this case, due to the restriction of the category classification structure of the product structure, it is necessary to perform a plurality of clicks in order to reach the desired product in the online shopping. This situation is similar to that when a consumer visits a new store whose product layout is unknown and goes to the display shelf of the desired product, the consumer goes around.

Further, as a social circumstance to be noted, the operation of a mouse is easily accepted by young people who are accustomed to the operation of a personal computer, but the elderly people feel a great deal of resistance. Especially in Japan, this tendency is large because the elderly people did not have the custom of using typewriters as in the West. On the other hand, in the sale of daily necessities, the purchasing power of the elderly is much larger than that of the younger generation, and it is indispensable for the spread of electronic commerce to make the elderly feel familiar with e-commerce in addition to the conventional face-to-face sales. is there.

Therefore, if a voice recognition program is installed in the browser terminal used by the consumer, the consumer can use the voice instead of clicking the mouse.
The e-commerce site operator thought that the elderly could feel familiar with e-commerce.

[0005]

However, the functions of the voice recognition program that can be installed in the browser are limited, and in applications such as online shopping that require voice recognition for a large number of products or daily products. , In order to show the response on the browser side according to the consumer's request, it is necessary to update the dictionary function on the browser side at any time,
Therefore, there is a problem that the communication load for updating the dictionary is significantly increased. That is, in order to complete online shopping with only voice, it is necessary to make the procedure words handled by the browser terminal and the electronic commerce site server side equal, which poses a problem that this equalization work is heavy. .

Therefore, it is conceivable that a voice recognition program is set on the electronic commerce site server side, the utterance of the consumer is digitized on the browser terminal, and the electronic commerce site server side performs voice recognition. But in this case,
There is a problem that a large amount of data needs to be exchanged between the browser terminal and the electronic commerce site server, and the response becomes poor due to the limitation of the communication capacity and the communication speed. For example,
In geostationary satellite communication, the call delay time is 0.2 seconds, but the caller feels slow even in such a short time. In addition, in order to reduce communication costs, VOIP (Voice Over Inte
Although it is possible to use the rnet Protocol), if the communication line is congested, the delay time becomes several seconds, and there is a problem that the conversation cannot be established.

The present invention solves the above problems, and a first object is to carry out a cooperative process between a consumer's browser terminal and an electronic commerce site server providing a product, and in the same oral manner as in the case of over-the-counter sales. It is to provide an online shopping method and a shopping site where online shopping can be performed by exchanging information. The second purpose is that when the voice recognition on the browser terminal does not work well for the utterance of the consumer, the voice recognition is accurately performed without the consumer feeling that the ability of the voice recognition is remarkably low. It is to provide a voice recognition device for ensuring the above. A third object is to provide a voice recognition support device in which the function of voice recognition can be easily upgraded so that the electronic commerce site server can appropriately cope with the product configuration that is changed daily.

[0008]

In the online shopping method of the present invention for achieving the first object, as shown in FIG. 5, the content uttered by the user is recognized by the user terminal 1 by voice. The procedure word is received via the communication network 4 (S14), the product database is searched based on the procedure word (S16), and the product search result matching the search is referred to by the user terminal 1 (S18).
It is an intention to select a product that is voice-recognized by the user terminal 1, and the product specified by the user from the product search results is stored in the cart (S20) and stored in the cart uttered by the user. The sale is completed based on the displayed purchase intention display of the product (S24).

In the online shopping method having such steps, the contents uttered by the user are recognized by the user terminal 1 as a procedure word and transmitted to the shopping site 2 by the user terminal 1. The shopping site 2 searches the database according to the procedure word and causes the user terminal 1 to refer to the product search result that matches the search. Next, since the product to be stored in the cart is voice-recognized from the user terminal 1 and transmitted to the shopping site 2, the shopping site 2 stores in the cart the product specified by the user from the product search results. . Then, since the purchase intention display of the product stored in the user's cart is transmitted from the user terminal 1 to the shopping site 2, the shopping site 2
Then the sale is completed.

As shown in FIG. 3, the shopping site of the present invention that achieves the first object transmits the procedure word recognized by the user terminal 1 to the contents uttered by the user through the communication network 4. Received by the procedure word receiving means 221, a product search means 222 for searching the product database 23 based on the procedure word, and a product search for making the product search results matching the search available in the user terminal 1. The result creation unit 223, the cart storage unit 224 that stores the product designated by the user from the product search result in the cart according to the user's designation, and the user's intention to purchase the product. On the basis of the above, the commodity transaction execution means 225 for executing the transaction of the commodity accommodated in the cart is provided.

In the apparatus configured as described above, the procedure word receiving unit 221 receives the procedure word recognized by the user terminal 1 by voice with respect to the content uttered by the user via the communication network. Next, the product search unit 222 searches the product database 23 based on the procedure word received by the procedure word reception unit 221, and the product search result creation unit 223 sends the product search result matching the search to the user terminal 1. And refer to it. Then, the user specifies the product by the user's utterance from the product search results, and the cart storage unit 224 stores the product specified by the user in the cart. When the user finishes the specification of the product to be purchased, the user indicates the purchase intention of the product, and the product transaction execution means 225 executes the transaction of the product accommodated in the cart.

Preferably, as shown in FIG. 2, FIG. 3, and FIG. 4, the user terminal 1 is provided with a voice recognition database 18, and a voice containing update information of the product database 23 provided at the shopping site 2 is provided. Shopping site 2 via common database 33 for recognition and communication network 4
The revision state of the voice recognition database 18 of the user terminal 1 that has accessed the user terminal and the voice recognition common database 3
3 and the equalization means 322 which detects the difference from the revised state and updates the contents of the voice recognition database 18 so that the same state as the voice recognition common database 33 is obtained. Even if the product handled on the site 2 is changed or the product transaction form is changed, the voice recognition function of the user terminal 1 is not changed by the voice server 3
It will be the latest version according to the update contents of.

Preferably, the product search means 222 is configured to search the product database 23 as a common noun of a product used by the brand when the procedure word is a brand instead of the common noun of the product. It is possible to search for a product corresponding to the recognition when the user purchases the product.

Preferably, the product search means 222 searches the product database 23 for the common noun of the product indicated by the procedure word, including the common noun of the related product used in association with the product. Then, when the user purchases the product, the related product used in association with the product is also displayed in the product search result, so that the product suitable for the consumption behavior of the user can be searched.

Preferably, when the procedure word indicates an attribute of a product, the product search means 222 extracts a common noun of a product having the attribute as an intermediate search word, and uses the intermediate search word to search the product database 23. With the search configuration, it is possible to search for a product that matches the consumption behavior when the user uses the product attribute as a reference for product selection.

[0016] Preferably, when the procedure word is a procedural sentence including an attribute term indicating a plurality of attributes of a product, the merchandise retrieval unit 222 extracts the attribute term constituting the procedural sentence and extracts the attribute term. If a common noun of a product having attributes of a plurality of attribute terms is extracted as an intermediate search word and the intermediate product search word and the plurality of extracted attribute terms are used to search the product database 23, the user purchases the product. When you do so, you can search for the product you want to buy by purchasing the product using the tone that the user speaks to the clerk, and you can search for products that match the daily business practice such as face-to-face sales. You can do it.

Preferably, the product search result creating means 223
If the products searched by the product search unit 222 are classified according to brand name and included in the product search result, the user's consumption behavior is appropriate when the brand name is used as a criterion for product selection. Product search results are created

[0018] Preferably, the commodity transaction fulfilling means 225 is
If the purchase intention display of the products accommodated in the cart is recorded by the voice recognition of the user terminal,
If there is a discrepancy in the commercial transaction between the user and the shopping site 2, it is possible to smoothly carry out the subsequent solution.

As shown in FIG. 2, the voice recognition device of the present invention which achieves the second object accesses the shopping site 2 and outputs the procedure word recognized by voice to the contents uttered by the user. In the voice recognition device used in the user terminal 1 for the shopping site that transmits and advances the purchase of products, in order to recognize the voice uttered by the user, the acoustic model database 17 of the user registered in advance, Input the voice spoken by the user,
When the voice recognition matching rate calculating means 151 for calculating the recognition matching rate of the voice uttered by the user with reference to the acoustic model database 17, and the calculated recognition matching rate of the voice is lower than a predetermined threshold value. Includes a reply means 152 for prompting the user to speak again.

Preferably, as shown in FIG. 2, when the voice recognition matching rate calculated by the voice recognition matching rate calculating means is lower than a predetermined threshold value, the calculated voice recognition matching rate is When the configuration is provided with the multiple candidate presenting unit 153 that presents the relatively high speech-recognized procedure words as a plurality of candidates, the procedure word that serves as an index when the user speaks again is displayed on the user terminal 1. The user's utterance content can be appropriately guided.

The voice recognition support device of the present invention which achieves the third object receives the procedure word recognized by the user terminal 1 through the communication network 4 and searches the product database based on the procedure word. Voice recognition by the user terminal 1 to the shopping site 2 that creates a product search result that matches the search, and executes the transaction of the product included in the product search result based on the user's intention to purchase the product. Is a voice recognition support device, which is configured to include means for managing the voice recognition common database 33 provided for each shopping site 2.

By using the voice recognition support device of the present invention, the operator of the shopping site 2 adds a new product or deletes a dead product in the voice recognition common database 33. The equalizing means can update the voice recognition common database 18 of the user terminal 1 to the latest contents of the voice recognition common database 33. Therefore, in each shopping site 2, it is possible not only to update the common database 18 for voice recognition without notifying the user of addition of new products or deletion of muscle products, but also to add a procedure word to a plurality of products. It is also possible to connect to cross-selling by connecting to.

[0023]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. In each drawing, the same or corresponding members are designated by the same reference numerals or similar reference numerals, and redundant description will be omitted. FIG. 1 is a configuration block diagram showing an embodiment of the present invention. In the figure, a user terminal 1, a shopping site server 2, and a voice server 3 are connected via the Internet 4. The user terminal 1 has a browser function and a voice recognition function, and for example, a personal computer, a portable information terminal, a mobile phone having an Internet connection function, or the like is used. For the shopping site server 2 and the voice server 3, computers having high-speed information processing capability such as minicomputers, office computers, and servers are used.

The Internet 4 as a communication network uses HTTP (Hyper Text Transfer Protocol) as a communication protocol.
l) and WAP (Wireless Application Protocol)
Communication information is HTML (Hyper Text Markup Language), X
ML (eXtensible Markup Language), WML (Wireless
Markup Language) for transmitting web page information described in a language such as an optical fiber communication line, a wired line using copper wire, and a wireless line for mobile communication. When the communication information is a moving image content file, for example, MPEG (Moving Picture Expe
rt Group) method. IMT (International Mobil) is used for internet communication for mobile communication.
e Telecommunication) 2000, CDMA (Code Division M
The user terminal 1 is provided with a transmission / reception function conforming to the standard for digital communication such as ultimate access).

FIG. 2 is a block diagram showing the configuration of the user terminal. The user terminal 1 includes a voice input unit 11, an acoustic analysis unit 12,
The display unit 13, the network card unit 14, the recognition decoder unit 15, the browser control unit 16, the acoustic model database 17, and the voice recognition database 18 are provided. The voice input unit 11 is a device that delivers a user's voice to the acoustic analysis unit 12 using a microphone or the like. Acoustic analysis unit 1
Reference numeral 2 is for extracting a characteristic parameter by performing coding, noise processing, correction, etc. of a voice (analog voice) spoken by a person. The display unit 13 displays Web data from the shopping site server 2, for example, CR
A T or liquid crystal display is used. The network card unit 14 is a communication interface device for transmitting / receiving data via the Internet 4 or the like.

The recognition decoder unit 15 recognizes a word string uttered from the characteristic parameters extracted by the acoustic analysis unit 12, and has a speech recognition matching rate calculation unit 151, a listening unit 152, and a plurality of candidate presentation units 153. have. The voice recognition coincidence rate calculating unit 151 inputs the voice uttered by the user and refers to the acoustic model database 17 to calculate the recognition coincidence rate of the voice uttered by the user. When the calculated voice recognition concordance rate is lower than a predetermined threshold, the listening unit 152 prompts the user to speak again, and the prompt is displayed by the audio function of the user terminal 1. It may be displayed, or may be a character graphic display using the display function of the user terminal 1. Multiple candidate presentation unit 153
Is similar to the procedure word at the time of speech recognition in which the speech recognition coincidence rate calculation unit 151 has a speech recognition coincidence rate of a user's speech lower than a predetermined threshold, but a speech recognition coincidence rate is high. If so, present the procedure word as a plurality of candidates. In the recognition decoder unit 15,
When recognizing a word string, the acoustic model database 17 and the voice recognition database 18 may be used to increase the recognition rate.

The browser control unit 16 includes the recognition decoder unit 1
5, a means for activating a program for searching for a product in the shopping site server 2, adding it to the shopping cart 25, and making a purchase decision according to the word recognized by voice. The acoustic model database 17 is a database that stores a model for calculating the probability of which phoneme the input characteristic parameter is, and reflects the speech language characteristics of the user. Speech recognition database 1
Reference numeral 8 is provided with a dictionary database 182 and a grammar database 184, and uses the contents of the voice recognition common database 33 of the voice server 3 as a master file. The dictionary database 182 is a database that stores words. Here, the names of products handled by the shopping site server 2 and terms that are convenient for searching are enriched. The grammar database 184 is a database that stores a model for calculating the probability of which word the input characteristic parameter is, and has a high recognition rate.

FIG. 3 is a block diagram of the shopping site server. The shopping site server 2 includes a network card unit 21, a web control unit 22, a product database 23, an order database 24, a shopping cart 25, and a log file 26. The network card unit 21 is a communication interface device for transmitting / receiving data via the Internet 4 or the like. The web control unit 22 is a means for executing a program for searching for a product requested from the user terminal 1, adding it to the shopping cart 25, and making a purchase decision, and is a procedure word receiving unit 221, a product search unit 222, and a product search result creating unit. 223, a cart storage means 224, and a commodity transaction fulfillment means 225.

The procedure word receiving unit 221 receives, via the communication network 4, the procedure word recognized by the user terminal 1 by voice with respect to the content uttered by the user. The product search unit 222 searches the product database 23 based on the procedure word. Product search result creation means 223
Makes the product search results that match the search by the product search means 222 available for reference on the user terminal 1. The cart storage unit 224 stores the product designated by the user from the product search result in the shopping cart 25 according to the user's designation. Commodity transaction fulfillment means 225
Executes the transaction of the product accommodated in the shopping cart 25 based on the user's intention to purchase the product.
The product database 23 is a database that stores product information. The order database 24 is a database for temporarily holding the product to be purchased and storing the product data after the purchase decision.

The shopping cart 25 is a file for temporarily storing the product purchased by the user at the shopping site server 2, and is stored until the purchase contract is established and the transportation and payment of the price are completed. The log file 26 stores the search results of the product search unit 222, the products stored in the shopping cart 25 by the cart storage unit 224, and the product transaction history of the product transaction fulfillment unit 225.
The contents to be stored may be, for example, items necessary for updating the dictionary database 332, and the storage period may be an appropriate period, for example, a period required for a product life cycle or a normal settlement of commercial transactions.

FIG. 4 is a block diagram of the configuration of the voice server. The voice server 3 includes a network card unit 31, a voice system management unit 32, and a voice recognition common database 33.
In addition, the site information database 34 is provided. The network card unit 31 is a communication interface device for transmitting / receiving data via the Internet 4 or the like. The voice system management unit 32 has a function as the equalization unit 322, and is a dictionary database 182 or a grammar database 18 set in the user terminal 1.
Check whether the contents of 4 are the latest, and if they are not the latest, dictionary database 182 or grammar database 18
In order to equalize the contents of 4 into the latest data, the latest data of the voice recognition common database 33 is downloaded to the user terminal 1.

The common database 33 for voice recognition has a dictionary database 332 and a grammar database 334. The dictionary database 332 is a database that stores words, and stores the content that serves as the master of the dictionary database 182 set in each user terminal 1. The grammar database 334 is a database that stores a model for calculating the probability of which word the input characteristic parameter is, and stores the content that will be the master of the grammar database 184 set in each user terminal 1. ing. The site information database 34 includes the shopping site server 2 and the dictionary databases 182, 3
32 and a grammar database 184, 334 is a database that manages the association of contents and versions.

Next, the operation of this embodiment will be described.
FIG. 5 is a flow chart for explaining the purchase of products on the shopping site by the user terminal. First, the shopping site server 2 receives access in the form of login of the user terminal 1 or the like (S10). Then, the shopping site server 2 receives the voice system management unit 32 of the voice server 3.
Thus, the contents of the voice recognition database 18 of the user terminal 1 are managed so as to be equalized with the latest data of the voice recognition common database 33 (S12). And
If the content of the voice recognition database 18 is the content of the voice recognition common database 33 of the old issue date, the latest data of the voice recognition common database 33 is downloaded and the content of the voice recognition database 18 is updated.

Next, the user views the product search page presented on the user terminal 1 from the shopping site server 2 and utters a procedure word. Then, the user terminal 1
The uttered procedure word is recognized by the recognition decoder unit 15. The shopping site server 2 receives the procedure word transmitted from the user terminal 1 via the communication network 4 (S14), searches the product database 23 based on the received procedure word (S16), and matches the search. The product search result is referred to by the user terminal 1 (S18). Then, the user views the product search result displayed on the user terminal 1 and utters the designation of the product to be purchased. The designation includes a number, a common name of a product, a place of origin, a brand, and the like. Then, the user terminal 1 voice-recognizes the intention to select the product and transmits it to the shopping site server 2. Then, according to the product selection intention received from the user terminal 1, the shopping site server 2 stores the product specified by the user from the product search results in the shopping cart 25 (S20).
If the user has an item to purchase (S22), the user subsequently utters a procedure word, so the shopping site server 2 returns to S14 and processes it.

On the other hand, when all the products that the user wants to purchase have been stored in the shopping cart 25 (S22), the shopping site server 2 uses the shopping cart 25.
Based on the user's intention to purchase, the sale and sale of the products accommodated in the store are completed, and the stage of fulfillment is entered (S2).
4). The establishment of a sale means, for example, determining a delivery method of the price and notifying the shopping site server 2 of information necessary for delivering the product. Payment methods include payment by credit card, payment by David card, cash on delivery service by TA-Q-BIN (registered trademark) and post office, bank transfer, postal transfer, and registered mail.
The information necessary for delivering the product includes a delivery place, delivery date and time, delivery form, and the like.

FIG. 6 is a flow chart for explaining the process between the user terminal, the shopping site server, and the voice server for shifting the state in which the user's utterance is effectively voice-recognized by the user terminal. The user accesses the shopping site server 2 corresponding to online shopping by voice from the user terminal 1 and requests display of page A of a certain shopping site A (S100). Upon receiving the connection request from the user terminal 1, the shopping site server 2 causes the user terminal 1 to display page A based on the voice recognition common database 33 for the shopping site A (S101).

On page A, it is confirmed whether the voice recognition database 18 of the user terminal 1 is the latest, and the latest contents of the voice recognition common database 33 for the shopping site A stored in the voice server 3 are stored. The function required for equalization is realized by, for example, an agent function. Therefore, it is confirmed whether a constituent element of the voice recognition database 18, for example, the dictionary database 182 for the shopping site A exists in the user terminal (S10).
2) If it does not exist, it requests the voice server 3 to download the dictionary database 332 (S10).
4). Even when the dictionary database 182 for the shopping site A of the user terminal 1 does not match the latest version (S103), a request for downloading the dictionary database 332 to the voice server 3 is made (S104). On the other hand, the latest version of the dictionary database 332 is downloaded from the voice server 3 to the user terminal 1 (S105). Then, when the dictionary database 182 for the shopping site A of the user terminal 1 is brought into a state of matching with the latest version, the voice recognition of the user by the user terminal 1 is started (S106).

If the user terminal 1 accesses the shopping site B instead of the shopping site A, steps S110 to S11 similar to steps S100 to S106.
6, the dictionary database 182 for the shopping site B of the user terminal 1 is brought into a state of matching with the latest version. The dictionary database 182, which differs depending on the shopping sites A and B, and the grammar database 184, if necessary, are provided because the procedure words vary greatly depending on the type of store. For example, if the shopping site A is a perishable food, there are words and phrases specific to meat, fish, vegetables, etc. If the shopping site B is a liquor store, words and phrases specific to sake, beer, wine, etc. Because there is turning.

FIG. 7 is a diagram for explaining the types of procedure words. The procedure word is the dictionary database 182, 33.
It is stored in No. 2, and there are types such as for product search and shopping cart operation. For product search, the attributes of the product are shown.
There are product names, product categories, and so on. The characteristics of the product include sweetness and saltiness in the case of food, and summer and winter in the case of clothing. There are distinctive uses for products such as gifts and personal use, and food products include allergies, infants, seniors, diets, and athletes. The product name includes important information about product quality and customer appeal such as manufacturer, seller, common name and brand. For the operation of the shopping cart, a display operation such as on the cart and a phrase such as product deletion are included.

FIG. 8 is a flow chart when the user searches for a product. The user utters a procedure word such as a specific product name, color, or keyword to be imaged through a microphone or the like (S121). Then, in the user terminal 1, the acoustic analysis unit 12 and the recognition decoder unit 15 effectively act to recognize the user's voice (S122). And
The procedure word for product search, which is voice-recognized by the user terminal 1, is transmitted to the shopping site server 2 (S1).
23). The shopping site server 2 searches the product database 23 based on the procedure word and the intermediate search word related to the procedure word (S124), and returns the product search result to the user terminal 1 (S125). Then, the user terminal 1 displays a list of product search results (S126) to facilitate the user's product selection.

FIG. 9 is a flow chart when the user selects a product by using the product search result. The user is the user terminal 1
Looking at the list display of the product search results displayed in, the product to be purchased is selected and stored in the shopping cart 25. In this case, according to the procedure word stored in the dictionary database 182 for the connected shopping site 2 (for example, the number or the number of the displayed product), the user utters the message between the user terminal 1 and the shopping site 2. Indication of intention to purchase goods is processed smoothly. As an example, consider a case where a user adds three red wines to the shopping cart 25. When the user utters "3 bottles of red wine" (S131), the recognition decoder unit 15 of the user terminal 1 recognizes the procedure word uttered by the user (S132). Then, the product name for product input, which is voice-recognized by the user terminal 1, is transmitted to the shopping site server 2 (S133). The shopping site server 2 adds the transmitted product name to the shopping cart 25 (S134). The information on the shopping cart 25 is managed by the shopping site server 2 for each user.

FIG. 10 is a flow chart when the user purchases a product. Dictionary database 182 even when making a purchase
Speaking in accordance with the procedure word stored in, the processing is smoothly performed. For example, when the user utters "Order decision" (S141), the acoustic analysis unit 1 is used in the user terminal 1.
2 and the recognition decoder unit 15 effectively operate to recognize the user's voice (S142). Then, the intention to purchase the product, which is voice-recognized by the user terminal 1, is transmitted to the shopping site server 2 (S143). The shopping site server 2 creates order data in accordance with the intention to purchase the product (S144), and returns the created order data to the user terminal 1 to enable the user's confirmation (S).
145).

When the user utters "Order decision" again (S146), the user terminal 1 recognizes the user's voice (S147) and recognizes the purchase confirmation intention. The purchase confirmation intention is transmitted from the user terminal 1 to the shopping site server 2 (S148). The shopping site server 2 stores the order database 24 as order data based on the received intention to confirm the purchase.
(S149). Although FIG. 10 shows an example in which reconfirmation is performed via the shopping site server 2,
The purchase confirmation may be performed again when the procedure word for purchase is uttered on the user terminal 1 side.

FIG. 11 is a flow chart for explaining the voice recognition processing. The user speaks at the voice input unit 11 through a microphone or the like (S201). Then, by the acoustic analysis unit 12,
The speech uttered by the user is subjected to encoding, noise processing, correction, etc., and its characteristic parameters are extracted (S20).
2). The acoustic model database 17, the dictionary database 182, and the grammar database 184 are taken into consideration for the extracted characteristic parameters (S204), and the word string uttered by the recognition decoder unit 15 is recognized (S).
203). At this time, a plurality of procedural words such as color, size, and place of production are extracted from the dictionary database 182 and the grammar database 184, so that it is not necessary for the user to be aware of the order in which they speak.

When the decoding result by the recognition decoder unit 15 is normal (S205), the event is activated using the recognized word string (S208). On the other hand, when the decoding result is NG (S205), the dictionary database 1
82 and the word string inferred from the grammar database 184 are presented to the user as candidates (S206). The user determines whether there is a corresponding word string from the word string candidates, and if there is a corresponding word string, selects it (S20).
7) The event is activated (S208). If there is no corresponding word string (S207), the process returns to S201 to perform voice input again. Here, the activation of an event includes searching for products, storing products in a cart, purchasing products, and the like.

FIG. 12 is a flow chart for leaving a history that the user has performed voice shopping. It has already been shown in FIG. 5 that voice shopping has processes of product search, input (add to shopping cart), and product purchase.
This is as described with reference to FIGS. In this case, the actions taken by the user in each process of voice shopping are collected as a history, and the shopping site 2
Should be improved to meet the needs of the user, for example, maintenance of the dictionary database 332 should be carried out to keep the search efficient. Specifically, it is as follows.

At the time of searching for products, the procedure word uttered by the user is recognized by the user terminal 1 (S251, S
252), and sends it to the shopping site server 2 side
(S253). The shopping site server 2 side records the transmitted procedure word in the log file 26 and leaves it as a history of product search (S254). : When the product is stored in the cart, the procedure word spoken by the user is recognized by the user terminal 1 (S251, S25).
2), send to the shopping site server 2 side (S2
53). On the shopping site server 2 side, the transmitted procedure word is recorded in the log file 26, and the product name stored in the shopping cart 25 is left as a history. : When purchasing a product, the procedure word uttered by the user is recognized by the user terminal 1 (S251, S252) and transmitted to the shopping site server 2 side (S253). On the shopping site server 2 side, the transmitted procedure word is recorded in the log file 26 and left as a history of the user's decision-making regarding product purchase.

FIG. 13 is a screen view showing an example of a web page displayed on the user terminal by the shopping site. The shopping site screen 100 includes a shopping field 102 and a community field 10 as a schematic display of a site map.
4. A banking column 106 and a schedule column 108 are provided. The shopping site screen 100 further includes an order method selection field 110, an order content confirmation field 120, a member registration field 130, a product sample display field 140, and a product detailed description field 1
50, product inventory display field 160, shopping cart field 1
70 and a voice input guidance display unit 180.

The order method selection field 110 includes a catalog order field 112, a bargain product order field 113, and a product search order field 11
4. It has an order column 115 for the purchased actual products and an order column 116 for the recommended recipe. Product search order column 11
As the details of the search mode, there are prepared in 4 a case of searching by genre, a case of AIUEO, and a case of searching by procedure word. The member registration column 130 includes a member agreement explanation column 132, a shopping member registration column 134, a member registration change column 136, and a shopping member suspension / withdrawal column 138.

In the product sample display field 140, for example, a sample of the product searched by the user in the product search order field 114 is explained with a photograph or a moving image. In the product detail description field 150, details of the product displayed in the product sample display field 140, for example, when the product is clothes, the product name, price, size, color, and quantity are displayed. The product inventory display field 160 is the product sample display field 14
This indicates the stock status of the product displayed at 0. In the case of clothes, the stock quantity is displayed for each type such as fleece, lambswool, aeric jacket, and hand knit. Preferably, if the color and size are displayed for each number of the stocked product, the correspondence with the product displayed in the product sample display field 140 becomes clear, which is useful not only for the user who purchases the product but also for store management.

The shopping cart column 170 is provided with a purchase reservation item display column 172, a purchase price display column 174, and an order determination column 176. Purchase reserved product display field 172
Displays the unit price, quantity, standard size, and price of the reserved product for purchase, and also enables cancellation of the reservation by deleting it. In the purchase price display field 174, the total price of the purchase reservation products registered in the purchase reservation product display field 172 is displayed, and a guide is also displayed regarding the delivery charge. When the user operates the order determination field 176, the order for the purchase reserved product registered in the purchase reserved product display field 172 is transmitted from the user terminal 1 to the shopping site server 2. The voice input guidance display unit 180 informs the user that voice can be input instead of the keyboard and mouse. In addition, since the voice input can be ambiguously interpreted, the shopping site server 2 side makes the user's intention. When is not confirmed, the display necessary for confirming the user's intention is displayed.

Since the shopping site screen 100 is a system for online shopping using voice, it can be applied to all kinds of products such as groceries, clothes and home appliances. First, the operator of the shopping site on the shopping site server 2 makes a contract with the operator of the voice server 3, and the dictionary database 3 of the relevant site is entered.
32 and the creation of the grammar database 334 are requested. At this time, for example, by connecting an abstract procedure word related to a preference or purpose that the consumer wants to know about a specific product such as “sweet wine” or “a wine suitable for fish dishes” with the specific product, It is also possible to maximize the convenience of.

FIG. 14 is a screen view showing an example in which the search results are displayed as a list of desired products for purchase. Here, a search result display field 190 and a shopping cart field 170 are displayed. The search result is, for example, the item search order column 114 by the user.
It is for the product name for which the search condition is entered. The user speaks “sweet red wine” to the user terminal 1, and the shopping site server 2 is the product database 2
3 is displayed, and the wine list of the corresponding brand is displayed in the search result display field 190 together with the price.

FIG. 15 is a screen view showing an example in which the product for which the consumer has made a purchase reservation is displayed on the shopping cart. When the user looks at the display of the search result display field 190, “3
When the user speaks "No. 2", the designated quantity of the product "Sweet Red Wine C" of the brand displayed as No. 3 in the wine list is added to the shopping cart 25.

FIG. 16 is a diagram showing an example of the purchase confirmation screen. When the user utters the order decision field 176 of the shopping cart field 170 or “Order decision”, the user's intention is transmitted from the user terminal 1 to the shopping site server 2. Then, shopping cart column 1
A purchase confirmation screen 178 is displayed on the screen 70. here,
When the user utters "order decision" again, the order is confirmed and the series of purchase process is completed.

FIG. 17 is a screen view for explaining the function of inquiring homonyms in the recognition decoder section. Since the number of vowels in Japanese is smaller than that in other foreign languages, it is necessary to enhance the function for inquiring homonyms. As an example of the homonyms, it is assumed that the user utters a procedure word “Sasuke”. Then, the voice input guidance display unit 180
In the squirt column 186, the user is asked again "whether it is" drinking sake "or" fish salmon "", and the inquiry contents are simply displayed in the voice recognition display column 187. Also, the voice input style display field 1
At 85, the style of the utterance content is displayed. In the order display field 188 by voice, the user is notified that the product can be ordered by voice input.

Next, the details of the product search function will be described. The product search function is related to the product search function in S124 of FIG. 8 and the product search order column 114 of FIG. 13, and shows an organic cooperative relationship between the browser control unit 16, the web control unit 22, and the product database 23. There is. If the product search function is provided with a compound procedure word search function, a heterogeneous category search function, and an abstract word procedure word search function as intermediate search words, convenience is increased. For example, if a product search function using a compound procedure word is used, it is closer to a natural conversation than a mere word, and is easily familiar to consumers. As an example of a compound procedure word,
When the user says "Full Zip Red M size", the target product is immediately displayed in the product sample display field 140 or the product detail description field 150 as shown in FIG.

FIG. 18 is a screen view for explaining the search function for different categories. As an example of searching for different categories,
The user utters a procedural word for a purpose such as "nabe cooking". Then, the recognized word or sentence is displayed in the voice recognition display field 187. Then, the product database 23 is searched using the procedure word recognized by voice, and the searched product is displayed in the search result display field 190. The displayed product search results are a list of related items from multiple categories. For example, in addition to Chinese cabbage, pickles, chicken balls, oden, chikuwa as ingredients for "nabe dishes," consumables for cooking utensils. A cylinder for a tabletop stove is displayed. Although they are products in different categories from consumables such as food and cooking utensils, they are displayed in the search result display field 190 in consideration of user convenience.

FIG. 19 is a screen diagram for explaining the abstract word procedure word search function. A voice recognition display column 202 is provided in the voice operation explanation column 200. As an example of the abstract word procedural word, when the user speaks “sweet” at a site that deals with natural food, “sweet” is displayed in the voice recognition display field 202 as a voice recognition result. Then, the product database 23 is searched using the procedure word recognized by voice, and the searched product is displayed in the search result display field 190. That is, the search result display field 190 displays a list of products related to the abstract word procedural word “sweet” such as jam, shortcake, ice cream, anpan, and juice.

As another example of the abstract word procedure word, there is a mode in which the top brand of a product is handled according to the common noun of the product. Many consumers adapt to the reality that they recognize the product as a brand name that they often buy rather than as a common noun. For example, if the user says “Super Dry” (registered trademark) or “Ichiban Shibori” (registered trademark) at a site that handles alcoholic beverages, the search result display field 190 displays the words that the user has spoken as a voice recognition result. Is displayed, and the search engine uses the common noun "beer" corresponding to these registered trademarks in the product database 2
3 is searched, and the searched products are displayed in the search result display field 190 for each brand or manufacturer.

A state in which the user purchases a product on the shopping site screen 100 having such a voice recognition function will be described in time series. FIG. 20 is a screen diagram showing an initial state in which the user has accessed the shopping site screen 100. The shopping cart field 170 is empty,
No products have been selected. Here, when the user utters the product “Potato Chips” in the product search order field 114, the search result display field 190 displays the product “Potato Chips”.
Several types of product search results included in are displayed. It is preferable that the product search results include products of representative brands of the respective manufacturers A, B, and C, and the product search results are displayed in a manner that is familiar to purchasers. When the user utters “Add 2 to Cart 3”, the voice recognition display field 202 displays “Add 2 to Cart 3” as the voice recognition result.

FIG. 21 is an explanatory view of a form in which the shopping cart stores the products to be purchased. Two 3rd products “without potato chips salt” are displayed in the purchase reservation product display field 172 of the shopping cart field 170 so as to match the voice recognition result of the voice recognition display field 202. The purchase price display field 174 displays the total price of the products in the purchase reservation product display field 172 including the delivery charge.

FIG. 22 is an explanatory diagram for explaining the establishment of a purchase contract for a product stored in a shopping cart. From the product search results displayed in the search result display field 190, the user utters a search procedure word for the product to be purchased,
Select a product and add it to the shopping cart field 170. Then, when the user utters “I will purchase” or operates the order determination field 176 with the mouse, “Thank you for your purchase” is displayed in the ejection field 186 of the voice input guidance display unit 180. .

FIG. 23 shows a delivery product registration screen for purchased products and a payment method confirmation screen. The user registers the client's name, address, telephone number, date of birth, gender, etc. In this case, if the user registers the destination of the purchased product in the shopping member registration column 134 of the member registration column 130 and registers the payment method in advance, the destination registration and the payment method are performed each time the purchase is made at the shopping site. It is convenient because you do not need to specify. For identification of the user, for example, a member number and password issued in the member registration field 130 may be used.

FIG. 24 is a screen view for prompting another utterance when the uttered voice cannot be recognized. When the voice recognition concordance calculating unit 151 determines that the recognition concordance of the voice uttered by the user is lower than a predetermined threshold value, the voice recognition display field 18 indicates that the voice cannot be recognized.
“????” is displayed on 7. At this time, in order to make clear to the user that voice recognition was not possible, “What?” Is displayed in the gush field 186, and further the prompting means 152 prompts the user to input a voice again. Other words can be freely used in the blowout column 186 in order to prompt the user to speak again.

FIG. 25 is a screen view when the spoken voice can be sufficiently recognized, but there are a plurality of candidates for the spoken voice. In the voice recognition coincidence rate calculating means 151, the recognition coincidence rate of the voice uttered by the user is lower than a predetermined threshold value, but the voice recognition coincidence rate is high, and is approximated to the procedure word at the time of voice recognition. If so, the plural candidate presenting unit 153 presents the procedure word as a plurality of candidates.
For example, if the user utters “Mat” but the voice recognition concordance rate is below the threshold value, the procedure words are “Sewing thread (knit)”, “Miscellaneous mat”, and “Food natto”. It is possible to present the candidate in the squirt field 186 as a candidate and prompt the user to speak again.

[0067]

EFFECT OF THE INVENTION The shopping site of the present invention is based on the procedure word receiving means for receiving, through the communication network, the procedure word speech-recognized by the user terminal for the content uttered by the user. A product search means for searching the product database with a product, a product search result creating means for making a product search result matching the search available on the user terminal, and a message uttered by the user from the product search results. Cart storing means for storing the designated product in the cart according to the user's designation, and product transaction fulfilling means for fulfilling the transaction of the product contained in the cart based on the user's intention to purchase the product. Since the configuration is provided, the user can use voice to perform online shopping using a communication network such as the Internet. Further, the product search unit and the product search result creation unit eliminate the need to go through many pages depending on the site configuration of the shopping site to reach the product desired by the user, and the user can enjoy high convenience. In particular, when a high-speed communication network capable of moving image distribution is used as the communication network, stress such as slow response time is not felt.

The voice recognition device of the present invention accesses the shopping site, transmits the procedure word recognized by voice to the content uttered by the user, and advances the purchase of the product. The user terminal for the shopping site. In the voice recognition device used for, in order to recognize the voice uttered by the user, the acoustic model database of the user registered in advance and the voice uttered by the user are input, and the acoustic model database is referred to. Voice recognition matching rate calculation means for calculating the recognition matching rate of the voice uttered by the user, and when the calculated recognition matching rate of the voice is lower than a predetermined threshold, Since it is configured to have a reply means for prompting the utterance of, even in online shopping using a communication network such as the Internet, Over The is in the case of the purchase of goods by using the voice, shopping site can be carried out smoothly the recognition of voice that speech of the user in conjunction with voice recognition device.

The voice recognition support device of the present invention receives the procedure word recognized by the user terminal through the communication network, searches the product database based on the procedure word, and displays the product search result matching the search. make,
A voice recognition support device that supports voice recognition by a user terminal for a shopping site that executes a transaction of a product included in the product search result based on the user's display of purchase intention of the product. Since the configuration is provided with means for managing the voice recognition common database, the shopping site operator notifies the users of addition of new products one by one by updating the contents of the voice recognition common database. Can be dealt with without.

[Brief description of drawings]

FIG. 1 is a configuration block diagram showing an embodiment of the present invention.

FIG. 2 is a configuration block diagram of a user terminal.

FIG. 3 is a configuration block diagram of a shopping site server.

FIG. 4 is a configuration block diagram of a voice server.

FIG. 5 is a flowchart illustrating a product purchase at a shopping site by a user terminal.

FIG. 6 is a flowchart illustrating processing between the user terminal, the shopping site server, and the voice server for shifting to a state in which the user's utterance is effectively voice-recognized by the user terminal.

FIG. 7 is a diagram illustrating types of procedure words.

FIG. 8 is a flowchart when a user searches for products.

FIG. 9 is a flow chart when the user selects a product using the product search result.

FIG. 10 is a flow chart when a user purchases a product.

FIG. 11 is a flowchart illustrating a voice recognition process.

FIG. 12 is a flowchart for leaving a history that a user has performed voice shopping.

FIG. 13 is a screen diagram showing an example of a web page displayed on the user terminal by the shopping site.

FIG. 14 is a screen diagram showing an example of displaying a list of search results as desired products for purchase.

FIG. 15 is a screen view showing an example in which a product for which a consumer has made a purchase reservation is displayed clearly on a shopping cart.

FIG. 16 is a diagram showing an example of a purchase confirmation screen.

FIG. 17 is a screen diagram illustrating a function of inquiring homonyms in the recognition decoder unit.

FIG. 18 is a screen diagram illustrating a search function for different categories.

FIG. 19 is a screen diagram illustrating a search function for abstract word procedure words.

FIG. 20 is a screen diagram showing an initial state in which the user has accessed the shopping site screen.

FIG. 21 is an explanatory diagram of a form in which a product to be purchased is stored in a shopping cart.

FIG. 22 is an explanatory diagram illustrating the establishment of a purchase contract for a product stored in a shopping cart.

FIG. 23 shows a destination registration screen for purchased products and a payment method confirmation screen.

FIG. 24 is a screen diagram for prompting another utterance when the uttered voice cannot be recognized.

FIG. 25 is a screen diagram when a spoken voice can be sufficiently recognized but there are a plurality of candidates for the spoken word.

[Explanation of symbols]

1 user terminal 2 shopping site server 3 voice server 4 Internet 11 Voice input section 12 Acoustic analysis section 14, 21, 31 Network card section 15 Recognition decoder section 151 Speech recognition matching rate calculation means 152 Means of reply 153 Multiple candidate presentation section 16 Browser control unit 17 Acoustic model database 18 Speech recognition database 182, 332 dictionary database 184,334 Grammar Database 22 Web control 221 Procedure word receiving means 222 Product Search Method 223 Product search result creation means 224 Cart storage means 225 Commodity Transaction Performance Means 23 product database 24 Order database 25 shopping cart 32 Voice System Management Department 33 Common database for voice recognition 34 Site Information Database

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) G10L 15/00 G10L 3/00 551A 15/22 561C F term (reference) 5B075 KK07 KK34 KK37 ND20 NK02 PP07 PP25 PQ02 PQ46 UU40 5D015 KK02

Claims (12)

[Claims] 1. The content uttered by a user,
The procedure word recognized by the user terminal is received through the communication network; the product database is searched based on the procedure word, and the product search result matching the search is referred to by the user terminal; Is a voice-recognized intention to select a product and stores the product specified by the user from the product search results in the cart; purchase of the product contained in the cart uttered by the user An online shopping method, characterized in that a sale is established based on an indication of intention. 2. The content uttered by the user,
A means for receiving a procedure word recognized by the user terminal through a communication network; a product search means for searching a product database based on the procedure word; and a product search result matching the search in the user terminal Product search result creating means for enabling a product; and a product specified by the user uttered from the product search results,
Shopping according to the user's designation; and means for storing a commodity in the cart based on the user's intention to purchase the commodity, the commodity transaction fulfilling means for fulfilling the transaction of the commodity accommodated in the cart. site. 3. The voice recognition database is provided in the user terminal; a voice recognition common database including update information of the product database provided in the shopping site; and a voice recognition database in the shopping site via the communication network. The difference between the revision state of the voice recognition database of the user terminal that has accessed and the revision state of the voice recognition common database is detected, and the voice is set to the same revision state as the voice recognition common database. 3. A shopping site according to claim 2, further comprising: an equalization means for updating the contents of the recognition database. 4. The product search means searches the product database as a common noun of a product used by the brand when the procedure word is a brand instead of the common noun of the product. The shopping site according to claim 2 or 3. 5. The product search means searches the product database for common nouns of products indicated by the procedure word, including common nouns of related products used in association with the product. The shopping site according to claim 2 or claim 3. 6. The product search means, when the procedure word indicates an attribute of a product, extracts a common noun of a product having the attribute as an intermediate search word, and searches the product database by the intermediate search word. The shopping site according to claim 2 or 3, wherein 7. The product search means, when the procedure word is a procedural sentence containing an attribute term indicating a plurality of attributes of a product, extracts the attribute term that constitutes the procedural sentence, and extracts the procedural sentence. The common noun of a product having attributes of a plurality of attribute terms is extracted as an intermediate search word, and the product database is searched by the intermediate search word and the extracted plurality of attribute terms. The shopping site according to Item 3. 8. The product search result creating means divides the products searched by the product search means by brand name and includes them in the product search result. Shopping site according to any one of 7. 9. The product transaction fulfillment means leaves a history of a purchase intention display of a product accommodated in the cart, which is performed by voice recognition of a user terminal. The shopping site described in one. 10. Accessing a shopping site,
In a voice recognition device used for a user terminal for a shopping site, which transmits a speech-recognized procedure word with respect to the content uttered by a user to proceed with the purchase of a product; In order to
An acoustic model database of the user registered in advance; means for inputting the speech uttered by the user, and calculating a recognition matching rate of the speech uttered by the user with reference to the acoustic model database; And a voice recognition device for prompting the user to utter again when the recognition recognition rate of the generated voice is lower than a predetermined threshold. 11. When the calculated voice recognition concordance rate is lower than a predetermined threshold, the calculated voice recognition concordance rate is relatively high. The voice recognition device according to claim 10, which is presented. 12. A procedure word recognized by a voice from a user terminal is received via a communication network, a commodity database is searched based on the procedure word, a commodity search result matching the search is created, and the commodity of the user. A voice recognition support device that supports voice recognition by the user terminal with respect to a shopping site that executes a transaction of a product included in the product search result based on the purchase intention display of the product; provided for each of the shopping sites. A voice recognition support device comprising means for managing a common database for voice recognition.