TW202343280A - Voice search device based on user's behavior including a database, a voice module, and a search module, to find a number of objects in which the object feature and the first object condition string match each other within a predetermined condition - Google Patents

Abstract

Disclosed provides a voice search device based on users' behaviors, which includes a database, a voice module, and a search module. The database is configured to store a plurality of objects, wherein each object includes an object feature; the voice module is configured to establish a first voice object condition based on frequent uses of predetermined sentence patterns, wherein the first voice object condition is related to one object feature and converts the first voice object condition, according to search scenarios, into a first object condition string by using voice recognition phrases relevant to the search scenarios; and the search module is configured to compare the first object condition string with the object feature and find a number of objects in which the object feature and the first object condition string match each other within a predetermined condition, thereby producing a search result containing a specific number of the plurality of objects.

Description

Voice search device based on usage habits

The present invention relates to a voice search device based on usage habits, and in particular to a search system for finding voice objects that match each other in predetermined conditions based on voice object search condition strings.

There are many aspects to consider when purchasing real estate, such as: price, location, public-to-facilities ratio, square footage, living functions, property types, etc. A common real estate search method uses the above aspects as search conditions for customers, for example, the total search price is less than one There are tens of millions of objects or objects that can be searched for two bedrooms and one living room. However, the object results displayed after the customer searches through the above-mentioned aspects are still very messy and disorderly. It still takes a lot of effort and energy to browse and filter each object one by one. This kind of The usual search methods cannot fully meet the needs of customers in purchasing considerations.

Furthermore, when purchasing real estate, many customers no longer only consider the conditions of the property itself, but give priority to whether the property where the property is located meets their preferences or needs. In other words, to find a good house, you must first The concept of finding good properties has gradually become a trend. However, the existing real estate search methods are still based on the conditions of the properties themselves, or they can only query what properties are in a certain administrative area. You must click one by one to enter each property category. Knowing the conditions of the property (such as house age, total price range, etc.) will require customers to spend more time to find a satisfactory property, and real estate agents will not be able to smoothly provide information about the properties they sell to customers through the system.

In addition, with the development of speech technology, speech recognition Counting has also been introduced piece by piece to meet the needs of real estate transactions, but voice recognition has always been one of the most troublesome problems. If the results that are often recognized are not what the user wants, over time, the user will no longer use voice recognition to search.

In addition, the condition parameters required when searching for real estate are not as simple as a few strings like ordinary searches. They often require very basic conditions such as administrative region, price range, etc. Otherwise, the search scope is too broad and it is easy to get confused. There is the problem of too many results, which can easily cause overload on the system.

Therefore, in view of the shortcomings of the conventional technology, the applicant invented the "voice search device based on usage habits" of the present invention to improve the above shortcomings.

Therefore, there is a need for a voice search device based on usage habits that can overcome the problems of the conventional art. In order to achieve the above object, the present invention provides a voice search device based on usage habits. The voice search device based on usage habits includes a database, a voice module and a search module. The database is configured to store a plurality of objects, wherein each object includes at least object characteristics; the speech module is configured to establish a first speech object condition based on the frequently used predetermined sentence pattern, wherein the first speech object The condition is related to an object feature, and according to the search context, the first voice object condition is converted into a first object condition string using a speech recognition phrase relative to the search context; and the search module is configured to compare the first An object condition string and the at least object characteristic are used to find the plurality of objects that match each other within a predetermined condition between the at least object characteristic and the first object condition string, thereby generating one of the plurality of objects containing a specific number. Search results.

In order to reduce the speech recognition error rate, the system of the present invention will use the speech recognition phrases relative to the search situation according to the search situation. The voice object conditions are converted into object condition strings, and then the search process is performed.

Using interactive voice guidance, the first voice object conditions that need to be known in order to perform a search can be completed in sequence, which can reduce the problem of too broad a search scope and too many results, and can also reduce the load on the system.

The voice search device based on usage habits provided by the present invention can fully meet the needs of customers in real estate purchase. That is, the voice search device based on usage habits of the present invention searches in an object-first retrieval method. The user inputs his or her information about real estate. After the demand for residential types such as objects, the system presents the results of the object collection. Users can first select the objects they are satisfied with, and then browse the objects for sale, because the objects can often reflect the living conditions such as the quality of life of the objects. , so the property-first search method can effectively allow real estate agents to provide information about the properties they sell to customers through the system, thereby increasing the chances of matching between the two parties.

10: Voice search device based on usage habits

101: User side

105:Database

106: Multiple objects

110: Voice module

120:Search module

130:Output module

106A:Object

106B:Object

A01, A02, A03, A04: Object characteristics

S01, S02: Sellable items

20: System search page

205: Search type options

206: Search type options

210:Search conditions

211:Search conditions

215:Search button

220:Map block

221: mark

222:mark

230: Search results block

231:Project

232:Project

30: Execution process

S301, S302, S303, S304, S305, S306, S307, S308: steps

Figure 1 is a schematic diagram showing a voice search device based on usage habits of the present invention;

Figure 2 is a schematic diagram showing the database of the present invention;

Figure 3 is a schematic diagram showing the system search page of the present invention;

Figure 4 is a schematic diagram showing the system search page of the present invention;

Figure 5 is a schematic diagram showing the system search page of the present invention; and

The present invention will be fully understood from the following embodiments, so that those skilled in the art can implement it accordingly. However, the present invention can also be widely implemented in other embodiments. That is to say, the implementation of the present invention is not limited to its implementation form by the following embodiments, but should be based on the present invention. The scope of the patent application submitted shall prevail.

Please refer to FIG. 1 , which shows a schematic diagram of the voice search device based on usage habits of the present invention. In Figure 1, the voice search device 10 based on usage habits includes a database 105, a voice module 110, a search module 120 and an output module 130. The database 105 stores a plurality of objects 106 and user habit records, and The plurality of objects 106 have at least one object characteristic (detailed in Figure 2), and each user habit record includes a predetermined sentence pattern and the selected number of uses of the predetermined sentence pattern. In the voice search device 10 based on usage habits, the voice module 110 uses a sound receiving and playing device (ie, a microphone, a speaker, and other conventional hardware components required for voice recognition) to perform a third operation based on the frequently used predetermined sentence pattern. A voice object condition is established to receive the first voice object condition regarding the characteristics of the object sent by the user terminal 101, and according to the search context, use the voice recognition phrase relative to the search context to convert the first voice object condition into The first object condition string.

The frequently used predetermined sentence pattern may be determined based on the number of times the predetermined sentence pattern is selected. More specifically, if the number of times the predetermined sentence pattern is selected is the first or the top three, the voice module 110 will designate or randomly designate one of the top three as the frequently used predetermined sentence pattern. However, the voice module 110 still needs to collect and analyze the user's habits before it can sort out the selected usage times of the predetermined sentence pattern. In this regard, before the number of selected uses of the predetermined sentence pattern reaches the predetermined threshold, the voice module 110 will still require the user to select the predetermined sentence pattern, and record the selected predetermined sentence pattern in the database 105. Then the first voice object condition will be established, and only after the number of selected uses of the predetermined sentence pattern reaches the predetermined threshold (for example, 10 times), the selected sentences of the predetermined sentence pattern will be retrieved from the database 105. The selected usage times are among the top three, and the voice module 110 randomly designates one of the top three as the frequently used predetermined sentence pattern.

The user can choose the predetermined sentence pattern and the order of each syllable in the sentence pattern through button clicks, voice selection or other means. Of course, the system can preset several groups for selection. , or defined by the user. The predetermined sentence pattern contains at least two syllables, such as the administrative area and the number of square meters. Of course, it can also be the administrative area and other frequently searched conditions, such as total price, house age, etc. When making the first voice object condition based on a predetermined sentence pattern, if the selected predetermined sentence pattern is an administrative area and a square meter, the system will only accept, for example, "Da'an District, Taipei City", "30 square meters", and the system will first check the first Check whether the syllable meets the definition of the administrative area (it can be Taipei City Da'an District, Taipei City Da'an, Taipei City Da'an District, Taipei City Da'an), and also check whether the second syllable meets the definition of the number of square meters (the number of square meters can be 30 or 30 square meters), Then the establishment of the first voice object condition will continue. The system will first check whether the first syllable meets the definition of an administrative region, mainly to see if there is a name for the administrative region, while the second syllable will check whether it contains a number.

The above search situations include different situations such as renting a house, buying a house, selling a house, etc. Under different search situations, there will be corresponding speech recognition phrases for recognition, and the first object condition string generated will generally belong to the corresponding speech recognition phrases. Speech recognition phrases to reduce erroneous speech recognition. Going a step further, learning functions can also be incorporated.

For example, in the search context of buying a house and selling a house, with the same first voice object condition, such as Shuanglian, the search context of buying a house is more likely to be looking for objects around the Shuanglian MRT station, so the first object condition The string is more likely to be Shuanglian MRT Station, and in the search context for houses for sale, it is more likely to be looking for objects around Shuanglian Street, so the first object condition string is more likely to be Shuanglian Street, not Shuanglianjie transportation station.

After receiving the first object condition string sent by the user terminal 101, the voice module 110 sends the received object condition string to the search engine. module 120, and then the search module 120 compares the object condition string with the object characteristics in the database 105, finds objects that match each other within the predetermined conditions, and sends the matching objects to the output module 130, and then outputs the module 130 sends the search results back to the user terminal 101. Preferably, the search module 120 finds objects that match each other within predetermined conditions and generates search results that include a specific number of objects, where the specific number can be zero or an integer greater than zero.

Please refer to Figure 2, which shows a schematic diagram of the database of the present invention. In Figure 2, the database 105 stores an object 106A, such as a residential building, and an object 106B, such as a luxury building. For example, the object may be a dependent village, a complete residence, a national residence, a collective residential area, a suitable residence, a social residence, or an apartment, a mansion, a residential building, a suite, a tenement house or other residences within a specific area. type. Object 106A has object characteristic A01, for example, the house is 15 years old, object characteristic A02, for example, the number of households in the community to which the object belongs is 20, object characteristic A03, for example, the living function is a shopping mall, and object characteristic A04, for example, a public swimming pool, and the object 106A also has There are sellable items S01 and sellable items S02. Object 106B has object characteristic A01, for example, the house is 15 years old, object characteristic A02, for example, the number of households in the community to which the object belongs is 15, and object characteristic A03, for example, the living function is the market, in which the salable objects of object 106B are zero. For example, the object characteristics can be the number of houses, the age of the house, the state of public facilities, the number of households in the community to which the object belongs, living functions, the number of square feet of the house, the orientation of the house, and the number of floors of the house. Preferably, the database 105 can store data of one or more objects. Preferably, an object may have one or more object characteristics. Preferably, the salable items of the object may be zero or an integer greater than zero.

The correspondence between the object 106A and the salable object S01 and the salable object S02 needs to be completed through some means, because there is no public data that can be queried for these correspondences. Specifically, it is based on the town, town, location, land section location or building area house address, total land transfer area square meters, urban land use zoning, transaction date, transaction building Number, transfer level, total number of floors, building type, main purpose, main building materials, construction completion year, total transferred area of the building in square meters, current layout of the building, whether there is a management organization, total price in yuan, unit price per square meter You can select all data such as feet or object names, or select several types to construct object characteristics of multiple objects, and use this to dismantle the transaction data into multiple fields for search and comparison purposes. Then, based on the housing conditions of the salable object S01 and the salable object S02, that is, comparing the address, total number of floors, and construction completion year in the housing conditions with the object characteristics of multiple objects in the database, to The search is performed to obtain at least one specific object information corresponding to the house condition that satisfies the salable object S01 and the salable object S0, and the establishment of the relationship between the object 106A and the salable object S01 and the salable object S02 is completed.

Please continue to refer to Figures 1 and 2. The search module 120 compares the received object condition string with the object characteristics of the plurality of objects 106 in the database 105 to find objects that match each other within the predetermined conditions, such as , the object condition string is that the house is 15 years old. The search module 120 compares the object characteristics in the database 105 and finds out that the object characteristic A01 of object 106A and object 106B is the house that is 15 years old, so it meets the comparison conditions. That is, object 106A and object 106B, and then the search module 120 sends the object 106A and the object 106B to the output module 130 to generate search results.

In the present invention, the search module 120 will compare within predetermined conditions, and the predetermined conditions include (1) the object condition string is a numerical value, the characteristics of the residential group are the same as the object condition string, and the search results display objects that completely match; (2) The object condition string is a numerical value. If the characteristics of the residential group are different from the object condition string, the difference range is the range of the object condition string plus or minus the predetermined value. The search results display partially matching objects; (3) The object condition string is Text, the characteristics of the residential group are the same as the condition string of the object, and the search results show objects that completely match; and (4) The condition string of the object is text, and the characteristics of the residential group are different from the condition string of the object, then the search results show that the characteristics of the residential group are the same as those of the object condition string. Objects that partially match the similar object condition string. For example, if the object condition string is 20 households in the community to which the object belongs, the search module 120 will compare the object characteristics in the database 105 with the search range of 15 to 25 of the value 20 plus or minus 5, and then find out the residential group. The object 106A is characterized by 20 households in the community to which the object belongs, and the object 106B is characterized by the residential group having 15 households in the community to which the object belongs. If the object condition string is the living function and is the market, the search module 120 in addition to finding the characteristics of the residential group has In addition to the objects 106B in the market, objects 106A with the characteristics of a residential group and a shopping mall will also be found with similar characteristics. Among them, markets, stores, etc., which have similar conditions, need to be summarized and build a database in advance.

Please refer to Figure 3, which shows a schematic diagram of the system search page of the present invention. In Figure 3, the system search page 20 includes search type options 205~206, search conditions 210~211, and search buttons 215. After the user selects the type option and enters the search conditions, the user clicks the search button 215, and then the object condition string is sent to the voice module of the system of the present invention. Preferably, the search type option 205 and the search condition 210 are required input data, and the search type option 206 and the search condition 211 are optional input data. If the user wants to input multiple search data, he can click on the system search page 20 "+" to generate new search type options and search criteria fields.

Please refer to Figure 4, which shows a schematic diagram of the system search page of the present invention. In Figure 4, the system search page 20 further includes a map block 220 and a search result block 230. The map block 220 displays the location of the search result object on the map. For example, the object 106A is represented by a mark 221 and the object 106B. It is represented by mark 222, and the search result block 230 displays the name of the object of the search result, for example, object 106A is represented by item 231 and object 106B is represented by item 232. For example, after the user sends the object condition string, for example, the house is 15 years old to the voice module of the system of the present invention, the search module compares the database with The output module sends the object 106A with the characteristics of the residential group A01, for example, the house is 15 years old, such as a certain suitable residence, and the object 106B, such as a certain luxury building, back to the system search page 20, where the search module finds the matching object after It will then confirm whether the matching objects have salable objects. If the matching objects do not have salable objects, they will be included in the ineligible objects and removed from the comparison results. That is, the search results will not show items that meet the comparison conditions. Objects without any salable object data. In this example, object 106A has salable object S01 and salable object S02, and object 106B has salable object S01, so they are all objects that have salable objects and meet the comparison conditions. In addition, , the search module will also sort the objects in the search results from most to least according to the number of salable object data in the objects that meet the comparison conditions. For example, object 106A has more salable objects than object 106B, so object 106A is sorted. Higher than object 106B, the user can thereby know which objects meet the search criteria and have sellable objects, and at the same time, the user can give priority to objects with more sellable objects.

Please refer to Figure 5, which shows a schematic diagram of the system search page of the present invention. Figure 5 is a continuation of Figure 4. After the user obtains the preliminary search results, the user then performs an operation on the second voice object condition related to one of the object characteristics, and the voice module 110 uses the search context relative to the search context. The speech recognition phrase converts the second speech object condition into a second object condition string, and then uses the second object condition string to perform a re-search in the preliminary search results. Since the data of the preliminary search results has been stored in the user end, so re-searching will only read the user-side cache and will not re-access the database to reduce the load on the database.

In more detail, the search module compares the second object condition string with an object characteristic of an object that meets the comparison condition in the preliminary search results, and finds out that at least one residential group characteristic and the second object condition string are in the predetermined condition. objects that match each other, and a second search result containing an object that meets the comparison condition is generated. If the search module compares the second object condition string with the search After searching for an object characteristic of an object that meets the comparison conditions in the results and finding an object that contains at least one residential group characteristic and a second object condition string that match each other within the predetermined conditions, the search module is removed from the search results. Objects that do not match will be displayed in the re-search results. For example, the user sends the second object condition string, such as a swimming pool, and in the preliminary search results, only object 106A has the residential group characteristics A04, such as a public swimming pool, so the system search page 20 is based on the second object condition. The string removes the object 106B from the map block 220 and the search result block 230, leaving only the matching object 106A, which is the re-search result. Preferably, if the search module does not find an object characteristic and the second object condition string that match each other within the predetermined conditions after comparing the second object condition string with the object characteristics that meet the comparison conditions in the preliminary search results. object, the re-search results will be the same as the initial search results.

The present invention is disclosed with reference to the above-mentioned preferred embodiments and examples. Readers should understand that these examples are for description rather than limitation. Those skilled in this art can make various combinations and modifications without departing from the spirit and scope of the invention, and they should still be within the scope of the patent of the invention.

10: Voice search device based on usage habits

101: User side

105:Database

106: Multiple objects

110: Voice module

120:Search module

130:Output module

Claims (10)

A voice search device based on usage habits, including: A database configured to store a plurality of objects and a plurality of user habit records, wherein each object includes an object characteristic, and each user habit record includes a predetermined sentence pattern and a selected one of the predetermined sentence pattern. number of uses; a communication module; A voice module configured to establish a first voice object condition based on the frequently used predetermined sentence pattern through the communication module and using a sound receiving and playing device, wherein the predetermined sentence pattern contains at least two syllables And the first voice object condition is related to an object feature, and then according to a search context, using a voice recognition phrase relative to the search context, the first voice object condition is converted into a first object condition string, wherein often used The predetermined sentence pattern may be determined based on the number of times the predetermined sentence pattern is selected; and A search module configured to compare the first object condition string and the object characteristics, find the plurality of objects that the object characteristics and the first object condition string match each other within a predetermined condition, and generate Contains a specific number of search results for one of the plurality of objects. The voice search device based on usage habits as described in item 1 of the patent application, wherein the mutual agreement within the predetermined conditions means that the first object condition string falls within a difference range of the characteristics of at least one residential group. . As for the voice search device based on usage habits described in item 2 of the patent application, the mutual agreement within the predetermined conditions includes: (1) the first object condition string is a numerical value, and the at least one residential group characteristic and The first object condition string is the same, and the search results display objects that completely match; (2) the first object condition string is a numerical value, and the at least one residential group characteristic is consistent with the first If the object condition strings are different, the difference range is the range of the first object condition string plus or minus a predetermined value, and the search results display partially matching objects; (3) the first object condition string is text, and the at least one The characteristics of the residential group are the same as the first object condition string, and the search results display objects that completely match; and (4) the first object condition string is text, and the at least one residential group characteristic is consistent with the first object condition string. If different, the search results display objects that partially match the characteristics of the at least one residential group and the first object condition string. The voice search device based on usage habits as described in item 3 of the patent application, wherein the voice module is configured to receive a second voice object condition related to the characteristics of the object, and according to the search context, use the information relative to the The speech recognition phrase of the search context converts the second speech object condition into a second object condition string. The voice search device based on usage habits as described in item 4 of the patent application, wherein the search module is configured to compare the second object condition string with the plurality of objects in the search results that meet the comparison condition. The object characteristics, find the plurality of objects that match the characteristics of at least one residential group and the second object condition string within the predetermined condition, and generate a re-search result including the plurality of objects that meet the comparison condition. , wherein if the search module compares the second object condition string with the object characteristics of the plurality of objects that meet the comparison conditions in the search results, it finds the characteristics of at least one residential group and the second object. If the condition strings match each other within the predetermined conditions, the search module will remove the non-matching objects from the search results and display the matching objects in the re-search results. The voice search device based on usage habits as described in item 5 of the patent application, wherein the comparison of the second object condition string is generated by reading the comparison of the first object condition string from cache data. data without re-accessing the data in the database. A voice search device based on usage habits as described in item 5 of the patent application, wherein the database is further configured to store a salable object information for each object and the search module is configured to match matching conditions. The plurality of objects and the salable object data they possess, and the search results and the re-search results are sorted from the largest to the smallest according to the quantity of the salable object data in the plurality of objects that meet the comparison conditions. of the plurality of objects. For example, in the voice search device based on usage habits described in Item 5 of the patent application, if there is no information on the salable objects among the plurality of objects that meet the comparison conditions, the search results and the re-search results will not be consistent. The plurality of objects of the comparison condition. As described in Item 4 of the patent application, a voice search device based on usage habits, wherein the object characteristics are selected from a number of houses, a house age, a public facility status, a number of households in the community to which the object belongs, a living function, a A group consisting of the number of square feet of a house, the orientation of a house, the number of floors of a house, and their combinations. As for the voice search device based on usage habits as described in item 1 of the patent application, the voice module will ask the user to select the predetermined sentence pattern before the number of times the predetermined sentence pattern is selected reaches a predetermined threshold. , the first voice object condition will be established. After the selected usage times of the predetermined sentence pattern reach the predetermined threshold, the selected usage times of the predetermined sentence pattern will be among the top three, and the voice model will The group will randomly designate one of the first three as the frequently used predetermined sentence pattern.

Images