JP2017059014A - Analysis device and data retrieval device, as well as control method of analysis device and data retrieval device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an analysis device, data retrieval device, a control method of the analysis device and data retrieval device, and a program thereof that can show the way to facilities and the like meeting desires of retrievers even when an address, name, telephone number and the like are not known.SOLUTION: A query is received from a microphone 15 of a smart phone 10. Voice data representing the query is transmitted from the smart phone 10 to a voice recognition device 2, where the voice data is converted to character string data. A query of the character string data is next transmitted to an analysis device 3, where the query is determined whether at least two words representing an object and predicate respectively are included in the query. When the given query includes, for example, only one word, and the word is the object, a word representing the predicate is inquired, and on the contrary, when the word is a word representing the predicate, a word expressing the object is inquired. A query including words representing the object and predicate respectively is obtained. In a retrieval device 4, data retrieval is conducted using both the object and the predicate.SELECTED DRAWING: Figure 7

Description

  The present invention relates to an analysis device, a data search device, and a control method thereof.

  In a navigation device that guides a route to a destination, a plurality of search methods such as name search, address area search, and telephone number search may be prepared. First, the search method is set (selected), and then it proceeds to specific data input. If it is a name search, a name search is set (selected) first, and then a specific facility name is input. If it is an address area search, the address area search is first set (selected), and then the process proceeds to input of a specific address. If it is a phone number search, the phone number search is set (selected) first, and then the process proceeds to input of a specific phone number. Japanese Patent Laid-Open No. 2004-151867 describes a technique in which a keyword is searched for by using only a keyword without having to set the search method in advance by storing the keyword and the search method in association with each other.

JP 2009-258805 A

  In the navigation device described in Patent Document 1, if a keyword specifying a facility such as a name, an address, and a telephone number can be input, it is possible to search for a facility that includes the keyword or is related to the keyword. . However, it may be difficult to enter keywords, especially when it is unfamiliar with the surrounding geography.

  An object of the present invention is to be able to guide a facility or the like along a searcher's query (inquiry, request, request, question) without knowing the name, address, telephone number, or the like.

  Another object of the present invention is to make it possible to use spoken words as queries.

  The analysis apparatus according to the present invention includes a word satisfaction determination means for determining whether or not character string data representing a given query includes at least two words corresponding to the object and the predicate, and a word corresponding to the object Or it is provided with a word satisfaction request means for requesting input of a word corresponding to an object not included or a word corresponding to a predicate when it is determined that a word corresponding to the predicate is not included.

  The present invention also provides a control method suitable for the analysis apparatus. An analysis apparatus control method according to the present invention is a method for controlling an analysis apparatus including a word satisfaction determination means and a word satisfaction request means, wherein the word satisfaction determination means adds an object and a predicate to character string data representing a given query. It is determined whether or not at least two words corresponding to each of the above are included, and it is determined by the word satisfaction request means that a word corresponding to the object or a word corresponding to the predicate is not included in the character string data. In this case, it is requested to input a word corresponding to an object not included or a word corresponding to a predicate.

  The present invention further provides a program for causing a computer to execute the processing of the analysis apparatus described above. The analysis program according to the present invention includes a word satisfaction determination process for determining whether or not character string data representing a given query includes at least two words corresponding to the object and the predicate, and a word corresponding to the object Or, when it is determined that a word corresponding to the predicate is not included, the computer executes word satisfaction request processing for requesting input of a word corresponding to the object not included or a word corresponding to the predicate. is there. The present invention also provides a computer-readable recording medium (CD-ROM, DVD-ROM, hard disk, etc.) on which the program is recorded.

  The present invention also provides an analysis system. The analysis system according to the present invention includes a word satisfaction determination device that determines whether or not character string data representing a given query includes at least two words corresponding to the object and the predicate, and a word corresponding to the object. Or, when it is determined that a word corresponding to a predicate is not included in the character string data, a word satisfaction requesting device that requests input of a word corresponding to an object not included or a word corresponding to a predicate is provided. ing.

  The analysis apparatus and analysis system according to the present invention determines whether a given query is a data search by determining whether or not character string data representing the given query includes at least two words corresponding to an object and a predicate. Analyzes whether or not it contains content suitable for. This is because if the object and the predicate are known, the searcher who performs the data search knows what the user wants to do.

  According to the present invention, when it is determined that a word corresponding to an object or a word corresponding to a predicate is not included in a given query, a word corresponding to an object not included or a word corresponding to a predicate is not included. Input is required. When the searcher responds to this request, a query including at least one of each of the object and the predicate is given, and it becomes easy to obtain a data search result that matches the searcher's request.

  Preferably, the analysis device performs morphological analysis processing on character string data representing a given query, and determines morphological analysis means for determining one or more words included in the character string data and the part of speech of each word, and the morphological analysis means described above. An assignment unit is further provided for assigning a word determined to be a noun to a word corresponding to the object and assigning a word determined to be a verb, an adjective, or an adjective verb to a word corresponding to the above description word. Natural sentences can be used as the given query.

  Preferably, the analysis device further includes voice input means for receiving the input of the query by voice, and voice recognition means for converting voice data into character string data. Spoken words spoken by people can be used for queries.

  The present invention also provides a data retrieval apparatus. A data search apparatus according to the present invention includes query input means for receiving input of character string data representing a query, and word satisfaction for determining whether or not at least two words corresponding to an object and a predicate are included in the accepted query. When it is determined that the determination means includes at least two words corresponding to the object and the predicate in the accepted query, the word corresponding to the object and the word corresponding to the predicate are used as search keywords. Corresponds to the word or predicate corresponding to the object not included in the query when it is determined that the search means to search and the accepted query do not include the word corresponding to the object or the word corresponding to the predicate Word satisfaction requesting means for requesting input of a word to be processed.

  The present invention also provides a control method suitable for the data retrieval apparatus. A method for controlling a data search apparatus according to the present invention is a method for controlling a data search apparatus comprising query input means, word satisfaction determination means, search means, and word satisfaction request means, and a character string representing a query by the query input means. Accepts data input, determines whether or not at least two words corresponding to the object and predicate are included in the character string data by the word satisfaction determination means, and at least two corresponding to the object and predicate When it is determined that the word is included in the character string data, the search means searches the data using the word corresponding to the object and the word corresponding to the predicate as the search keyword, and corresponds to the object. When it is determined that the character string data does not contain a word or a word corresponding to a predicate, Is a request for input of a word corresponding to the word or the predicate corresponding to object that is not contained by satisfaction requesting means.

  The present invention also provides a program for causing a computer to execute the processing of the above-described data search device, and a computer-readable recording medium (CD-ROM, DVD-ROM, hard disk, etc.) on which the program is recorded.

  According to the present invention, when a given query includes a word corresponding to an object and a word corresponding to a predicate, the process proceeds to data search, and on the other hand, a word corresponding to the object or a word corresponding to a predicate is included. If it is determined that there is no such word, input of a word corresponding to an object not included or a word corresponding to a predicate is requested. This is because if the object and predicate are known, the searcher knows what he / she wants to do (request and request of the searcher). When the searcher responds to this request, a query including at least one of each of the object and the predicate is given, and it becomes easy to obtain a data search result that matches the searcher's request.

  Data search is performed by using both the word corresponding to the object and the word corresponding to the predicate as search keywords. There may be a plurality of objects and predicates used as search keywords. In any case, at least one word corresponding to the object and one word corresponding to the predicate are required to be included in the query, and if the query does not satisfy this requirement, the requirement is required to be satisfied. .

  In one embodiment, the character string data is subjected to a morphological analysis process, and one or a plurality of words included in the character string data and a part of speech of each word are determined, and a word identified as a noun by the morphological analysis means Is assigned to the word corresponding to the object, and a word determined as one of a verb, an adjective and an adjective verb is further assigned to a word corresponding to the above description word. Natural sentences can be used for queries.

  Preferably, voice input means for receiving the query input by voice and voice recognition means for converting voice data into character string data are further provided. Spoken words spoken by people can be used for queries.

  In one embodiment, a plurality of explanatory text data for each of a plurality of facilities is stored in the storage device, and a word corresponding to the object is selected from the plurality of explanatory text data stored in the storage device. One or a plurality of explanatory text data is searched using a word corresponding to the predicate as a search keyword. Facility information (facility name, facility location, etc.) corresponding to the explanatory text data found by the search is output from the output means as a search result for the query. The output of the search result may be an audio output (the output means is a speaker), or may be a character display or an image display on the display screen (the output means is a display device). Facilities that meet the searcher's needs can be introduced to the searcher. The facility may be a station, a sightseeing spot, a park, an art museum, a museum, an aquarium, a store, or the like, or may be a narrower range, for example, each of a plurality of sales floors in one store.

  In one embodiment, the search means performs processing for each of a plurality of search keywords including a word corresponding to the object word and a word corresponding to the above description word, with each of the plurality of explanatory text data being processed. Appearance degree calculating means for calculating the degree of occurrence representing the presence ratio of the search keyword in the explanatory text data based on the number of target explanatory text data, and each of the plurality of explanatory text data as processing targets, For each search keyword, density calculating means for calculating the density representing the degree of concentration of the search keyword in the description text data, and the appearance degree calculated for each of the plurality of search keywords in each of the plurality of description text data Multiply the density and add the value obtained by multiplication for each explanation data. Comprises a matching degree calculation means for calculating a coincidence degree for the search term number, the size of the matching degrees calculated by the matching degree calculating unit is used in search results for the query. A large degree of match is easily given to explanatory text data having a large value obtained by multiplying the appearance degree and density of a plurality of search keywords, that is, explanatory text data appropriately including a plurality of search keywords (objects and predicates). The magnitude of the degree can be used as a search result for the query. For example, the facility information corresponding to the explanatory text data having the highest degree of match is output as a search result for the query.

  Preferably, the descriptive text data for the plurality of facilities stored in the storage device includes predicate data representing a human action on the facility. For example, in the case of an aquarium facility, the explanatory text data includes predicate data (verb, adjective or adjective verb) representing a person's action on the aquarium, such as “see”, “play”, “appreciate”. In the case of a hot spring facility, the explanatory text data includes predicate data representing a person's action on the hot spring, such as “enter” or “take a bath”. For example, if you are going to an aquarium, the query will naturally include “Aquarium”, and what you want to do in the aquarium, typically predicates such as “see” and “appreciate” fish Is likely to be included. By including the predicate data representing the behavior of the person with respect to the facility in the explanatory text data to be searched, it is possible to improve the accuracy of the search result for the query, particularly when the query is accepted by spoken language.

  In one embodiment, erroneous recognition determination means for determining whether or not a word included in the query is included in the dictionary data, and the word included in the query by the erroneous recognition determination means is included in the dictionary data. First, correction candidate presenting means for presenting word candidates with correct misrecognition when it is determined that they are misrecognized is provided. When a query is received by voice input, misrecognition is particularly likely to occur. When a recognition error occurs, a correction candidate is presented so that the recognition error can be corrected. Data retrieval can be performed with the corrected words.

It is a block diagram which shows the hardware constitutions of a data search system. It is a block diagram which shows the other example of the hardware constitutions of a data search system. It is a block diagram which shows the further another example of the hardware constitutions of a data search system. It is a block diagram which shows the further another example of the hardware constitutions of a data search system. It is a block diagram which shows the further another example of the hardware constitutions of a data search system. Indicates a smartphone. It is a block diagram which shows the hardware constitutions of a data search system with the electrical constitution of a smart phone. It is a flowchart which shows the process of a data search system. It is a flowchart which shows the process of a data search system. It is a flowchart which shows the process of a data search system. The processing results of the search process are summarized in a table format. Indicates search target data.

(1) Hardware Configuration FIGS. 1 to 5 are block diagrams schematically showing the hardware configuration of the data search system. Each of the data retrieval systems shown in FIGS. 1 to 5 accepts input of a query (inquiry, request, request, question) given by voice, performs data retrieval according to the accepted query, and outputs a data retrieval result. To do.

  The data search system shown in FIG. 1 includes an input / output terminal 1, a voice recognition device 2, an analysis device 3, and a search device 4.

  The input / output terminal 1, the speech recognition device 2, the analysis device 3, and the search device 4 are all realized by a computer device in which a dedicated program is installed. For example, the input / output terminal 1 includes a CPU, a memory, a storage device, a communication device, and the like, and functions as the input / output terminal 1 by the CPU executing a dedicated program stored in the storage device. The same applies to the speech recognition device 2, the analysis device 3, and the search device 4. The program can be recorded on a tangible recording medium (CD-ROM, DVD-ROM, hard disk, etc.) (not shown). The program recorded on the recording medium is installed in a storage device provided in each of the input / output terminal 1, the speech recognition device 2, the analysis device 3, and the search device 4.

  The input / output terminal 1, the speech recognition device 2, the analysis device 3, and the search device 4 may exist in one place or may exist in different places. When they exist in different places, the input / output terminal 1, the speech recognition device 2, the analysis device 3, and the search device 4 are connected to each other via a network such as the Internet (cloud computing). The input / output terminal 1, the speech recognition device 2, the analysis device 3 and the search device 4 are all equipped with a communication device for transmitting and receiving data.

  A query is input from the microphone 6 provided in the input / output terminal 1. A query accepted by the microphone 6, that is, voice data (voice query) including the contents of the query is transmitted from the input / output terminal 1 to the voice recognition device 2 through the network, where the voice data is converted into character string data.

  Character string data representing a query (character string query) is then given from the speech recognition device 2 to the analysis device 3. The analysis device 3 includes a morphological analysis unit 3A, a word satisfaction determination unit 3B, and a word satisfaction request unit 3C. Morphological analysis processing is performed on the character string data in the morphological analysis unit 3A. The morpheme analysis process is a process of dividing a character string (sentence, sentence) represented by the character string data into morphemes (minimum unit words that have meaning in the language) and determining their respective parts of speech. As will be described in detail later, the word satisfaction determination unit 3B determines whether the character string represented by the character string data includes a word corresponding to the object and a word corresponding to the predicate. At least one of the word corresponding to the object and the word corresponding to the predicate is included in the character string as a requirement for the search processing to be executed in the subsequent search device 4. When it is determined that at least one of the word corresponding to the object and the word corresponding to the predicate is not included in the character string, the word satisfaction request unit 3C selects the word or predicate corresponding to the object not included. A satisfaction request process (reverse query) for inquiring (inquiring) a word corresponding to is also performed. The reverse query is output from the speaker 7 provided in the input / output terminal 1, for example.

  When both the word corresponding to the object and the word corresponding to the predicate are included in the character string query, the word corresponding to the object and the word corresponding to the predicate are input to the search device 4. The search device 4 executes data search processing using a word corresponding to the object and a word corresponding to the predicate as search keywords. Data (search result) found by the data search process in the search device 4 is transmitted from the search device 4 to the input / output terminal 1.

  The input / output terminal 1 that has received the search result from the search device 4 executes processing for outputting the search result. Most simply, it is possible to execute a process of outputting the search result transmitted from the search device 4 as it is from the speaker 7 by voice, or outputting an image on the display screen of the display device included in the input / output terminal 1. Instead of outputting the search result transmitted from the search device 4 as it is, a predetermined process using the search result transmitted from the search device 4 is executed in the input / output terminal 1, and the execution result is input to the input / output terminal 1. Can also be output. For example, if the search result transmitted from the search device 4 is address data (latitude and longitude indicating the location of a specific facility), the input / output terminal 1 outputs a route from the current location to the point specified by the address data. (Display on the display screen).

  The data search system shown in FIG. 2 includes an input / output terminal 1, a voice recognition device 2, and a computer system 5A. The computer system 5A is a system (personal computer, workstation, etc.) including a CPU, memory, storage device, input device, display device, CD-ROM drive, input / output interface, and the like.

  A program (analysis program, search program) for executing the processing of the analysis device 3 and the processing of the search device 4 is stored in the storage device of the computer system 5A. When the analysis program and the search program are executed by the CPU of the computer system 5A, the computer system 5A functions as the analysis device 3 and the search device 4 described above. The computer system 5A that executes the processing of the analysis device 3 and the processing of the search device 4 can also be positioned as a data search device.

  The data search system shown in FIG. 3 includes an input / output terminal 1 and a computer system 5B. Programs (speech recognition program, analysis program, and search program) for executing the processes of the voice recognition device 2, the analysis device 3, and the search device 4 described above are stored in the storage device of the computer system 5B. The computer system 5B functions as the voice recognition device 2, the analysis device 3, and the search device 4 described above. The computer system 5B that executes the processing of the speech recognition device 2, the processing of the analysis device 3, and the processing of the search device 4 can also be positioned as a data search device.

  The data retrieval system shown in FIG. 4 includes an input / output terminal 1 and a computer system 5C. 3 is different from the data retrieval system shown in FIG. 3 in that the voice recognition program is stored in the storage device of the input / output terminal 1 instead of the storage device of the computer system 5C. The input / output terminal 1 performs voice recognition processing on the voice query input from the microphone 6 to generate a character string query. The character string query generated in the input / output terminal 1 is transmitted to the computer system 5C via the network.

  The data search system shown in FIG. 5 includes an input / output terminal 1, and a speech recognition program, an analysis program, and a search program are stored in the storage device of the input / output terminal 1. This is a mode in which the input / output terminal 1 itself functions as a data search system. The data to be retrieved may be stored not in the storage device of the input / output terminal 1 but in a storage device that can be connected through a network.

  As described above, the data retrieval system is constituted by one or a plurality of hardware. As shown in each of FIGS. 1 to 5, each of a plurality of processes performed in the data retrieval system can be executed on separate hardware, or some or all of the plurality of processes can be performed in a single manner. It can also be executed in hardware.

(2) Specific Example of Data Search System Hereinafter, a specific example of the data search system will be described. Here, a data search system having the hardware configuration shown in FIG. 1 will be exemplified.

  FIG. 6 shows a smartphone (mobile phone) 10 used as the input / output terminal 1. FIG. 7 shows a block diagram of the data search system together with details of the electrical configuration of the smartphone 10.

  The smartphone 10 can be used for calls (calls) and provides various functions to users according to installed application programs. One of application programs installed on the smartphone 10 is a navigation program. FIG. 6 shows an example of a screen displayed on the display screen of the smartphone 10 on which the navigation program is executed.

  The navigation program is a program for showing the current location to the user of the smartphone 10 and guiding the route to the destination. The current location (latitude and longitude) is determined using, for example, GPS (Global Positioning System).

  When the navigation program is activated, a road map is displayed on the display screen of the smartphone 10, and the current location determined by the GPS is superimposed on the road map. Data representing the road map is downloaded to the smartphone 10 via the network. The route guidance to the destination is indicated by superimposing a symbol indicating the route on the road display included in the road map.

  With reference to FIG. 7, the smartphone 10 includes a CPU 11, and the operation of the smartphone 10 is totally controlled by the CPU 11. The CPU 11 includes a memory 12 for temporarily storing road map data, voice guidance data, etc., a navigation program, an operating system (OS) program, a hard disk (HD) 13 for storing other programs and data, voice communication and A communication device 14 for data communication through a network (such as the Internet), a microphone 15 for inputting sound, a display device 16 for displaying a road map, a speaker 18 for outputting sound, and the GPS described above A GPS device 19 for use is provided. A touch panel 17 is formed on the display screen of the display device 16, and a destination can be input by touching the touch panel 17.

  As described above, the navigation program has a basic function of guiding the route from the current location to the destination. For this purpose, the destination must be input to the smartphone 10. If the destination is known, the destination can be input to the smartphone 10 by its address, telephone number, facility name, designation of the destination on the road map (screen touch), and the like. However, especially when it is unfamiliar with the surrounding geography, it may take time or trouble to enter the destination.

  As will be described below, by using the data search system, it is stored in the facility information database 22 (details will be described later) connected via the network without inputting the destination itself into the smartphone 10. One or a plurality of facilities can be introduced to the user of the smartphone 10 from the plurality of facility information data. As explained below, a query is entered by voice, and facilities that match the query (along the query) are introduced.

  The input query is required to include “object” and “predicate” as described below (the object and predicate are required to be used as search keywords). No subject is required. This is because in the case of the smartphone 10, the subject is the user of the smartphone 10, and it is determined in advance that the subject of the query is the user. Nor is it required for the time of the query (when). This is because the time when the query is input is the time. Of course, the subject may be included in the query depending on the language, but the word representing the subject is not used in the data retrieval system.

  As described above, the data search system includes not only the smartphone 10 but also the voice recognition device 2, the analysis device 3, and the search device 4. The analysis device 3 includes a dictionary database 21, and the search device 4 includes a facility information database 22.

  FIG. 8 to FIG. 10 are flowcharts showing processing performed in the smartphone 10, the speech recognition device 2, the analysis device 3, and the search device 4 from when a query is received until search processing is performed. The hardware for executing each process is shown in association with each process in the flowchart.

  First, a query is received by the smartphone 10 (step 31). The query is accepted by uttering a voice (word) toward the microphone 15 included in the smartphone 10.

  The voice query (voice data) is transmitted from the smartphone 10 to the voice recognition device 2 via the network, and converted into character string data in the voice recognition device 2 (voice recognition processing, step 32).

  Character string data is given from the speech recognition device 2 to the analysis device 3. The analysis device 3 performs a morphological analysis process on the character string data (step 33). In the morphological analysis process, the character string data is divided into words (morphemes), and the part of speech (noun, verb, adjective, adjective verb, particle, article, etc.) of each word is determined. Each word is assigned (associated) with the determined part of speech.

  Of the words that have been divided in the morphological analysis process and whose part of speech has been identified, words having part of speech of noun, verb, adjective, and adjective verb are selected (extracted) (step 34). Other part-of-speech words such as particles, articles, and prepositions are not used in the following processing. For example, if the query is “I want to go to a delicious pasta restaurant”, “pasta” (noun), “no” (particle), “delicious” (adjective), “store” (noun), “ni” (particle) , “Go” (verb), “tai” (auxiliary verb). The four words "pasta" (noun), "delicious" (adjective), "store" (noun), and "go" (verb) are used for the following processing: "no", "ni", and "tai" Is not used in the following process (ignored or discarded).

  If the query is in English, for example, if the query is “I want to go to a delicious shop of pasta.”, “Go” (verb), “delicious” (adjective), “shop” (noun), and “pasta” The four words (noun) are used for the following processing. The subject “I”, the definite article “a”, and the preposition “of” are not used in the following processing. Since it is known that the query itself represents a request, “want” is basically not used in the following processing. In English, “to” is basically not used in the following processing.

  Next, for each of the extracted words, it is determined whether the word is a word stored in the dictionary database 21 provided in the analysis device 3 (step 35). If the word is not stored in the dictionary database 21, there is a possibility that erroneous recognition has occurred in the speech recognition described above. For example, the user has spoken as “I want to eat pasta”, but has been voice-recognized as “I want to eat pasta”. If “Pasata” is a word that is not stored in the dictionary database 21, the process proceeds to a false recognition correction process (N0 in step 35).

  When a word not stored in the dictionary database 21 is input, a word (correction candidate) similar to the word is transmitted from the analysis device 3 to the smartphone 10 and displayed on the display screen of the display device 16 in a list. (Step 36). A correct word is selected from the correction candidates displayed in the list by, for example, tapping the display screen (step 37). The selected correct word is transmitted from the smartphone 10 to the analysis device 3, whereby “Pasata” is corrected to “Pasta”. In the above case, “I want to eat pasta” is treated as a query accepted instead of “I want to eat pasata”. For this process, for example, a misrecognition correction process described in Japanese Patent Application No. 2014-240350 by the applicant can be used.

  The next process branches depending on whether the query includes one or more words (including words corrected by the misrecognition correction process) (step 38).

  First, a case where only one word (a word having a part of speech of noun, verb, adjective or adjective verb) is included in the query will be described (FIG. 9). For example, “pasta”, “aquarium”, “meal” or “want to eat” are accepted as queries.

  It is determined whether one word included in the query is a noun (step 41). The part of speech that has already been determined in the morphological analysis process (step 33) is used to determine whether it is a noun or other (verb, adjective, adjective verb).

  If the word accepted as a query is a noun (YES in step 41) (eg, “pasta”), the word is positioned as the “object”. Of the above-mentioned “object” and “predicate” required as search keywords in the search process, “object” is satisfied, and “predicate” is missing.

  A predicate is asked (step 42). Using the noun “pasta” included in the query, for example, an inquiry (voice data representing the question) such as “What do you want to do with the pasta?” Is created in the analysis device 3 and transmitted to the smartphone 10 from the speaker 18. Output (reverse query). An answer to the inquiry, for example, “I want to eat” (a predicate) is uttered and accepted by the microphone 15 of the smartphone 10 (step 43). It is also determined whether the accepted predicate is stored in the dictionary database 21, and if not stored, the above-described erroneous recognition correction process (steps 35, 36, 37) may be performed.

  Since the “object” and “predicate” have been prepared by the above processing, the search device 4 can proceed to the search processing (step 46). Details of the search process will be described later.

  If a single word is accepted as a query and the word is not a noun (verb, adjective verb or adjective verb) (NO in step 41) (for example, “eat” when “I want to eat” is spoken) "), The word is positioned as a predicate. Of the “object” and “predicate” necessary for the search processing, the “predicate” is satisfied and the “object” is missing.

  The object is asked (step 44). For example, the predicate “eating” included in the query is used, and an inquiry (voice data representing the question) that says “What do you eat?” Is created in the analysis device 3 and transmitted to the smartphone 10. Output from the speaker 18 (reverse query). An answer to the inquiry, for example, “pasta” (object) is spoken and accepted by the microphone 15 of the smartphone 10 (step 45). It is also determined whether or not the accepted object is stored in the dictionary database 21, and if it is not stored, the above-described misrecognition correction process may be performed. When the object and the predicate are prepared, the search process (step 46) is performed. A word representing an object and a word representing a predicate are given from the analysis device 3 to the search device 4.

  Next, a case where a query includes a plurality of words will be described (FIG. 10). For example, “pasta, store”, “I want to eat pasta”, “Delicious pasta”, “I want to see dolphins”, “I want to eat delicious Japanese food”, etc. are accepted as queries.

  The processing branches depending on whether a plurality of words included in the query include only two or more nouns or whether a noun and a predicate are included.

  For example, when “pasta, store” is issued toward the microphone 15 of the smartphone 10, “pasta” (noun) and “store” (noun) are identified by the morphological analysis process (step 33). This is the case when only multiple nouns are included in the query. First, it is determined whether or not each of the plurality of nouns is a word stored in the dictionary database 21 (step 35). If it is not stored in the dictionary database 21, words similar to the word (correction candidates) are displayed in a list on the display screen of the display device 16 of the smartphone 10 as described above, and the correct word is selected from the list. (NO in step 35, steps 36 and 37).

  If the query contains only multiple nouns (YES in step 51), the predicate is asked using the first noun recognized among the multiple nouns, for example "pasta, store" (Step 52). For example, an inquiry “What do you do with pasta?” Is output from the speaker 18 of the smartphone 10 (reverse query). If the query is in the order of “store, pasta”, an inquiry “What do you do with the store?” Is output from the speaker 18. A predicate is accepted by uttering “I want to eat”, “I want to go” or the like (step 53). When two object words “pasta” and “store” and one predicate (for example, “eat” or “go”) are arranged, the process proceeds to the search process (step 54). A word representing the object and a word representing the predicate are given from the analysis device 3 to the search device 4.

  If the query includes both a noun and a non-noun word (verb, adjective, adjective verb) (NO in step 51), the noun included in the query is treated as satisfying the object and the non-noun word as satisfying the predicate. For example, if the above-mentioned “I want to go to a delicious pasta restaurant” is issued to the microphone 15 of the smartphone 10, two objects (“pasta”, “store”) and two predicates (“delicious”, “go” ”) Is included, the search process using these four words is performed (step 54).

  FIG. 11 shows the processing results of the search processing (step 46, step 54) executed in the search device 4 in a tabular form. FIG. 12 shows search target data (facility information data) stored in the facility information database 22 (see FIG. 7) included in the search device 4.

  In order to make it easy to understand, it is assumed that the facility information data to be searched stored in the facility information database 22 is data relating to twelve facilities A to L here. The facility information data includes explanatory text (article, text) data. In the search processing in the search device 4, the 12 pieces of facility information data of the facilities A to L are ranked in the order considered to be suitable as a search result corresponding to the search keyword including the above-described object and predicate. The In accordance with the obtained ranking, the facility with the highest ranking is introduced to the user (the facility name is output from the speaker 18 and the facility name is displayed on the display device 16). Of course, for example, the top three facilities may be introduced. In addition, route guidance to the highest ranking facility may be started.

  Referring to FIG. 11, here, three words (XXX, XXX, .DELTA..DELTA..DELTA.) Are used as search keywords. These search keywords are the object and predicate obtained by the above-described processing. There may be one or more object words used as search keywords. The same applies to predicates. As described above, at least one object and at least one predicate are always included in the search keyword.

  For each of the three search keywords, the appearance degree, density, and degree of match are calculated for each of the facility information data A to L to be searched.

  The “appearance degree” is a value that indicates how much a specific search keyword is included in all the facility information data A to L. The degree of appearance is calculated by the following equation.

  Appearance level = number of specific keywords included in specific facility information data / average value of specific keywords included in all facility information data

  “Number of specific keywords included in specific facility information data”, for example, the number of specific keywords “XXX” included in facility information data A counts the number of keywords “XXX” included in facility information data A It is obtained by doing. “Existence average value of specific keywords included in all facility information data”, for example, the presence average value of specific keywords “XX” included in 12 facility information data of facility information data A to facility information data L is It is obtained by counting the number of specific keywords “XXX” included in all of the facility information data A to facility information data L and dividing this by the number of all facility information data, here twelve.

  “Density” is a value representing how concentrated a specific search keyword exists in each piece of facility information data. The density is calculated by the following formula.

  Density = number of specific keywords included in specific facility information data / number of characters in specific facility information data

  “Number of specific keywords included in specific facility information data”, for example, the number of specific keywords “XXX” included in facility information data A counts the number of keywords “XXX” included in facility information data A It is obtained by doing. The “number of characters of the specific facility information data”, for example, the number of characters of the facility information data A is obtained by counting the number of characters of the facility information data A.

  When the above-described appearance degree and density are calculated for the specific keyword in the specific facility information data, the degree of coincidence for the specific keyword in the specific facility information data is calculated by the following equation.

  Consistency for specific keywords in specific facility information data = appearance level x density

  Depending on the degree of appearance, the existence ratio of specific keywords in specific facility information data based on the number of all facility information data is known. The density indicates the degree of concentration of specific keywords based on each specific facility information data. The degree of match, which is the result of multiplying the appearance level and the density, takes into account both the ratio of the specific keyword to the entire number of facility information data to be searched and the ratio to the facility information data itself to be searched. Value.

Even if the facility information data has a high degree of matching with respect to one search keyword, there is also facility information data with a low degree of matching with respect to other search keywords. As shown in FIG. 11, in the search process, the sum of the matching degrees calculated for each search keyword for each of the above-mentioned specific facility information data is calculated as the facility matching degree (c ₁ + c ₂ + c ₃ ). . Then, the facility information data to be searched is ranked in descending order of the facility matching degree. High ranking is given to facility information data having a large degree of coincidence on average (with good balance) for a plurality of search keywords.

  Referring to FIG. 12, the facility information data includes facility name, major classification, middle classification, minor classification, action classification, classification feature, and explanatory text data related to the feature. Facility information data is roughly classified by major classification, middle classification, and minor classification. The “behavior classification” is a predicate (generally a verb) that represents the action performed by the user for the facility. “Classification feature” is a general description of the facility, and “Feature” is a detailed description of the facility. In particular, the “behavior classification” may include data representing a classification feature and a predicate that is not included in the feature and represents a behavior performed by the user on the facility. The facility information data is registered in advance in the facility information database 22 (FIG. 7).

  For example, if a query “I want to eat where Mt. Fuji can be seen” is given, five search keywords including the objects “Mt. Fuji”, “Place” and “Meal”, and the predicates “Visible” and “I want to” from that query. Is extracted, and facility information data including the five extracted search keywords in a balanced manner is determined (ranked). Even facility information data not including any of a plurality of search keywords can be ranked. The determined facility name in the facility information data is transmitted from the search device 4 to the smartphone 10. In the smartphone 10, a guidance such as “I recommend restaurant XX” is output from the speaker 18 by using the received facility name by the navigation program being executed. By transmitting data representing the facility address (latitude and longitude) as well as the facility name from the search device 4, the route guidance to the facility can be displayed on the display screen of the display device 16 of the smartphone 10. it can.

  The facility matching degree described above may be calculated for the facility information data narrowed down by the narrowing-down process in advance. For example, facility information data within a predetermined distance (for example, 2 km range) from the current location can be extracted from the facility information database 22 in advance, and the rank based on the above-mentioned facility matching degree can be calculated for the extracted facility information data.

  In the above-described embodiment, the search process has been described by taking a navigation program for performing route guidance as an example. However, the above-described data search process can also be used in, for example, sales floor search in one store.

1,1A, 1B I / O terminal 6,15 Microphone (voice input means)
7, 18 Speaker (output means)
2 Voice recognition device (voice recognition means)
3. Analysis device (word satisfaction determination means, word satisfaction request means, morpheme analysis means, allocation means)
4. Search device (search means)
5A, 5B, 5C Computer system (data search device, data search system)
10 Smartphone (input means, output means)
11 CPU
12 memory
13 Hard disk
14 Communication equipment
16 Display device (output means)
17 Touch panel
22 Facility information database

Claims (16)

Word satisfaction determination means for determining whether or not at least two words corresponding to the object and the predicate are included in the character string data representing the given query, and a word corresponding to the object or a word corresponding to the predicate A word satisfaction request means for requesting input of a word corresponding to an object not included or a word corresponding to a predicate when it is determined not to be included in the character string data;
Analysis device. Character string data representing a given query is subjected to morphological analysis processing, and morpheme analysis means for determining one or more words included in the character string data and the part of speech of each word; Assignment means for assigning to a word corresponding to the object and assigning a word determined to be a verb, an adjective or an adjective verb to a word corresponding to the above descriptive word;
The analysis device according to claim 1. A control method for an analysis apparatus comprising a word satisfaction determination means and a word satisfaction request means,
The word satisfaction determination means determines whether the character string data representing the given query contains at least two words corresponding to the object and the predicate,
Corresponding to a word or predicate corresponding to an object not included when the word satisfaction request means determines that a word corresponding to the object or a word corresponding to a predicate is not included in the character string data Require input of words,
Control method of analysis device. A word satisfaction determination process for determining whether or not at least two words corresponding to the object and the predicate are included in the character string data representing the given query, and a word corresponding to the object or a word corresponding to the predicate A word satisfaction request process for requesting input of a word corresponding to an object not included or a word corresponding to a predicate when it is determined not to be included in the character string data;
Analysis program that causes a computer to execute. A word satisfaction determination device for determining whether or not character string data representing a given query includes at least two words corresponding to the object and the predicate, and a word corresponding to the object or a word corresponding to the predicate A word satisfaction requesting device for requesting input of a word corresponding to an object not included or a word corresponding to a predicate when it is determined not to be included in the character string data;
Analysis system. Query input means that accepts input of character string data representing a query,
Word satisfaction determination means for determining whether or not at least two words corresponding to each of an object and a predicate are included in the character string data;
When it is determined that at least two words corresponding to the object and predicate are included in the character string data, data search is performed using the word corresponding to the object and the word corresponding to the predicate as search keywords. When it is determined that the character string data does not include the word corresponding to the search means and the word or predicate corresponding to the object, the word corresponding to the word or predicate corresponding to the object not included A word satisfaction requesting means for requesting input,
Data retrieval device. Character string data representing the query input by the query input means is subjected to morphological analysis processing, and morpheme analysis means for discriminating one or more words contained in the character string data and the part of speech of each word; Assignment means for assigning a word determined to be a word corresponding to the object and assigning a word determined to be a verb, an adjective, or an adjective verb to a word corresponding to the above descriptive word,
The data search device according to claim 6. Voice input means for receiving the query input by voice; and voice recognition means for converting voice data into character string data.
The data search device according to claim 6. A storage device storing a plurality of explanatory data for each of a plurality of facilities;
The search means uses the word corresponding to the target word and the word corresponding to the upper description word as search keywords from among the plurality of description text data stored in the storage device as one or more description text data. Is to search,
The data search device according to any one of claims 6 to 8. An output means for outputting facility information corresponding to the explanatory text data found by the search as a search result for the query;
The data search device according to claim 9. The search means is
Using each of a plurality of explanatory text data as a processing target, for each of a plurality of search keywords including a word corresponding to the object word and a word corresponding to the above description word, the number of explanatory text data of all processing targets is used as a reference. A degree-of-appearance calculating means for calculating a degree of appearance representing the presence ratio of the search keyword in the explanatory text data;
Density calculation means for calculating the density indicating the degree of concentration of the search keyword in the description text data for each of the plurality of search keywords, and a plurality of description text data In each case, the degree of coincidence of the explanation text data for the plurality of search keywords is calculated by multiplying the appearance degree and density calculated for each of the search keywords and adding the value obtained by the multiplication for each explanation text data. A matching degree calculation means for
The degree of match calculated by the match level calculation means is used in the search result for the query.
The data search device according to claim 9. The descriptive text data for the plurality of facilities stored in the storage device includes predicate data representing human behavior with respect to the facilities.
The data search device according to claim 9. A method for controlling a data search apparatus comprising query input means, word satisfaction determination means, search means and word satisfaction request means,
Accepts input of character string data representing a query by the query input means,
Determining whether or not at least two words corresponding to the object and the predicate are included in the character string data by the word satisfaction determination means;
When it is determined that at least two words corresponding to the object and the predicate are included in the character string data, the word corresponding to the object and the word corresponding to the predicate are used as search keywords by the search means. Search for data,
When it is determined that a word corresponding to an object or a word corresponding to a predicate is not included in the character string data, it corresponds to a word or a predicate corresponding to an object not included by the word satisfaction request means Require input of words,
Control method of data retrieval apparatus. Word satisfaction determination processing for determining whether character string data representing a given query includes at least two words corresponding to the object and the predicate,
Data for data retrieval using the word corresponding to the object and the word corresponding to the predicate as the search keyword when it is determined that at least two words corresponding to the object and the predicate are included in the character string data Search processing and input of a word corresponding to an object not included or a word corresponding to a predicate when it is determined that a word corresponding to an object or a word corresponding to a predicate is not included in the character string data Word satisfaction request processing that requires
A program that causes a computer to execute. Query input device that accepts input of character string data representing a query,
A word satisfaction determination device for determining whether or not at least two words corresponding to an object and a predicate are included in an accepted query,
A search that searches for data using the word corresponding to the object and the word corresponding to the predicate as the search keyword when it is determined that the accepted query includes at least two words corresponding to the object and the predicate. When it is determined that the device and the accepted query do not contain the word corresponding to the object or the word corresponding to the predicate, the input of the word corresponding to the object or the predicate not included is input. Equipped with requesting word satisfaction request device,
Data retrieval system.   The computer-readable recording medium which recorded the program of Claim 4 or 14.

Images