JP5300576B2 - SEARCH DEVICE, SEARCH METHOD, AND SEARCH PROGRAM - Google Patents

Description

  The present invention relates to a search device, a search method, and a search program for searching based on a search sentence input by text or voice.

  There are search systems and search engines that search information on a predetermined database or network based on a search condition input by a user. As a service for providing a search engine, for example, there is Non-Patent Document 1.

“Go search”, [online], [search April 1, 2009], Internet <URL: http://search.goo.ne.jp/web.jsp?MT=>

  When entering search conditions with text using an input device such as a keyboard, users who are unfamiliar with keyboard operations need time to input text, and input errors may occur. Is big.

  In addition, when a search condition is input by voice using an input device such as a microphone, erroneous recognition is likely to occur when the voice is recognized, and the search is performed by the erroneously recognized search condition. There is a problem that results are presented.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to reduce a user's input load and more easily input search conditions intended by the user, thereby further improving the convenience of the user. There is.

  In order to solve the above-described problem, the present invention provides a search device, a text input receiving unit that receives an input of a text search sentence, and a morphological analysis of the text search sentence, and a common semantic ID for each morpheme of the analysis result Morphological analysis means for generating text data with a common meaning ID, voice input acceptance means for accepting input of a voice search sentence, voice recognition of the voice search sentence and converting it into text, and the converted A speech recognition unit that assigns a common meaning ID to each morpheme of text and generates text data with a common meaning ID, and analyzes the text data with a common meaning ID generated by the morpheme analysis unit or the speech recognition unit, and a search condition Search condition analysis means for extracting a common meaning ID corresponding to the search condition, the search condition and the corresponding common meaning ID as a search condition. Has a search condition management means for storing in the storage means, and presenting means for presenting the search conditions stored in the search condition storage means, and searching using the search condition, a search means for outputting a search result.

  In addition, the present invention is a search method performed by a computer, wherein an input receiving step for receiving an input of a text search sentence or an input of a voice search sentence, and when the input of the text search sentence is received, the text search sentence Morphological analysis is performed, a common meaning ID is assigned to each morpheme of the analysis result, text data with a common meaning ID is generated, and when the voice search sentence is received, the voice search sentence is recognized and converted into text. A conversion step is performed, a common meaning ID is assigned to each morpheme of the converted text, and text data with common meaning ID is generated, and the text data with common meaning ID generated in the generation step is analyzed and searched. A search condition analysis step for extracting a common semantic ID corresponding to the search condition and the search condition; A search condition management step for storing a semantic ID in a search condition storage unit, a presentation step for presenting the search condition stored in the search condition storage unit, and a search step for performing a search using the search condition and outputting a search result And do.

  Further, the present invention is a search program executed by a computer, wherein the computer receives an input of a text search sentence or an input of a voice search sentence, and an input of the text search sentence is received The text search sentence is morphologically analyzed, a common meaning ID is assigned to each morpheme of the analysis result, text data with a common meaning ID is generated, and when the voice search sentence is accepted, the voice search sentence is Recognizing and converting to text, adding a common meaning ID to each morpheme of the converted text to generate text data with common meaning ID, and text data with common meaning ID generated in the generating step A search condition analysis step for extracting a search condition and a common semantic ID corresponding to the search condition; A search condition management step for storing the search condition and the corresponding common meaning ID in a search condition storage unit, a presentation step for presenting the search condition stored in the search condition storage unit, and a search using the search condition, And a search step for outputting a search result.

  According to the present invention, it is possible to reduce a user's input load, to more easily input a search condition intended by the user, and to further improve user convenience.

It is a block diagram which shows the structure of the search device with which one Embodiment of this invention was applied. It is a flowchart of the search process of this embodiment. It is a figure which shows an example of a common dictionary.

  Embodiments of the present invention will be described below.

  FIG. 1 is a configuration diagram showing a configuration of a search device according to an embodiment of the present invention. The search device 1 according to the present embodiment analyzes input information (search text) input by voice or text from a PC (Personal computer) 3 via a network such as the Internet, generates a search condition, and uses the search condition. The retrieved search result is transmitted to the PC 3.

  The illustrated search apparatus 1 includes a text input receiving unit 11, a morpheme analyzing unit 12, a voice input receiving unit 13, a voice recognizing unit 14, a search condition analyzing unit 15, a search condition managing unit 16, and a search condition storage. A unit 17, a search condition display unit 18, a search unit 19, a search database (search DB) 20, a search result output unit 21, and a common dictionary 22 are included.

  The text input accepting unit 11 accepts input of a search sentence input as text from the PC 3, and the voice input accepting unit 13 accepts input of a search sentence input as speech from the PC 3. The morpheme analysis unit 12 performs morphological analysis on the text search sentence received by the text input reception unit 11, assigns a common meaning ID to each morpheme of the analysis result by referring to the common dictionary 22, and text data with a common meaning ID Is generated.

  The voice recognition unit 14 recognizes a voice search sentence received by the voice input reception unit 13 and converts it into text, and assigns a common meaning ID to each morpheme of the converted text, and text data with a common meaning ID. Is generated.

  The search condition analysis unit 15 analyzes the text data with common meaning ID generated by the morpheme analysis unit 12 and the speech recognition unit 14, and extracts the search condition and the common meaning ID corresponding to the search condition. The search condition management unit 16 stores the search condition analyzed by the search condition analysis unit 15 and the corresponding common meaning ID in the search condition storage unit 17.

  The search condition display unit 18 transmits the search conditions stored in the search condition storage unit 17 to the PC 3 and displays it on the display of the PC 3 to present it to the user. The search unit 19 uses the search conditions stored in the search condition storage unit 17 to search the search DB 20 storing various information or a database existing on the network, and outputs the search results to the search result output unit 21. To do. The search result output unit 21 transmits the search result of the search unit 19 to the PC 3 and displays it on the display of the PC 3 to present it to the user. The common dictionary 22 stores a word (morpheme) to which a common meaning ID is assigned and a reading and a notation of the word in association with each other.

  The search device 1 described above can use a general-purpose computer system including a CPU, a memory, an external storage device, an input device, an output device, and a communication control device for connecting to a network. . In this computer system, each function of the search device 1 is realized by the CPU executing a predetermined program loaded on the memory.

  Next, the search process of this embodiment will be described.

  FIG. 2 is a flowchart showing search processing according to the present embodiment. First, the search device 1 accepts a search request from the PC 3 and starts accepting input of a voice or text search sentence (S11). The user inputs a search sentence by voice to a voice input device such as a microphone provided in the PC 3 or inputs a search sentence by text (characters) using a text input device such as a keyboard. As described above, in the present embodiment, a search sentence can be input by a desired method (speech input or text input) arbitrarily selected by the user.

  When a text search sentence is input, the text input reception unit 11 receives an input of the search sentence. When a voice search sentence is input, the voice input reception unit 13 receives an input of the search sentence (S12).

  In order to prevent simultaneous input of voice and text, when a text search sentence is input (S13: text), the voice input receiving unit 13 stops inputting the search sentence by voice (S14), and searches for the voice. When the sentence is input (S13: voice), the text input receiving unit 11 stops the input of the search sentence in the text (S15). In this way, exclusive control of voice input and text input is performed.

  When a voice search sentence is input (S13: voice), the voice recognition unit 14 analyzes the voice content and converts it into text by voice recognition technology, and each morpheme (word) of the converted text. A common meaning ID is assigned with reference to the common dictionary 22 to generate text data with a common meaning ID (S16). That is, the speech recognition unit 14 converts speech into text by speech recognition processing, and adds a common meaning ID to each morpheme recognized during the speech recognition processing. The morpheme will be described later.

  The speech recognition technology is a process of analyzing a spoken language spoken by a person by a computer and extracting the content being spoken as character data, and there are various technologies such as a hidden Markov model and an acoustic model. The voice recognition unit 14 may use a voice recognition engine such as “VoiceRex”, for example, to recognize the voice for each morpheme (word) and convert it into text.

  “VoiceRex” is described in “http://www.ntt.co.jp/RD/OFIS/active/2007pdf/pdf/g_ct07.pdf#search='VoiceRex '”.

  Specifically, for example, it is assumed that a search sentence with a voice “Ah, I want to know information on a fire that occurred in 2008” is input in S12. In this case, the voice recognition unit 14 divides into morphemes such as “2008 / In / Wake up / Take / Fire / No / Information / Know / Want”, and the common morpheme acquired from the common dictionary 22 is obtained. A semantic ID is added.

  FIG. 3 shows an example of the common dictionary 22. The common dictionary 22 is a dictionary used for both a voice search sentence and a text search sentence. The common dictionary has a common meaning ID, a meaning ID, a reading, and a notation for each morpheme having the same meaning. In FIG. 3, “2008” and “2008” are shown as words expressing the year as a specific example.

  Since there is no concept of “notation” at the time of voice input, it is sufficient that it can be recognized as “2008” regardless of whether “Nisenhachinen” or “Heisei Nijuunen” is input. Therefore, there is a portion (2 records) of the voice data 221 as a dictionary. do it. On the other hand, at the time of text input, similarly, the text data 222 (2 records) may be included as a dictionary.

  The common dictionary 22 of the present embodiment has three records as illustrated as a dictionary that includes and integrates the voice data 221 and the text data 222, and corrects or inputs text to a search sentence input by voice. The common meaning ID (Y2008) common to the voice and text is assigned so that the retrieved text can be corrected by voice. By using such a common dictionary, it is possible to cope with fluctuations in various input methods (a plurality of different inputs for the same meaning).

  The speech recognition unit 14 refers to the common dictionary 22, and if the recognized morpheme is “2008”, the speech recognition unit 14 acquires the common meaning ID (Y2008) of the word using the speech data 221, and also for other morphemes Similarly, the common meaning ID is acquired. Then, text data with a common meaning ID is generated by adding the common meaning ID of each morpheme to the text data divided into morphemes.

  Specific examples of text data with a common meaning ID include, for example, “2008 / In / Wake / Ta / Fire / No / Information / To know / want: Y2008 / IDXX / IDYY /... .

  On the other hand, when a text search sentence is input (S13: text), the morpheme analysis unit 12 performs morphological analysis of the text search sentence input in S12 (S17). Morphological Analysis is a morpheme (the smallest linguistic unit that has meaning in a language) using a grammatical rule or dictionary (word list with information such as parts of speech) in the target language. ). The morpheme analyzer 12 refers to the common dictionary 22 (text data 222), assigns a common meaning ID to each of the divided morphemes, and generates text data with a common meaning ID (S18).

  Then, the search condition analysis unit 15 analyzes the text data with the common meaning ID generated in S16 or S18 using syntax analysis and natural language processing technology, and searches the search condition included in the text data and the corresponding meaning ID. Is extracted (S19). Natural language processing is a technique for inputting a natural language into a computer and performing information processing according to a predetermined purpose.

  In the specific example shown in FIG. 2, the process is a process for searching for information related to disasters. Therefore, the search condition analysis unit 15 calculates search conditions related to time, search conditions related to disaster types, search conditions related to regions, etc. from text data with a common meaning ID. Extract.

  In this case, the search condition analysis unit 15 extracts the following search condition and a common meaning ID corresponding to the search condition.

<Search conditions> Time: 2008 [Y2008]
Disaster type: Fire [FIRE]
Note that the speech (reading) input in S12 is “2008 (Heisei Jiunen)”, but the notation of the common dictionary 22 is “2008”, so the search condition is “2008”.

  The search condition management unit 16 registers or updates the search condition and the common meaning ID analyzed by the search condition analysis unit 15 in the search condition storage unit 17 (S20). Then, the search condition display unit 18 transmits the search condition registered / updated in the search condition storage unit 17 to the PC 3 and displays it on the display device of the PC 3 (S21).

  The user using the PC 3 browses the search conditions displayed on the display device, and inputs an additional correction instruction to the PC 3 when the search conditions are not intended or when a new search condition is to be added. As a result, there is a search condition addition / correction instruction (S22: present), and the process returns to S11 to start accepting input of a voice or text search sentence.

  In this embodiment, by registering the common meaning ID together with the search condition in the search condition storage unit 17, after inputting the search sentence by voice first, not only by inputting an additional correction instruction by voice but also by adding correction by text It is possible to input an instruction, and after inputting a search sentence in text first, it is possible to input not only an additional correction instruction in text but also an additional correction instruction in speech (S12).

  Here, it is assumed that the user inputs an additional correction instruction in text using the PC 3 (for example, “I want to know information on 2007, not 2008”). In this case, the input of the search sentence by voice is stopped (S14), and the morpheme analysis unit 12 performs morpheme analysis (S17), refers to the common dictionary 22 and assigns a common meaning ID to each morpheme, and thus the common meaning. ID-added text data (for example, “2008 / not / not / 2007 /...: Y2008 / IDaa / IDbb /...)” Is generated (S18).

  Then, the search condition analysis unit 15 analyzes the text data with the common meaning ID of the additional correction instruction, and updates information for updating the search condition and the common meaning ID registered in the search condition storage unit 17 (for example, “time” (Change): 2008 [Y2008] → 2007 [Y2007] ”is generated (S19).

  The search condition management unit 16 updates the search condition and the common meaning ID registered in the search condition storage unit 17 based on the update information generated by the search condition analysis unit 15 (S20). Thereby, the following search condition and common meaning ID are stored in the search condition storage unit 17.

<Search conditions> Time: 2007 [Y2007]
Disaster type: Fire [FIRE]
Then, the search condition display unit 18 transmits the updated search condition to the PC 3 and displays it on the display device of the PC 3 (S21). When there is no additional search condition correction instruction (S22: none), the search unit 19 Then, the predetermined search DB 20 is searched with the search condition (S23), and the search result output unit 21 transmits the search result to the PC 3 (S24). The search DB 20 may be a database existing on the network as well as the case where the search device 1 is provided.

  In the present embodiment described above, the user can select an arbitrary input method (text input, voice input) and input a search sentence. This allows users who are unfamiliar with keyboard operation to input by voice, users who are accustomed to keyboard operation input by text, etc., so that the user can freely select the input method according to his / her characteristics and situation. It is possible to reduce the input load of the search sentence and improve user convenience.

  For example, in a place where the PC 3 is placed, if there is no space for the keyboard, it is conceivable that voice input is performed using a microphone, and text input is performed in a place where noise is high. In addition, it is conceivable that a search sentence is first input by voice, and after a search condition is presented, an additional correction instruction is input in text only for a part that has an unintended search condition due to misrecognition. Basically, it is conceivable that a search sentence is input by voice that is easy to input, and text is input when it is difficult to input by voice, such as a place name that is not understood.

  In this embodiment, before the search unit 19 performs a search, the search condition is transmitted to the PC 3 and presented to the user. Thereby, the user can correct and add the search condition as necessary after confirming the search condition. That is, since the search is performed under the search condition intended by the user, the convenience for the user can be improved, and more efficient search processing can be realized.

  In the present embodiment, the search condition presented to the user is not the search sentence itself input by the user, but the search condition analyzed by the search device. Therefore, the user determines whether the search condition is intended at a glance. Judgment can be made.

  In this embodiment, a common meaning ID is assigned to each morpheme of a search sentence that is input by text and each of the search sentences that are input by voice using a common dictionary. As a result, the same common meaning ID is assigned to the morphemes (words) having the same meaning regardless of whether the input is text input or voice input, and is stored in the search condition storage unit. Thereby, the search condition of the search sentence inputted by voice can be additionally corrected by text input, or the search condition of the search sentence inputted by text can be additionally corrected by voice input.

  In addition, this invention is not limited to said embodiment, Many deformation | transformation are possible within the range of the summary.

DESCRIPTION OF SYMBOLS 1: Search apparatus 11: Text input reception part 12: Morphological analysis part 13: Speech input reception part 14: Speech recognition part 15: Search condition analysis part 16: Search condition management part 17: Search condition storage part 18: Search condition display part 19: Search unit 20: Search result output unit 21: Search DB
22: Common dictionary 3: PC

Claims (9)

A search device,
Text input accepting means for accepting input of a text search sentence;
Morphological analysis of the text search sentence, assigning a common meaning ID to each morpheme of the analysis result, and generating text data with a common meaning ID;
Voice input receiving means for receiving input of a voice search sentence;
Speech recognition means for recognizing the speech search sentence and converting it to text, giving a common meaning ID to each morpheme of the converted text, and generating text data with a common meaning ID;
A search condition analysis unit that analyzes the text data with common meaning ID generated by the morpheme analysis unit or the speech recognition unit, and extracts a search condition and a common meaning ID corresponding to the search condition;
Search condition management means for storing the search condition and the corresponding common meaning ID in the search condition storage means;
Presenting means for presenting the search conditions stored in the search condition storage means;
And a search unit that searches using the search condition and outputs a search result. The search device according to claim 1,
The text input receiving means and the voice input receiving means receive a search condition correction instruction presented by the presenting means in text or voice,
The morpheme analyzing means performs a morphological analysis on the correction instruction inputted with the text, gives a common meaning ID to each morpheme of the analysis result, generates text data with a common meaning ID,
The voice recognition means recognizes the correction instruction inputted by voice and converts it into text, assigns a common meaning ID to each morpheme of the converted text, generates text data with a common meaning ID,
The search condition analysis unit analyzes the text data with the common meaning ID of the correction instruction generated by the morpheme analysis unit or the voice recognition unit, and extracts the search condition of the correction instruction and the common meaning ID corresponding to the search condition. ,
The search condition management means updates the search condition and the corresponding common meaning ID stored in the search condition storage means to a common meaning ID corresponding to the search condition of the correction instruction. The search device according to claim 1 or 2, wherein
For a morpheme having the same meaning, further has a common dictionary that associates reading, notation, and common meaning ID,
The morpheme analyzing unit and the speech recognition unit assign a common meaning ID to each morpheme using the common dictionary. A search method performed by a computer,
An input receiving step for receiving input of a text search sentence or input of a voice search sentence;
When the input of the text search sentence is accepted, the text search sentence is subjected to morphological analysis, a common meaning ID is assigned to each morpheme of the analysis result, and text data with a common meaning ID is generated. If accepted, the speech search sentence is recognized and converted into text, a common meaning ID is assigned to each morpheme of the converted text, and a generation step of generating text data with a common meaning ID;
A search condition analysis step of analyzing the text data with common meaning ID generated in the generation step and extracting a search condition and a common meaning ID corresponding to the search condition;
A search condition management step of storing the search condition and the corresponding common meaning ID in a search condition storage unit;
A presenting step of presenting search conditions stored in the search condition storage unit;
And a search step of performing a search using the search condition and outputting a search result. The search method according to claim 4,
The input receiving step receives a search condition correction instruction presented in the presenting step in text or voice,
In the generation step, when the correction instruction is input as text, the correction instruction is morphologically analyzed, a common meaning ID is given to each morpheme of the analysis result, and text data with a common meaning ID is generated. When a correction instruction is input by voice, the correction instruction input by voice is recognized and converted into text, a common meaning ID is assigned to each morpheme of the converted text, and text data with a common meaning ID is Generate
The search condition analysis step analyzes the text data with the common meaning ID of the correction instruction generated in the generation step, extracts the search condition of the correction instruction and the common meaning ID corresponding to the search condition,
The search condition management step updates the search condition and the corresponding common meaning ID stored in the search condition storage unit to a common meaning ID corresponding to the search condition of the correction instruction. The search method according to claim 4 or 5, wherein:
The computer has a common dictionary in which readings, notations, and common meaning IDs are associated with morphemes having the same meaning,
The generating step assigns a common meaning ID to each morpheme using the common dictionary. A search program executed by a computer,
In the computer,
An input receiving step for receiving input of a text search sentence or input of a voice search sentence;
When the input of the text search sentence is accepted, the text search sentence is subjected to morphological analysis, a common meaning ID is assigned to each morpheme of the analysis result, and text data with a common meaning ID is generated. If accepted, the speech search sentence is recognized and converted into text, a common meaning ID is assigned to each morpheme of the converted text, and a generation step of generating text data with a common meaning ID;
A search condition analysis step of analyzing the text data with common meaning ID generated in the generation step and extracting a search condition and a common meaning ID corresponding to the search condition;
A search condition management step of storing the search condition and the corresponding common meaning ID in a search condition storage unit;
A presenting step of presenting search conditions stored in the search condition storage unit;
A search step for performing a search using the search condition and outputting a search result. The search program according to claim 7,
The input receiving step receives a search condition correction instruction presented in the presenting step in text or voice,
In the generation step, when the correction instruction is input as text, the correction instruction is morphologically analyzed, a common meaning ID is given to each morpheme of the analysis result, and text data with a common meaning ID is generated. When a correction instruction is input by voice, the correction instruction input by voice is recognized and converted into text, a common meaning ID is assigned to each morpheme of the converted text, and text data with a common meaning ID is Generate
The search condition analysis step analyzes the text data with the common meaning ID of the correction instruction generated in the generation step, extracts the search condition of the correction instruction and the common meaning ID corresponding to the search condition,
The search condition management step updates the search condition and the corresponding common meaning ID stored in the search condition storage unit to a common meaning ID corresponding to the search condition of the correction instruction. A search program according to claim 7 or claim 8, wherein
The computer has a common dictionary in which readings, notations, and common meaning IDs are associated with morphemes having the same meaning,
The generating step assigns a common meaning ID to each morpheme using the common dictionary.

Images