JP2006079462A - Interactive information providing method for information retrieval and interactive information providing apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide information even if an input from a user includes a word (unknown word) which is not included in a database. <P>SOLUTION: An input analysis part 2 analyses an input, an interactive processing part 3 retrieves a topic field database 4 while interacting with the user, and when a word is an unknown word which is not included in the database, an unknown word sort determination part 6 decides whether the unknown word is the same sort as an entry of the topic field database 4 or not, an unknown word solution candidate selection part 7 selects a candidate word for solving the unknown word, and an unknown word relationship checking part 8 allows the user to select a candidate word suitable for solving the unknown word and provides a retrieval result obtained by using the candidate word. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to an unknown word processing method in information retrieval. In particular, the present invention relates to an interactive information providing apparatus that performs operations such as railway information guidance, schedule management, weather information guidance, and car navigation that perform unknown word processing when interactively communicating with people through a keyboard and voice.

2. Description of the Related Art Conventionally, an apparatus that achieves a predetermined operation through interactive interaction with a user (hereinafter referred to as an interactive information providing apparatus as appropriate) is known (see Patent Document 1 and Non-Patent Document 1). FIG. 5 shows a conventional interactive information providing apparatus. Such an interactive information providing apparatus provides information according to the user's request content obtained through interactive exchange performed under the function of the interactive control 401 as shown in FIG. In order to provide this information, the interactive information providing apparatus has an information providing database 402. For example, in the case of a railway information guidance system, the information providing database 402 is associated with information such as timetables of stations, time required for each railway between stations, and fares and express charges between stations. Become.
JP 2002-132811 (paragraphs 0005 to 0007, FIG. 1) Natural language processing, Vol.9, No.1, pp. 43-63 (2002)

  In a conventional interactive information providing apparatus (see FIG. 5), when an input from a user includes a word that is not included in the information providing database 402 (hereinafter referred to as an unknown word as appropriate) There is a problem that information cannot be provided even if there is relevant useful information. Furthermore, since the information providing database 402 is structured in a complicated manner, there has been a problem that a designer of the interactive information providing apparatus needs a lot of labor to add a new word. (Refer nonpatent literature 1).

  The present invention realizes an unknown word processing that can appropriately provide information in response to an input from a user including an unknown word, and an interactive information providing apparatus based on the unknown word processing.

  According to the present invention (claim 1), there is provided an interactive information providing method for providing information through interactive exchange in a natural language through a keyboard or voice with a user, wherein an unknown word solution supplement selection unit is provided by the user. An unknown word solution supplement is obtained for an unknown word included in the input, and the unknown word association confirmation unit associates the unknown word with information in the database through confirmation to the user, and the information providing unit is based on the associated information. Provide information.

  According to this method, it is possible to provide information for a search including an unknown word by associating the unknown word with information in the database.

  The present invention (Claim 2) provides interactive information provision in which a computer including input means, output means, and arithmetic means provides predetermined information to the user through interactive interaction with the user. In the method, the calculation means receives the user input via the input means, the input analysis section analyzes the user input and extracts a word, the unknown word type determination section A procedure for determining whether or not the extracted word is an unknown word by referring to a dictionary database, and an unknown word solution candidate selection unit that selects a candidate having a predetermined relationship with the unknown word by referring to the dictionary database In the unknown word association confirmation unit, the candidate is presented to the user via the output means, and the confirmation result of the user is input via the input means. In providing unit, the procedure for providing the extracted information via said output means to the user, to run.

  According to this method, information desired by the user can be provided even in a search including an unknown word.

  In the present invention (Claim 3), the unknown word type determination unit refers to an apparatus that stores a weight value indicating an appearance pattern of a word extracted in advance from a general document file and a list of related words. Determine the word.

  According to this method, information can be provided for a search including an unknown word more quickly.

  The present invention (Claim 4) provides interactive information provision for providing predetermined information to the user through interactive interaction with the user using a computer including input means, output means, and arithmetic means. The calculation means refers to an input receiving unit that receives the user's input via the input unit, an input analysis unit that analyzes the user's input and extracts a word, and a dictionary database An unknown word type determination unit that determines whether or not the extracted word is an unknown word, an unknown word solution candidate selection unit that selects a candidate having a predetermined relationship with the unknown word by referring to the dictionary database, and the output unit The unknown word association confirmation unit that presents the candidate to the user via the input and inputs the confirmation result of the user via the input means, the confirmation result with reference to the topic field provision database Comprising more interactive processing unit that extracts information related to the confirmed candidate, the information providing unit, for providing to the user the extracted information through an output unit.

  According to this apparatus, information desired by the user can be provided even in a search including an unknown word.

In the present invention (Claim 5), the unknown word type determination unit further includes an apparatus for storing a numerical value of a weight indicating an appearance pattern of a word extracted in advance from a general document file and a list of related words.

  According to this apparatus, information can be provided for a search including an unknown word more quickly.

  According to the present invention, it is possible to quickly provide useful information for a user even for a search including an unknown word that is not included in the database for providing information to be searched that could not be handled conventionally. .

  An embodiment of an interactive information providing method (unknown word processing method) and an interactive information providing apparatus in information search will be described with reference to the drawings. FIG. 1 shows a block diagram of the interactive information providing apparatus of this embodiment. As shown in FIG. 1, the interactive information providing apparatus 1 according to the present embodiment operates in conjunction with the input analysis unit 2, the dialogue processing unit 3 (3A, 3B), the topic field database 4, the information providing unit 5, and the information providing unit 5. In addition to the information providing database 51, an unknown word type determination unit 6, an unknown word solution candidate selection unit 7, a question answering system 71, and an unknown word association confirmation unit 8 are provided. The unknown word type determination unit 6 is an unknown word determination device that stores a numerical value of a weight indicating an appearance pattern of a word extracted in advance from a general document file such as newspaper text information and a multiple set of related words. A configuration including the information storage unit 61 may be employed. However, without adopting such a configuration, it is also possible to obtain on the spot a numerical value of the weight indicating the appearance pattern of the word extracted from the general document file by a method given by the user and a multiple set of related words .

[Input analysis section]
The input analysis unit 2 converts the user input into an attribute / key pair that can be processed by the dialogue apparatus. If the input format is a table in which attributes are shown to the user in advance, the input can be output as it is. When using a natural sentence from the keyboard as an input, a generally known parsing technique can be used. When using speech, generally known speech recognition technology can be used in addition to the above-described technology. Among the output values, those not described in the topic field database 4 to be described later are unknown words, and are therefore output with the unknown word.

[Dialogue processing section]
The dialog processing unit 3 sequentially monitors the input analysis result of the input analysis unit 2, and if sufficient information for providing information is not yet included in the key information input from the user, The analysis result is updated by performing interactive output for prompting input, and the updated analysis result is output. In addition, it has functions normally provided in a dialog processing unit of a general interactive information providing apparatus, such as history reference and resolution of ambiguity of homonyms.

[Topic database]
The topic field database 4 serves as a dictionary corresponding to each attribute name used for input analysis. That is, the information providing database 51 holds a list of words that can be accepted for providing information with the attribute. Words not included in this word list are unknown words.

[Unknown word type determination unit]
The unknown word type determination unit 6 determines whether the type of the unknown word being input is the same type as the entry in the topic field database 4 of the device. For example, if a station name is listed as a vocabulary that can be accepted as an attribute “departure location”, such as “departure location” or “destination” in “example of topic field database” in Table 3 to be described later It is determined whether the input unknown word is a word of the same kind as the station name or a different word from the station name.

  By making such a determination, for an unknown word that is essentially impossible to provide information depending on the association, the interactive information providing device informs the user that information cannot be provided promptly and terminates the dialogue. Realize that. For example, if an unknown word “Kyoto Station”, which is the same type of word as a station name, is input to a railway information guidance system that only has information on the Tokyo metropolitan area, various information related to Kyoto Station (such as timetables and trains) Unless there is a train or the like), the interactive information providing apparatus 1 cannot appropriately provide information. In such a case, it is desirable to end the dialog promptly rather than trying to associate. On the other hand, in the case of different words, there is a possibility that useful information can be provided if the word of the station name is successfully associated.

  Entries in a general database such as the information providing database 51 are not configured by a specific concept such as “station name” as in the above example. Therefore, an existing dictionary method can be used directly, but is not appropriate. In the present embodiment, each word listed in the topic field database serving as a dictionary for searching the target information providing database and the word to be determined are general sentences (for example, Judgment is made using whether the pattern of words appearing together in a newspaper or the like is a similar appearance method. The determination method will be described in the description of the operation of the interactive information providing apparatus 1 described later.

[Unknown word solution candidate selection section]
The unknown word solution selection complement selection unit 7 receives the result of the input analysis unit 2 as an input, and receives a generally known question answering system 71 (see, for example, Patent Document 1) for the unknown word in the analysis result. Used to output association candidates.

[Question answering system]
In this embodiment, a known question answering system is used as the question answering system 71. Since the example taken up in this embodiment is an example of railway information guidance, when a question sentence is given to a set of general documents including at least information on the station, the answer sentence to the question and the document on which it is based This includes systems that list examples. At this time, if there are a plurality of documents as the basis, a plurality of answers and the documents as the basis are listed. However, in the present embodiment, only the output of the answer is used, and the document information as the basis is not used.

[Unknown word association confirmation section]
The unknown word association confirmation unit 8 receives the unknown word to be associated and the candidate output by the unknown word solution candidate selection unit 7 and outputs the confirmed association candidate to the user.

[Information provision department]
The information providing unit 5 searches the information providing database 51 based on the analysis result obtained from the dialogue processing unit 3 (3B) to provide information. In the present embodiment, since the analysis result received by the information providing unit 5 does not include an unknown word, a conventional information providing unit that does not correspond to the unknown word and is similar to the interactive information providing apparatus can be used as it is. .

[Information providing database]
The information providing database 51 is a database that stores information to be provided to the user as the interactive information providing apparatus 1. In this embodiment, information such as the timetable of each station, the time required for each railroad between the stations, the fare and the express charge between the stations, and the like are associated with each other and referred to by the information providing unit 5.

[Operation of interactive information provider]
Here, the case where the interactive information providing apparatus 1 performs railway information guidance will be described as an example. In the railway information guidance, when the user inputs a desired departure place and a desired date place, the interactive information providing apparatus 1 guides a transfer route by train from the departure place to the destination.

  FIG. 1 shows the overall operation of the interactive information providing apparatus 1. The operation at each stage will be described separately for each part. As shown in FIG. 1, the interactive information providing apparatus 1 that has received an input from the user passes the input to the input analysis unit 2, where an attribute and key pair (table) that can be processed by the interactive apparatus. 1).

  At this time, the dialogue processing unit 3 communicates with the input analysis unit 2 to determine an attribute that is not given among information necessary for the search, and a message that prompts the user to input the uninput information. Output. In this process, when all the attribute values necessary for the search are determined, the topic field database 4 is searched. Here, with reference to the topic field database 4, unknown words that are not included in the topic field database 4 in the search are determined. At this time, if there is no unknown word, the contents of the search question are sent to the information providing unit 5, the information providing unit 5 outputs the search result for the information providing database 51, and the interactive processing is terminated. On the other hand, if the unknown word is found, the unknown word type determination unit 6 continues the unknown word processing.

  The unknown word type determination unit 6 determines whether the unknown word is the same type or different from the entry in the topic field database 4 by a method for determining the unknown word type described later. If the unknown word type is different, the processing of the unknown word solution candidate selection unit 7 To continue. If the determination results are the same type, the process is transferred to the information providing unit 5 to indicate to the user that there is no corresponding information, and the dialogue process is terminated.

  The unknown word solution candidate selection unit 7 receives an input including the unknown word analyzed by the input analysis unit 2, calls the question answering system 71, and selects a word that is a candidate for solving the unknown word. When there are a plurality of candidates, the plurality of candidates are output to the unknown word association confirmation unit 8, and when these candidates are output with priority according to the evaluation in the question answering system 71, the candidates according to the ranking are output. Is output. At this time, if no candidate is output, the process proceeds to the information providing unit 5 to indicate to the user that there is no corresponding information, and the dialogue process is terminated.

  And the unknown word correlation confirmation part 8 shows a candidate with a priority which the unknown word solution candidate selection part 7 output to a user with an unknown word in an order from a high priority. A normal search is performed on the topic field database 4 using the solution candidate word selected by the user, and the search result is output to the information providing unit 5. Using the same key used in this search, the information providing unit 5 searches the information providing database 51 and displays the result. Hereinafter, the operation in each part will be described in order.

表１は、鉄道情報案内を行う対話型情報提供装置における入力解析部の出力例である。目的地を属性とするキーは「東京タワー」になっており、話題データベースにおける目的地を属性とするエントリのなかに「東京タワー」がないために、この語が未知語と判断されている。 [Operation of input analysis unit]
The input analysis unit 2 in FIG. 1 analyzes input from the user. In the example of railway information guidance, input such as voice and natural language sentences given by the user is analyzed, and keys corresponding to the attributes of the departure place, destination, date and desired arrival time necessary for the search are obtained. decide.

Table 1 shows an output example of the input analysis unit in the interactive information providing apparatus that provides railway information guidance. The key attributed to the destination is “Tokyo Tower”, and since there is no “Tokyo Tower” among the entries attributed to the destination in the topic database, this word is determined to be an unknown word.

たとえば、入力解析部２による解析結果が表２のような場合、「目的地」に相当する値がないと情報提供を行えないので、対話処理部３Ａでは「目的地を指定してください」といった出力を利用者に対して行う。これによって、話題データベース４への検索が可能になるために必要な値（単語）が全て揃うように検索のキーを更新する。 [Operation of Dialogue Processing Unit]
The operation in the dialog processing unit 3A in FIG. 1 is an operation for performing a dialog in which the user inputs a key for an attribute necessary for performing a search on the topic database 4.

For example, if the analysis result by the input analysis unit 2 is as shown in Table 2, the information cannot be provided unless there is a value corresponding to the “destination”, so the dialog processing unit 3A may “specify the destination”. Output to the user. As a result, the search key is updated so that all values (words) necessary for searching the topic database 4 can be obtained.

表３は鉄道情報案内システムにおける話題分野データベース４の一例である。この例に対して、出発地を「京都駅」、目的地を「東京駅」、日にちを「４月１２日」、到着希望時刻を「１７時３０分」として与えると、話題分野データベース４の該当するエントリを検索し、該当するエントリがあることを出力する。しかし、該当するエントリがない場合には、どの属性のキーが未知語であったかを出力する。 [Topic database operation]
When the topic field database 4 is given all the keys for the attributes necessary for the search, the topic field database 4 responds to the search result.

Table 3 is an example of the topic field database 4 in the railway information guidance system. For this example, if the departure place is “Kyoto Station”, the destination is “Tokyo Station”, the date is “April 12”, and the desired arrival time is “17:30”, the topic field database 4 Search for the corresponding entry and output that there is a corresponding entry. However, if there is no corresponding entry, it outputs which attribute key was an unknown word.

[Operation of unknown word type determination unit]
The unknown word type determination unit 6 has a function of determining whether the given unknown word and the entry in the topic field database 4 are of the same type or different types using information on words appearing in the text of general literature. FIG. 2 is a diagram illustrating the operation of the unknown word determination unit 6. As shown in FIG. 2, when the word appears, a value A _u for determining the similarity of words appearing in the vicinity is calculated (S101), and a comparison is made as to whether A _u exceeds a threshold ( In the case of S102), Yes, the same type (S103) is output, and in the case of No, the output of different type (S104) is performed, and the operation of the unknown word determination unit 6 is terminated. This determination method will be described more specifically with reference to FIG.

  In the present embodiment, each word listed in the current topic database 4 and the word to be determined are determined through the following five steps.

[Judgment step 1]
For a large amount of general sentences (such as newspapers), the words collected in a specific width (5 words in this description, this width is called a window) The collection is enumerated (S201). Here, unlike a set, a collection of elements including duplicates is referred to as a multiple set, and a multiple set including words included in a specific width as described above is referred to as a window multiple set. A set made from the window multiple set is N (S202), and a multiple set made from the window multiple set is M. When this N and M are compared, since the occurrence of duplicate elements is not allowed in N, when the same window multiplex set overlaps, only one of the window multiplex sets appears.

  For example, the sentence “Today, there are many mistakes in receiving extra cash from customers at Hikaridai Station.” Is “(Today) (Kohdai Station) (in) (customer) (from) (extra) ( ) (Cash) () (received) (miss) (ga) (many) (occurrence) ”and if the window is 5, the following 10 window multiple sets are listed. The

b1 = {(from), (de), (today), (Hikaridai Station), (customer)}
b2 = {(from), (de), (extra), (Hikaridai Station), (customer)}
b3 = {(from), (de), (ni), (extra), (customer)}
b4 = {(from), (to), (extra), (customer), (cash)}
b5 = {(from), (to), (to), (extra), (cash)}
b6 = {(to), (to), (extra), (cash), (received)}
b7 = {(to), (to), (miss), (cash), (received)}
b8 = {(G), (O), (Miss), (Cash), (Received)}
b9 = {(), (), (miss), (received), (many)}
b10 = {(g), (miss), (received), (many), (occurrence)}

In this case, since the 10 window multiple sets do not overlap,
N = M = {b1, b2, b3, b4, b5, b6, b7, b8, b9, b10}
It becomes.

[Judgment step 2]
From each window multiplex set collected in the determination step 1, window multiplex sets including words included in the existing topic field database are listed. For example, if the only word contained in the topic field database 4 is “Hikaridai Station”, the window multiple sets selected in this case are b1 and b2 (S203).

[Judgment step 3]
For each word included in the window multiplex set collected in the decision step 2, a set B of multiplex sets including the word is created (S204), and the weight w _t for each word is calculated according to the equation (1). (S205).

ただし、ｆ_t,bは、単語ｔがウィンドウ多重集合ｂに出現した回数であり、ｎ_tは、単語ｔが含まれるようなウィンドウ多重集合が多重集合Ｍ中に出現した回数である。この式に具体的な値を代入すれば、それぞれの単語について、重みｗ_tを計算する。

たとえば、単語「光台駅」に対しては、数２の式のように、単語「今日」に対しては
数３の式のように計算する。

Here, f _{t, b} is the number of times that the word t appears in the window multiplex set b, and n _t is the number of times that the window multiplex set including the word t appears in the multiplex set M. If a specific value is substituted into this equation, the weight w _t is calculated for each word.

For example, for the word “Hikaridai Station”, the calculation is performed as shown in Expression 2 and for the word “Today” as shown in Expression 3.

[Determination Step 4]
From the set N, a window multiplex set that includes the word u that is the determination target of the unknown word type is selected (S206). A set S _u including all elements (words) of all window multiplex sets selected in this way and a multiplex set T _u of similar words are created (S207), and a determination value A _u is calculated according to the equation (4) (S208). ). However, when the word t does not appear in the window multiplex set including the word included in the existing topic field database of the device (that is, when the weight for t is not defined), the weight is calculated as 0.

ただし、ｆ_tは単語ｔが多重集合Ｔ_uに出現した回数であり、ｗ_tは判定ステップ３で算出した重みである。Ｔ_uはｕを含む全てのウィンドウ多重集合ｂの要素からなる多重集合であり、Ｓ_uはＴ_uから重複した要素を取り除いた集合である。

However, f _t is the number of times that the term t appeared to multiset T _u, the weight calculated in w _t the determination step 3. T _u is a multiple set including elements of all window multiple sets b including u, and S _u is a set obtained by removing duplicate elements from T _u .

For example, when the word “customer” is a target of unknown word type determination, a set of words S _u and a similar set of similar words T _u included in a window multiplex set including the word _u to be determined Once created, it will be as described below.

S _u = {(From), (De), (To), (Today), (Extra), (Hikaridai Station), (Customer), (Cash)}
_Tu = {(From), (From), (From), (From), (In), (In), (In), (To), (To), (Today), (Extra), (Extra ), (Extra), (Kodai Station), (Kodai Station), (customer), (customer), (customer), (customer), (cash)}

この例に示すように、「客」という単語に対してＡ_uを計算するためには、単語の集合Ｓ_uの要素を対象としたｆ_tとｗ_tを必要とする。 When the judgment value is calculated for the word “customer” using S _u and T _u , the result shown in the formula 5 is obtained.

As shown in this example, in order to calculate the A _u against the word "customer" requires targeting the elements of the set S _u word f _t and w _t.

[Determination Step 5]
It is compared with a certain threshold value of A _u for unknown word u at decision step 4 (S209). When the threshold value is exceeded, the determination target word is determined to be the same type as the word in the topic field database used in S203 (S210). Otherwise, it is determined that they are different (S211). For example, assuming that the threshold value is 10 and the determination value 8.35 is obtained as shown in Equation 5 for the word “customer” in determination step 4, it is determined that this is a different type, and an unknown word solution candidate is selected. Part 7 continues the dialogue process. In the case of the same type, control is transferred to the information providing unit 5 to notify the user that there is no information, and the dialogue processing is terminated.

[Operation of unknown word solution candidate selection section]
The unknown word solution candidate selection unit 7 makes an inquiry to a general question answering system (for example, Patent Document 1) using the unknown word determined by the unknown word type determination unit 6 as being different from the topic database entry. In the case of this embodiment, the question answering system 71 is a question answering system equipped with a railway information guidance function. For example, for an unknown word “CS Park” having an attribute of destination, if a question such as “Where is the nearest station of CS Park” is requested to this question answering system, generally the question answering system 71 Will output multiple response candidates. These multiple answer candidates are used as the output of the unknown word solution / supplement selection section. If the question answering system 71 outputs answer candidates with priorities, the unknown word solution candidate selection unit 7 outputs the answer candidates in order according to the answer candidates.

[Operation of Unknown Word Association Confirmation Unit]
FIG. 4 is a diagram for explaining the operation of the unknown word association confirmation unit 8. First, if there is an output, the highest-order solution candidate is extracted from the output of the unknown word solution candidate selection unit 7 (S301). Next, using the extracted solution candidate and the unknown word, a question for confirming whether the association is correct is made to the user (S302). If the response from the user is affirmative, the solution compensation and unknown word pair is terminated as an output of the unknown word association confirmation unit 8 (S303). If the user's response (S304) is not positive, the next solution candidate is taken out from the unknown word solution candidate selection unit 7. It is checked whether or not there is a next candidate (S305). If there is, the same procedure is repeated from the confirmation of the association with the solution candidate (S302). If the unknown word solution complement selection unit 7 confirms all the solution candidates, and there is no next solution candidate, the unknown word association confirmation unit 8 ends without outputting anything. (S306).

  For example, let us consider a case where the solution candidate to be associated with the unknown word “CS Park” is {“Aiko Ishida Station”, “Hakuen Station”}. In this case, first, “Aiko Ishida Station” which is the first solution compensation is taken out. Next, the following confirmation question is made to the user using the unknown word “CS Park” and the solution supplement “Aiko Ishida Station”.

  “This device is not familiar with CES Park, but should I tell you about Aiko Ishida Station?”

  If an affirmative reply such as “Yes” or “Yes” is obtained from the user, the pair of “CS Park” and “Aiko Ishida Station” is output as an associated solution candidate and the process ends. Otherwise, if the user does not get a positive response (for example, “I don't know” or “I don't know”), I will take out the next solution candidate “Hakuen Station” and do the same Ask a question. If no positive response is obtained for either “Aiko Ishida Station” or “Housen Station”, the process ends without outputting anything.

[Operation of Dialogue Processing Unit]
In the dialog processing unit 3B, after receiving the word used for the search from the input analysis unit 2 first, after some processes are completed, the unknown word is resolved again and then the topic field database 4 is searched. At this time, if a result that there is no unknown word is obtained, a key used for the search is sent to the information providing unit 5.

[Operation of information provider]
The information providing unit 5 receives an output from the dialogue processing unit 3A, the unknown word type determination unit 6 or the dialogue processing unit 3B. When the output from the dialogue processing unit 3A or the dialogue processing unit 3B is received, the search result is displayed on the information providing database 51 by the key used in the dialogue processing unit to the user, or the system voice is displayed. Speak at For example, in the case of the analysis result as shown in Table 4, the following utterance is made and the dialogue processing is terminated.

「京都駅から東京駅まで４月１２日１７時３０分に到着するためには、京都駅を１６時に出発する新幹線にご乗車ください。」

"To get from Kyoto Station to Tokyo Station on April 12 at 17:30, please take the Shinkansen departing from Kyoto Station at 16:00."

  When the output from the unknown word type determination unit 6 is received, it is a case where the unknown word has not been resolved successfully, so the user is informed that information cannot be provided, and the dialog processing is terminated.

  As described above, in the present embodiment, information can be provided based on information on how to use words in a general document even when the user's input includes an unknown word. . Furthermore, it is possible to detect that the action for associating an unknown word is useless and promptly inform the user that information cannot be provided. Thereby, the interactive information providing apparatus 1 can handle unknown words. Furthermore, since it is possible to handle unknown words without adding vocabulary by the designer of the information providing apparatus, it is possible to reduce the effort by the designer of the interactive information providing apparatus 1 for increasing the vocabulary. . In addition, it is possible to construct an information providing apparatus that is more useful to the user, and it is possible to improve the convenience of the information system that involves voice recognition and understanding of natural language sentences. Note that these unknown word processing and interactive information providing method are realized by computing means.

The figure explaining the flow of the unknown word processing method in an interactive information provision apparatus. The figure explaining the outline of the method of unknown word kind determination. The figure explaining the detail of the method of unknown word kind determination. The figure explaining operation | movement of an unknown word correlation confirmation part. The block diagram of the conventional interactive information provision apparatus.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Interactive information provision apparatus 2 Input analysis part 3 Dialog processing part 4 Topic field database 5 Information provision part 6 Unknown word kind determination part 7 Unknown word solution candidate selection part 8 Unknown word association confirmation part 51 Information provision database 61 For unknown word determination Information storage unit 71 Question answering system

Claims (5)

An interactive information providing method for providing information through interactive exchange in natural language through a keyboard or voice with a user,
The unknown word solution compensation selection part asks for unknown word solution compensation for unknown words included in the input from the user,
The unknown word association confirmation unit associates the unknown word with the information in the database through confirmation to the user,
Information guidance based on information associated with the information provider,
An interactive information provision method characterized by A computer comprising an input means, an output means, and an arithmetic means provides an interactive information providing method for providing predetermined information to the user through interactive exchange with the user,
The computing means is
A procedure for accepting the user's input via the input means;
In the input analysis unit, a procedure for extracting a word by analyzing the user input,
In the unknown word type determination unit, a procedure for determining whether or not the extracted word is an unknown word with reference to a dictionary database,
In the unknown word solution candidate selection unit, a procedure for selecting a candidate having a predetermined relationship with the unknown word with reference to a dictionary database;
In an unknown word association confirmation unit, the candidate is presented to the user via the output means, and the confirmation result of the user is input via the input means,
In the information providing unit, a procedure for providing the extracted information to the user via the output means;
Performing the
An interactive information provision method characterized by The unknown word type determination unit determines an unknown word with reference to a device that stores a numerical value of a weight indicating an appearance pattern of a word extracted in advance from a general document file and a list of related words;
The interactive information providing method according to claim 2. An interactive information providing apparatus that provides predetermined information to the user through interactive exchange with a user using a computer including an input unit, an output unit, and a calculation unit,
The computing means is
An input receiving unit that receives the user's input via the input means;
An input analysis unit that analyzes the user's input and extracts words;
An unknown word type determination unit that determines whether the extracted word is an unknown word with reference to a dictionary database;
An unknown word solution candidate selection unit that selects candidates having a predetermined relationship with the unknown word with reference to the dictionary database;
An unknown word association confirmation unit that presents the candidate to the user via the output means and inputs the confirmation result of the user via the input means,
A dialogue processing unit that extracts information related to the candidate confirmed by the confirmation result with reference to the topic field providing database;
An information providing unit for providing the extracted information to the user via an output unit;
Having
An interactive information providing device characterized by the above. The unknown word type determination unit further includes a device for storing a numerical value of a weight indicating an appearance pattern of a word extracted in advance from a general document file and a list of related words,
The interactive information providing apparatus according to claim 4.

Images