JP2013069140A - Document analysis device and document analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate topic distribution with high accuracy even in an environment where new words may emerge.SOLUTION: Vocabulary set generation means 301 generates a first vocabulary set representing vocabulary of a first document set that is obtained in the past and a second vocabulary set representing vocabulary of a second document set that is obtained after the first document set. Prior distribution generation means 302 generates prior distribution of a frequency of appearance of each word included in the second vocabulary set on the basis of first word distribution which is polynomial distribution of a frequency of appearance of each word included in the first vocabulary set. Topic distribution generation means 303 generates second word distribution which is polynomial distribution of a frequency of appearance of each word included in the second vocabulary set on the basis of the prior distribution, and topic distribution which is polynomial distribution of a frequency of appearance of each topic characterized by the second word distribution. Prior state accumulation means 304 accumulates the second word distribution and the second vocabulary set.

Description

  The present invention relates to a technique for extracting a topic from a document.

Massive documents are generated every day on the Internet, such as text posted on microblogs such as Twitter (registered trademark), articles provided on news special sites, comments exchanged on SNS (social networking service), etc. .
Buzztter is known as one of the techniques for extracting matters that have become a hot topic on the Internet. Buzztter extracts words that appear frequently from text posted on Twitter in real time and distributes them to the user's terminal.

As one of topic models for extracting a topic, which is a potential factor that governs the appearance frequency of words on a document, from documents, there is Latent Dirichlet Allocation (hereinafter referred to as “LDA”, for example, Non-Patent Document 1). . FIG. 1 is a diagram showing a graphical model of LDA. Table 1 is a table showing the meaning of each symbol in FIG.

  In LDA, there are a plurality of topics in each document, each topic appears in each document with a certain probability, a plurality of words corresponding to each topic exist, and each word appears in each topic with a certain probability. Under the assumption, each topic is characterized by a multinomial distribution of the number of occurrences of each word (hereinafter referred to as “word distribution”), and a multinomial distribution of the number of occurrences of each topic (hereinafter referred to as “topic distribution”). This is a technique for characterizing each document. For example, if one sentence posted on Twitter (hereinafter referred to as “tweet”) is captured as a document, immediately after the earthquake occurs, many documents containing many topics related to earthquakes and tsunamis will be generated. The topic about is characterized by a multinomial distribution of the number of occurrences of each word (including word combinations) such as “earthquake”, “seismic intensity”, “earthquake breaking news”, and the topics about tsunami are “tsunami”, “height ”,“ Arrival time ”and the like, which are characterized by a multinomial distribution of the number of appearances of each word.

The conventional LDA is configured to extract topics from a specific document set given in advance. Therefore, topic changes cannot be extracted from a document set whose contents change over time. On the other hand, Dynamic Topic Model (hereinafter referred to as “DTM”, for example, Non-Patent Document 2) is a technique that extends LDA so that it can be applied to analysis of a document set whose contents change over time. is there. FIG. 2 is a diagram showing a graphical model of DTM. t is the time when the document set is acquired. A document set D ^(t) is input for each acquisition time t, and a word distribution β ^(t) and topic distribution θ ^(t) for each acquisition time t are generated. The feature of DTM is that the word distribution β ^(t−1) at the acquisition time ^t−1 is a prior distribution of the word distribution β ^{(t) at} the acquisition time t, in other words, the current topic from the past topic. It is a point to extract. Therefore, by comparing β ^(t) with β ^(t-1) , it is possible to observe changes in the topic over time. In addition, since the influence of past topics is reflected on the current topic, a result that does not overlearn the current document set can be obtained as compared with the case of extracting a topic from a document set only at a specific acquisition time. As a result, topics included in the document set at the next acquisition time t + 1 can be predicted with high accuracy.

D. M. Blei, A. N. Ng, and M. I. Jordan. Latent dirichlet allocation. In Journal of Machine Learning Research archive., 2003. D. M. Blei and J. D. Lafferty. Dynamic Topic Models. In Proc. Of the 23rd international conference on Machine learning., 2006.

FIG. 3 is a graph showing the results of analyzing 200,000 tweets per day posted on Twitter during the period from February 1 to March 31, 2011. The cumulative word count is a cumulative value of the number of words that appear in the tweet. The number of new words is the number of words (that is, new words) observed for the first time after the start of the period among the words appearing in the tweet. As shown in the figure, even after two months have passed since the start of observation, it can be seen that about 5000 new words continue to be generated per day. However, since DTM performs topic extraction within a fixed vocabulary set, even if a new word occurs at a certain time, topic extraction is not performed in consideration of it, and a new word is added to the word representing the topic. Will not be added. Therefore, as shown in FIG. 3, in an environment in which new words continue to be generated every day, the accuracy of topic extraction decreases with time.
The present invention has been made in view of the above-described background, and an object thereof is to generate a topic distribution with high accuracy even in an environment where a new word can be generated.

  The document analysis apparatus according to claim 1 is an acquisition unit that acquires a document set from an external device, a first vocabulary set that represents a vocabulary of the first document set acquired by the acquisition unit in the past, and the first A vocabulary set generating means for generating a second vocabulary set representing the vocabulary of the second document set based on the second document set acquired by the acquiring means after the document set of A prior distribution generation means for generating a prior distribution of the number of appearances of each word included in the second vocabulary set based on a first word distribution that is a multinomial distribution of the number of appearances of each word included in the vocabulary set; Based on the prior distribution, a second word distribution that is a multinomial distribution of the number of occurrences of each word included in the second vocabulary set is generated, and each topic characterized by the second word distribution is generated. For topics that are multinomial distributions of occurrences And having a Topic distribution generating means for generating, a pre-condition storing means for storing said second vocabulary set and the second word distribution.

  The document analysis apparatus according to claim 2 is the document analysis apparatus according to claim 1, wherein the vocabulary set generation unit is included in the first vocabulary set among words included in the second document set. A new word set that is a set of non-words and a known word set including part or all of the first vocabulary set, and the prior distribution generation means generates a prior distribution corresponding to the new word set On the other hand, a word distribution corresponding to the known word set in the first word distribution is used as a prior distribution corresponding to the known word set.

The document analysis apparatus according to claim 3 is the document analysis apparatus according to claim 2, wherein the prior distribution generation unit generates a prior distribution corresponding to the new word set based on an average value of the first word distribution. It is characterized by doing.
The document analysis device according to claim 4 is the document analysis device according to claim 2, wherein the vocabulary set generation means has a threshold of occurrence in any topic among words included in the first vocabulary set. The known word set is generated from the above words.

  The document analysis method according to claim 5 is based on a first vocabulary set representing a vocabulary of the first document set acquired in the past and a second document set acquired after the first document set. A vocabulary set generation step for generating a second vocabulary set representing the vocabulary of the second document set, and a first word distribution that is a multinomial distribution of the number of occurrences of each word included in the first vocabulary set A prior distribution generating step for generating a prior distribution of the number of occurrences of each word included in the second vocabulary set based on the first distribution, and a prior distribution of each word included in the second vocabulary set based on the prior distribution A topic distribution generation step of generating a second word distribution that is a multinomial distribution of the number of occurrences, and generating a topic distribution that is a multinomial distribution of the number of appearances of each topic characterized by the second word distribution; 2 word distribution and the second vocabulary And having a pre-condition storage step for storing and covering.

  According to a sixth aspect of the present invention, there is provided a program for acquiring a document set from an external device, a first vocabulary set representing a vocabulary of a first document set acquired by the acquisition unit in the past, and the first A vocabulary set generating means for generating a second vocabulary set representing a vocabulary of the second document set based on a second document set acquired by the acquiring means after the first document set; Pre-distribution generation means for generating a prior distribution of the number of appearances of each word included in the second vocabulary set based on a first word distribution that is a multinomial distribution of the number of appearances of each word included in the vocabulary set of Based on the prior distribution, each topic characterized by the second word distribution is generated by generating a second word distribution which is a multinomial distribution of the number of occurrences of each word included in the second vocabulary set. Multinomial distribution of the number of occurrences of And topic distribution generating means for generating a topic distribution that is characterized in that to function as a pre-condition storing means for storing said second vocabulary set and the second word distribution.

  According to the present invention, it is possible to generate a topic distribution with high accuracy even in an environment where a new word can be generated.

It is a figure which shows the graphical model of LDA. It is a figure which shows the graphical model of DTM. It is a graph which shows the result of having analyzed the tweet posted by Twitter. 1 is a diagram illustrating a configuration of a communication system 1. FIG. 2 is a block diagram showing a hardware configuration of a document analysis device 30. FIG. 4 is a block diagram illustrating functions of the document analysis device 30. FIG. It is a figure which shows the flow of the process which the document analysis apparatus 30 performs. It is a figure which shows the size of a vocabulary set, the number of new words, and the number of words deleted.

An embodiment of the present invention will be described.
(1) Configuration of Embodiment FIG. 4 is a diagram illustrating a configuration of the communication system 1 according to the embodiment of the present invention. The communication system 1 includes a mobile communication network 10, a mobile communication device 20, an Internet 50 connected to the mobile communication network 10 via a gateway device 60, a document analysis device 30 connected to the gateway device 60, and an Internet 50. And a plurality of web server devices 40 connected to each other.

  The mobile communication device 20 is a communicable computer such as a cellular phone, for example, and includes a control unit including an arithmetic device such as a CPU (Central Processing Unit) and a storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory). And a storage unit such as an EEPROM (Electronically Erasable and Programmable ROM) or SRAM (Static Random Access Memory) with a backup power source, a wireless communication unit including an antenna and a wireless communication circuit, a speaker, a microphone, and a sound processing circuit A voice input / output unit, an operation unit having a plurality of operators such as keys and a touch screen, and a display unit including a liquid crystal panel and a liquid crystal driving circuit are provided. In this mobile communication device 20, the control unit accesses the web server device 40 via the mobile communication network 10 and the Internet 50 by controlling the communication unit in accordance with a user operation received by the operation unit, and the web server. Information stored in the device 40 is acquired and displayed on the display unit. Thereby, the user can access and browse various information on the Internet 50.

  The mobile communication network 10 is a network that provides a communication service to the mobile communication device 20. The mobile communication network 10 includes a base station that performs wireless communication with a mobile communication device 20 that exists in a wireless cell of the mobile station, an exchange that performs routing of data transmitted in the network, and the mobile communication device 20. Are provided with various nodes such as a control station for performing location registration, and a communication line for interconnecting these nodes.

  The web server device 40 is a computer, and includes a control unit including a calculation device such as a CPU and a storage device such as a ROM and a RAM, a storage unit such as a hard disk device, and a communication unit connected to the Internet 50. . The web server device 40 has a function of performing data communication with the mobile communication device 20 via the Internet 50 and the mobile communication network 10. The web server device 40 is a server that provides a microblog service. When a user registers in the web server device 40 as a user of this service, the user is allowed to post to the microblog. When a user posts a text to the microblog using the mobile communication device 20, the web server device 40 sends the text back to the mobile communication device 20, whereby the text is displayed on the display unit of the mobile communication device 20. Is done. In addition, when the user registers another user, the text posted by the other user is also displayed on the display unit. In this way, it is possible to communicate among a plurality of users.

(2) Configuration of Document Analysis Device FIG. 5 is a block diagram showing a hardware configuration of the document analysis device 30. The document analysis apparatus 30 is a computer and includes a control unit 31, a communication unit 32, and a storage unit 33. The control unit 31 includes an arithmetic device such as a CPU and a storage device such as a ROM and a RAM. The CPU controls the operation of each unit of the document analysis apparatus 30 by executing a program group stored in the ROM or the storage unit 33 using the RAM as a work area.

The communication unit 32 includes a communication interface and is connected to the gateway device 60. The communication unit 32 acquires a document set distributed from the web server device 40 to the mobile communication device 20 relayed by the gateway device 60 from the gateway device 60 and inputs the document set to the document analysis device 30. That is, the communication unit 32 is an example of an acquisition unit that acquires a document set from an external device.
The storage unit 33 is a writable nonvolatile storage unit, for example, a hard disk device. The storage unit 33 stores a program group in which a procedure of processing executed by the control unit 31 is described. In addition, the storage unit 33 has a storage area for storing information acquired from the gateway device 60 and distributed to the mobile communication device 20 by the web server device 40.

FIG. 6 is a block diagram illustrating a functional configuration of the document analysis apparatus 30 according to the present embodiment. Table 2 shows symbols used in the description of the embodiments and their meanings.

The document analysis apparatus 30 includes a vocabulary set generation unit 301, a prior distribution generation unit 302, a topic distribution generation unit 303, and a prior state accumulation unit 304. There are mainly four processes executed by the document analysis device 30: vocabulary set generation, prior distribution generation, topic distribution generation, and prior state accumulation.
Here, terms used in the following description will be described.
In this embodiment, a sentence (text data) posted on a microblog is an object of analysis, and a sentence transmitted from the mobile communication device 20 to the web server device 40 in one posting is called a document. A document set is a set of one or more documents.
A word is a word constituting a document and corresponds to a morpheme in Japanese. The word can be any part of speech.

The acquisition time t is a time when the document analysis device 30 acquires the document set distributed from the web server device 40 to the mobile communication device 20 via the gateway device 60. The document analysis apparatus 30 executes a series of processes described below, triggered by acquisition of a document set. For example, the acquisition time t may be set at regular intervals such as every 24 hours or every 6 hours, or the administrator of the document analysis device 30 or the user of the mobile communication device 20 may set the intervals at any time without setting the intervals. The document analysis apparatus 30 may be instructed to acquire a document set.
The acquisition time t is represented by an integer. That is, if the time when the document set being processed by the document analysis apparatus 30 is acquired is t, the acquisition time of the document set that was the object of the previous processing is t-1.

The documents that make up the document set may be selected in any way. For example, 100,000 postings may be acquired retroactively from the latest posting date, or 100,000 postings may be randomly extracted from postings for the past 24 hours.
A new word is a word that is not included in a vocabulary set generated in the past. A known word is a word included in a vocabulary set generated in the past.
The prior distribution is a prior probability distribution in Bayesian estimation, and a posterior distribution is generated by multiplying the prior distribution by a likelihood function.

Next, the contents of each process will be described.
(2.1) Generation of Vocabulary Set Generation of a vocabulary set is executed by the vocabulary set generation unit 301. The vocabulary set corresponds to a combination of a new word set and a known word set. That is, the new word set and the known word set are both subsets of the vocabulary set, and generation of the vocabulary set includes generation of a new word set and generation of a known word set.
The vocabulary set generation unit 301 decomposes the document set D ^(t) acquired at the acquisition time t into morphemes, and executes the following processing using each morpheme as a word.
The vocabulary set generation unit 301 generates a new word set W _n ^(t) at the acquisition time t from the document set D ^(t) acquired at the acquisition time t ⁽ generation of a new word set). The new word set W _n ^(t) is a set of words that are not included in the vocabulary set W ^(t−1) at the acquisition time t−1 among the words included in the document set D ^(t) at the acquisition time t. The vocabulary set W ^(t-1) is a vocabulary set representing the vocabulary of the document set D ^(t-1) acquired up to the acquisition time t-1, and is stored in the prior state storage unit 304. The vocabulary set generation unit 301 reads the vocabulary set W ^(t-1) from the prior state storage unit 304, and among the words included in the document set D ^(t) , the words not included in the vocabulary set W ^(t-1). To generate a new word set W _n ^(t) .

Further, the vocabulary set generation means 301 generates a known word set W _o ^(t) at the acquisition time t from the vocabulary set W ^(t−1) (generation of a known word set). The known word set W _o ^(t) may include all the words included in the vocabulary set W ^(t−1) , but in that case, the size of the vocabulary set may increase as time passes. Therefore, in this embodiment, the known word set W _o ^(t) is generated by selecting a known word from the vocabulary included in the vocabulary set W ^(t−1) .
As an example of a known word selection method, in this embodiment, an obsolete word is deleted from the vocabulary set W ^(t−1) . Specifically, since each topic is characterized by a multinomial distribution of the number of occurrences of each word, a word with a smaller number of appearances contributes less to the topic generation. Words that are less likely to contribute to topic generation are words of low importance that have become obsolete in topic generation, and even if the words are deleted, it is considered that the accuracy of the topic model will not be reduced.

Specifically, the word selection is performed according to the following equation.

The above is the contents of the processing executed by the vocabulary set generation unit 301. A vocabulary set W ^(t) is a combination of the new word set W _n ^(t) generated in this way and the known word set W _o ^(t) .
In short, the vocabulary set generation unit 301 includes a first vocabulary set (W ^(t-1) ) representing the vocabulary of the first document set (D ^(t-1) ) acquired by the acquisition unit in the past, Based on the second document set (D ^(t) ) acquired by the acquisition means after the first document set, a second vocabulary set (W ^{(t (t)} ) representing the vocabulary of the second document set. ⁾ ) Is an example of a means for generating.
The vocabulary set generation means 301 includes a new word set (W _n ^(t) ) that is a set of words that are not included in the first vocabulary set among the words included in the second document set, and the first It can also be specified as a means for generating a known word set (W _o ^(t) ) including a part or all of one vocabulary set.
The vocabulary set generation unit 301 can also be specified as a unit that generates the known word set from words included in the first vocabulary set, the number of appearances of which is greater than or equal to a threshold in any topic.

(2.2) Generation of prior distribution

要するに、トピック分布生成手段３０３は、前記事前分布に基づいて、前記第２の語彙集合に含まれる各単語の出現回数の多項分布である第２の単語分布（β^(t)）を生成し、当該第２の単語分布で特徴付けられた各トピックの出現回数の多項分布であるトピック分布（θ^(t)）を生成する手段の一例である。 (2.3) Topic distribution generation

In short, the topic distribution generation unit 303 generates a second word distribution (β ^(t) ) that is a multinomial distribution of the number of appearances of each word included in the second vocabulary set based on the prior distribution. This is an example of means for generating a topic distribution (θ ^(t) ) that is a multinomial distribution of the number of appearances of each topic characterized by the second word distribution.

(2.4) Accumulation of Prior State Accumulation of the prior state is executed by the prior state accumulation unit 304. The prior state storage unit 304 stores the word distribution β ^(t) at the acquisition time t generated by the topic distribution generation unit 303 and the vocabulary set W ^(t) at the acquisition time t generated by the vocabulary set generation unit 301. To do. The accumulated word distribution β ^(t) and the vocabulary set W ^(t) are referred to by the vocabulary set generation unit 301, the prior distribution generation unit 302, and the topic distribution generation unit 303 when the word distribution is generated at the next acquisition time. Is done.
In short, the prior state accumulation unit 304 is an example of a unit that accumulates the second word distribution and the second vocabulary set.

(3) Operation of Embodiment The operation of the embodiment will be described (see FIGS. 5 and 7). FIG. 7 is a diagram showing a flow of processing executed by the document analysis apparatus 30.
When the document set D (t) is input at the acquisition time t, the vocabulary set generation unit 301 generates a new word set W _n ^(t) from the document set D ^(t) . The new word set W _n ^(t) is a set of words that are not included in the vocabulary set W ^(t−1) among the words included in the document set D ^(t) .
Further, the vocabulary set generation unit 301 generates a known word set W _o ^(t) from the vocabulary set W ^(t−1) . At this time, the vocabulary set generation unit 301 generates a known word set W _o ^(t) by selecting a known word from the vocabulary included in the vocabulary set W ^(t−1) .

FIG. 8 is deleted from the size of the vocabulary set W ^(t-1) , the number of new words, and the vocabulary set W ^(t-1) when the threshold κ is set to 1% of the number of occurrences of words in each topic. It is a figure which shows the number of words. The same vocabulary set W ^(t-1) as that of FIG. 3 was used. As can be seen from FIG. 8, the size of the vocabulary set W ^(t-1) converges to about 250,000 words, which is half or less of the cumulative number of words in FIG. Thus, by removing the obsolete word from the vocabulary set W ^(t-1), the size of inhibition vocabulary set W ^(t-1), preventing a decrease in processing speed by calculating the amount of increase in It becomes possible.
However, if too many words are deleted from the vocabulary set W ^(t-1) , not only obsolete words but also important words that contribute to topic generation may be deleted. In this case, the accuracy of the model decreases. there's a possibility that. Therefore, it is necessary to adjust how much the threshold value κ is set in actual operation according to the processing capability of the calculation system and the accuracy of the required topic model.

生成されたトピック分布は、様々な用途に利用され得る。例えば、トピック分布に基づいてトピックのランキングを作成し、ランキングの上位何番目かまでに位置するトピックを示す情報を移動通信装置２０のユーザに配信してもよい。また、同様の情報を予め契約した企業や団体に提供してもよい。

The generated topic distribution can be used for various applications. For example, a ranking of topics may be created based on the topic distribution, and information indicating topics located up to the top number in the ranking may be distributed to the user of the mobile communication device 20. Moreover, you may provide the same information to the company and organization which contracted beforehand.

According to the above embodiment, the vocabulary set is updated each time a document set is acquired. Therefore, when a new word is generated, the document is analyzed based on the vocabulary set to which the new word is added. Therefore, it is possible to generate a topic distribution with high accuracy even in an environment where new words can occur.
In addition, for known words, a topic distribution generated in the past is used as a prior distribution, so that the time required for processing can be reduced.
In addition, since a prior distribution is generated based on an average value of topic distributions generated in the past for a new word, a highly accurate prior distribution can be generated.
In addition, since the known word set is generated from words whose appearance count in the topic distribution generated in the past is greater than or equal to the threshold value, it is possible to prevent an obsolete word from being added to the known word set.

(4) Modifications The above embodiment may be modified as follows. Moreover, you may implement combining the following modifications.
(4.1) Modification 1
In the embodiment, an example of analyzing a document set posted on a microblog has been shown. However, other types of document sets may be analyzed.
For example, when the web server device 40 is a server that distributes news articles, articles distributed during a specific period may be acquired as a document set, and the same processing as in the embodiment may be performed on the document set. Good. Further, when the web server device 40 is a server that manages SNS, a comment exchanged on the SNS for a specific period is acquired as a document set, and the same processing as that of the embodiment is performed on this document set. Also good.

(4.2) Modification 2
In the embodiment, the distribution destination of the document set is not limited, but the document set distributed to a specific destination may be acquired and the same processing as in the embodiment may be performed. Specifically, it is only necessary to acquire a document whose destination address written in the header of the document is a specific address. The topic distribution generated from the document set acquired in this way may be transmitted to the specific address.

(4.3) Modification 3
In the embodiment, the example in which the document delivery destination is the mobile communication device has been described, but the document delivery destination may be any device. For example, a stationary computer connected to the Internet may be used.
In the embodiment, an example in which the control unit 31 of the document analysis apparatus 30 executes processing by executing a program has been described, but the same function may be implemented by hardware. The program is provided by being recorded on a computer-readable recording medium such as an optical recording medium or a semiconductor memory, and the program is read from the recording medium and stored in the storage unit 33 of the document analysis apparatus 30. Also good. Further, this program may be provided via a telecommunication line.

DESCRIPTION OF SYMBOLS 1 ... Communication system, 10 ... Mobile communication network, 20 ... Mobile communication apparatus, 30 ... Document analysis apparatus, 40 ... Web server apparatus, 50 ... Internet, 60 ... Gateway apparatus, 31 ... Control part, 32 ... Communication part, 33 ... Storage unit 301 ... vocabulary set generation means 302 ... prior distribution generation means 303 ... topic distribution generation means 304 304 prior state storage means

Claims (6)

An acquisition means for acquiring a document set from an external device;
Based on a first vocabulary set representing the vocabulary of the first document set acquired by the acquisition means in the past and a second document set acquired by the acquisition means after the first document set, Vocabulary set generation means for generating a second vocabulary set representing the vocabulary of the second document set;
A prior distribution for generating a prior distribution of the number of appearances of each word included in the second vocabulary set based on a first word distribution that is a multinomial distribution of the number of appearances of each word included in the first vocabulary set Generating means;
Based on the prior distribution, a second word distribution that is a multinomial distribution of the number of occurrences of each word included in the second vocabulary set is generated, and each topic characterized by the second word distribution is generated. Topic distribution generation means for generating a topic distribution which is a multinomial distribution of the number of occurrences;
A document analysis apparatus comprising: a prior state storage unit that stores the second word distribution and the second vocabulary set. The vocabulary set generation means includes a new word set that is a set of words that are not included in the first vocabulary set among words included in the second document set, and part or all of the first vocabulary set. And a known word set containing
The prior distribution generation unit generates a prior distribution corresponding to the new word set, and uses a word distribution corresponding to the known word set in the first word distribution as a prior distribution corresponding to the known word set. The document analysis apparatus according to claim 1, wherein: The document analysis apparatus according to claim 2, wherein the prior distribution generation unit generates a prior distribution corresponding to the new word set based on an average value of the first word distribution. The vocabulary set generation unit generates the known word set from words included in the first vocabulary set and having a frequency of appearance equal to or greater than a threshold in any topic. The document analysis device described. Based on the first vocabulary set representing the vocabulary of the first document set acquired in the past and the second document set acquired after the first document set, the vocabulary of the second document set A vocabulary set generation step for generating a second vocabulary set representing
A prior distribution for generating a prior distribution of the number of appearances of each word included in the second vocabulary set based on a first word distribution that is a multinomial distribution of the number of appearances of each word included in the first vocabulary set Generation step;
Based on the prior distribution, a second word distribution that is a multinomial distribution of the number of occurrences of each word included in the second vocabulary set is generated, and each topic characterized by the second word distribution is generated. A topic distribution generation step for generating a topic distribution which is a multinomial distribution of occurrences;
A document analysis method comprising: a prior state accumulation step of accumulating the second word distribution and the second vocabulary set. Computer
An acquisition means for acquiring a document set from an external device;
Based on a first vocabulary set representing the vocabulary of the first document set acquired by the acquisition means in the past and a second document set acquired by the acquisition means after the first document set, Vocabulary set generation means for generating a second vocabulary set representing the vocabulary of the second document set;
A prior distribution for generating a prior distribution of the number of appearances of each word included in the second vocabulary set based on a first word distribution that is a multinomial distribution of the number of appearances of each word included in the first vocabulary set Generating means;
Based on the prior distribution, a second word distribution that is a multinomial distribution of the number of occurrences of each word included in the second vocabulary set is generated, and each topic characterized by the second word distribution is generated. Topic distribution generation means for generating a topic distribution which is a multinomial distribution of the number of occurrences;
The program for functioning as prior state storage means for storing the second word distribution and the second vocabulary set.

Images