JP5528787B2 - Stock price impact company detection system - Google Patents

Description

  The present invention relates to a stock price-affected company detection system and program, and more particularly to a technology for identifying a company that has a positive or negative impact on the stock price due to the occurrence of a specific event such as the occurrence of a natural disaster or fluctuations in exchange rates.

Stock prices are an important indicator of corporate value, and of course fluctuate depending on the company's own business performance and future prospects, but the occurrence of earthquakes, fluctuations in exchange rates, outbreak of war, bankruptcy of business partners It can be greatly affected by external factors.
For example, if a coup occurs in an African country, the stock price of a manufacturer that owns a major factory in that country can easily be expected to crash.
For this reason, investors regularly work to collect information on the global situation through newspapers, specialized magazines, the Internet, television, etc., and once something happens, immediately identify the companies affected by the event. If the impact is positive, it is required to buy the stock of the company and sell it if it is negative.
What moves the stock price / War, terrorism, political instability (geopolitical risk) Internet URL: http://kabu-shosinsha-kabuneko.net/14/16/000136.html Search date: November 24, 2009

  However, in order to do this, it is necessary to have a thorough understanding of each company's activities, areas of activity, related companies, etc., and to know in advance the specific company names that will have a positive or negative impact on various events. In addition, since it is necessary to constantly check news information and the like, there are currently only a few investors who can move to investment behavior immediately after the occurrence of a sudden event.

  The present invention has been devised to overcome such a current situation, and automatically detects the presence or absence of an event that affects a company's stock price, and is specific to being affected by the stock price by this event. The purpose is to provide technology that can present company names.

In order to achieve the above object, the stock price-affected company detection system according to claim 1 registers event information storage means for storing event information having specific attributes that affect stock prices, and the attribute information of each company. An inference rule that defines whether the impact on the stock price is positive or negative for each combination pattern of the company information storage means, the event attribute, and the company attribute that is affected by the stock price due to the event A rule storage means, a means for applying the inference rule to each event information stored in the event information storage means, and identifying an attribute of a company that affects the stock price due to the event; and the company information storage means Refer to and generate a list of stock price-affected companies and a stock price-affected company extraction means for extracting companies with the above attributes as stock price-affected companies, And the attribute of the event information is a combination of a character string indicating the type of the event and a character string indicating the region where the event occurs, and the specific information related to the company is used as the attribute information of the company. The above inference rules define whether the effect on the stock price is positive or negative for each combination pattern with the event type and the specific type of region related to the company. The company extraction means is a stock price-affected company detection system that extracts, as a stock price-affected company, a company in which an event information generation area is registered as a specific type of area related to the company defined in the inference rule, An event dictionary that defines the correspondence between abstract tags indicating the types of higher-level conceptual events and character strings that represent specific event names, An abstract tag that indicates the correspondence between an abstract tag that indicates a conceptual region and a character string that represents a specific region name, an abstract tag that indicates the occurrence of a higher-level conceptual event, and a specific event An event predicate dictionary that defines a correspondence relationship with a character string that represents the occurrence fact, a means for decomposing a sentence in text data into morpheme units and identifying a part of speech of each morpheme, and referring to each of the above dictionaries, For each morpheme, a character string that represents the name of the event, a character string that represents the region name, and a character string that represents the fact that the event has occurred. In the sentence, a case slot is provided with a means for associating an abstract tag that indicates an event, an abstract tag that indicates the occurrence of an event, and a storage column for corresponding words at least for each particle attached to the subject and for each particle attached to the region. A case slot filling means for associating a morpheme with an abstract tag indicating the occurrence of an event as a predicate with the case slot, an abstract tag of the subject to be extracted, and an attachment to the subject An extraction frame definition in which at least a condition for specifying a particle to be extracted, a condition for specifying an abstract tag for a predicate to be extracted, an abstract tag for an area to be extracted, and a condition for specifying a particle attached to the area are defined. , By extracting a plurality of extracted frame definition storage means and applying the extracted frame definition to the case slots filled with corresponding words, extract at least information elements corresponding to the subject, predicate, and region of the sentence, And an information extraction means stored in the event information storage means, wherein the case slot filling means executes the following processing.
(1) The case slot in which the filling of the corresponding word with respect to the preceding predicate is completed is inherited for the subsequent predicate.
(2) Overwrite the corresponding word for the following predicate in the corresponding word storage field of the case slot.
(3) When the corresponding term storage column of the particle representing the subject of the sentence is overwritten with a word representing the subject related to the subsequent predicate, the corresponding word filled with respect to the preceding predicate is deleted.

The stock price-affected company detection system according to claim 2 is the system according to claim 1 , and further includes identification information indicating that the sentence is a title sentence in advance for the sentence corresponding to the title in the above sentence. A word attached to the subject regardless of the presence or absence of a particle for a word to which an abstract tag indicating the type of event in the title sentence to which the identification information is given is provided. The word having the abstract tag indicating the region is filled in the corresponding word storage column of the particle attached to the region name regardless of the presence or absence of the particle.

The stock price-affected company detection system according to claim 3 includes an event information storage means for storing event information having a specific attribute that affects the stock price, and a company information storage means for registering attribute information of each company. An inference rule storage means for storing an inference rule that defines whether the influence on the stock price is positive or negative for each combination pattern of the attribute of the event and the attribute of the company affected by the stock price by the event, and the event information The inference rule is applied to each event information stored in the storage means, and the attribute of the company that affects the stock price due to the event is identified, and the company information storage means is referred to, and the attribute is provided. Stock price-affected company extraction means for extracting a company as a stock price-affected company, and means for generating and outputting the list of stock price-affected companies. The attribute of the event information consists of a combination of a character string that identifies the name of the company that is the subject of the event and a character string that represents the type of event related to the company situation. The above inference rules define whether the impact on the stock price is positive or negative for each combination pattern of event type and company type related to the corporate situation. The stock price-affected company extraction means extracts a stock price-affected company that extracts a company in which the company name that is the subject of the event information relating to the corporate situation is registered as a specific type of related company defined in the above inference rules. a system, further, the abstract tag indicating higher conceptual company name, a company name dictionary that defines the correspondence between the string representing a specific corporate name, An event predicate dictionary that defines the correspondence between an abstract tag that indicates the occurrence of a high-level conceptual event related to the business situation and a character string that indicates the specific event occurrence related to the business situation, and text data Means for decomposing sentences into morpheme units and identifying parts of speech of each morpheme, referring to each dictionary above, a character string representing a specific company name in each morpheme, and generating an event related to the company situation A means for associating an abstract tag indicating a company name, an abstract tag indicating the occurrence of an event relating to the corporate situation, and at least for each particle attached to the subject, corresponding to a character string representing a fact A case slot with a corresponding word storage field is filled with the corresponding word in the sentence, and a morpheme with an abstract tag indicating the occurrence of the event is added to the case slot as a predicate. An extraction frame in which case slot filling means for associating the above descriptive word, a condition for specifying the abstract tag of the subject to be extracted and a particle attached to the subject, and a condition for specifying the abstract tag of the predicate to be extracted are defined. Extracting frame definition storage means for storing a plurality of definitions, and applying the extracted frame definition to the case slots filled with corresponding words, extracts at least information elements corresponding to the subject and predicate of the sentence, A stock price-affected company detection system comprising: information extraction means for storing in event information storage means, wherein the case slot filling means executes the following processing.
(1) The case slot in which the filling of the corresponding word with respect to the preceding predicate is completed is inherited for the subsequent predicate.
(2) Overwrite the corresponding word for the following predicate in the corresponding word storage field of the case slot.
(3) When the corresponding term storage column of the particle representing the subject of the sentence is overwritten with a word representing the subject related to the subsequent predicate, the corresponding word filled with respect to the preceding predicate is deleted.

The stock price-affected company detection system according to claim 4 is the system according to claim 3 , and further includes identification information indicating that the sentence is a title sentence in advance for the sentence corresponding to the title in the above sentence. The case slot filling means is attached to the subject, regardless of the presence or absence of the particle, with respect to the word with the abstract tag indicating that it is the company name in the title sentence to which the identification information is assigned. It is characterized in that it fills the corresponding word storage column of the particle to be processed.

The stock price-affected company detection system according to claim 5 is the system according to claims 1 to 4 , and further determines a part of speech of the compound word for each connection pattern of a plurality of parts of speech to be a compound word. Compound word analysis rule storage means for storing a compound word analysis rule in which a standard is defined, and the compound word analysis rule with reference to the compound word analysis rule, and a morpheme corresponding to the part-of-speech concatenation pattern specified in the compound word analysis rule in the sentence If there is a combination, it comprises a compound word analyzing means for recognizing these morphemes as compound words, and the case slot filling means is a compound word unit for a combination of morphemes recognized as compound words. In this case, the case slot filling process is executed.

The stock price-affected company detection system according to claim 6 is the system according to claims 1 to 5 , further comprising an abstract rule storage means for storing an abstract rule for estimating a morpheme type, and And a means for associating an abstract tag indicating the type with a morpheme that matches the abstract rule.

Stock influence companies detection system according to claim 7, The system of claim 6, further each anaphor, stores anaphora resolution rule that defines the criteria for determining the antecedent An anaphoric analysis rule storage means to be provided, an anaphoric analysis means for referring to the anaphoric analysis rule, determining a corresponding antecedent for the anaphoric in the sentence, and replacing the anaphor with the antecedent It is characterized by.

According to the stock price-affected company detection system described in claim 1, since specific companies that have a positive or negative impact on the stock price due to the occurrence of each event information are automatically listed and provided to the user, Even users who do not have time to check news information or who are not familiar with the specific activities and regions of each company and are not aware of the impact of each event on stock prices are required immediately. It becomes possible to shift to investment behavior (buying and selling stocks of the company).
Here, in the case where the geographical range to be affected is a local event, such as the occurrence of a natural disaster or a conflict, the relationship between the event occurrence area and each company is a point to estimate the influence on the stock price.
For this reason, in the stock price-affected company detection system according to claim 1, a character string indicating the type of event and a character string indicating the occurrence region are set as attributes of the event information, and the specific type of area information related to the company is It is set as a company attribute. In addition, for each combination pattern with the event type and the specific type of regional information related to the company, the influence (plus or minus) of the event on the stock price is set.
As a result, when an event such as a natural disaster that has a local impact occurs, it is possible to list companies related to the region where the event occurred as a stock price affected company.
There are various companies that provide information about events that occur regularly, such as changes in crude oil transaction prices and exchange rates, and it is easy to obtain relatively well-formed event information. For events that occur suddenly, such as disasters and conflicts, the first report is often provided in text format data (news articles on the Web, etc.) written in natural language, so It cannot be used as event information.
On the other hand, the stock price-affected company detection system according to claim 1 prepares a typical case slot in which a necessary particle type is set in advance and an extraction frame definition that defines the type of information to be extracted. If there is a corresponding particle in the sentence, the previous independent word is filled in the corresponding word storage field of the particle, and the extracted frame definition is applied to this filled case slot, so that the word order is changed. Because it has a function to automatically extract structured information such as “subject (events of natural disasters and conflicts)-predicate-region” from natural sentences, event information on natural disasters and conflicts, etc. It is possible to list stock price-affected companies based on this.
Moreover, in principle, the case slot related to the previous predicate is inherited by the next predicate, so even if there is a subject or region abbreviation for the subsequent predicate, it is easy for the subject and region of the previous predicate. Can make up for.

Trends in related companies (changes in circumstances) such as bankruptcy of business partners, increase in profits, decrease in profits, and public offerings also have a major impact on the company's stock price.
For this reason, in the stock price-affected company detection system according to claim 3, a character string that identifies the company name and a character string that represents the type of the event are set as attributes of the event information, and the specific type related to the company Related company information (customer, subsidiary, parent company, partner, etc.) is set as an attribute of the company. In addition, for each combination pattern of the event type and a specific type of related company related to the company, the influence (plus or minus) of the event on the stock price is set.
As a result, when an event such as bankruptcy, increase in profit, decrease in profit, disclosure, etc. occurs for a specific company, companies that have a predetermined relationship with the company that is the subject of the event may be listed as a stock price affected company. It becomes possible.
Information on corporate trends, like natural disasters and conflicts, is often provided as text information (news articles on the Web, etc.) in the form of text written in natural language. Not available.
On the other hand, the stock price-affected company detection system according to claim 3 prepares a typical case slot in which a necessary particle type is set in advance and an extraction frame definition that prescribes the type of information to be extracted. If there is a corresponding particle in the sentence, the previous independent word is filled in the corresponding word storage column of the particle, and the extracted frame definition is applied to this filled case slot, regardless of the word order. Since it has a function to automatically extract event information related to the company situation structured as “subject (company name)-predicate (company situation)” from the natural sentence, the stock price impact based on this event information It becomes possible to list companies.
Moreover, in principle, the case slot related to the previous predicate is inherited by the next predicate, so even if the subject (company name) is omitted for the subsequent predicate, it is easy to use the subject of the previous predicate. Can make up for.

According to the stock price-affected company detection system described in claims 2 and 4 , it is possible to extract the necessary words by applying the above case slots to title sentences in which particles are often omitted. It becomes.

According to the stock price-affected company detection system according to the fifth aspect , it is possible to accurately identify a compound word composed of a combination of a plurality of morphemes in a sentence, and as a result, it is possible to extract information elements with an optimum granularity.

According to the stock price-affected company detection system described in claim 6 , it is possible to add an abstract tag based on a rule base even for morphemes that are not recorded in the dictionary, and it is possible to extract many information elements from the text accordingly. It becomes.

According to the stock price-affected company detection system according to the seventh aspect , it is possible to replace the anaphor in the sentence with the corresponding antecedent, so that many information elements can be extracted from the text.

FIG. 1 is a block diagram showing the overall configuration of a stock price-affected company detection system 10 according to the present invention. A morphological analysis processing unit 12, a compound word analysis processing unit 14, an abstraction processing unit 16, and an anaphoric analysis processing unit. 18, sentence type determination processing unit 20, case slot filling processing unit 22, event extraction processing unit 24, stock price affected company extraction unit 26, dictionary storage unit 28, compound word analysis rule storage unit 30, abstract A rule rule storage unit 32, an anaphora analysis rule storage unit 34, an extracted frame definition storage unit 36, an event information storage unit 38, an inference rule storage unit 40, and an ontology storage unit 42.
A Web server 44 is connected to the stock price affected company extraction unit 26 via a network. The Web server 44 provides an investment information providing service to a plurality of client terminals 47 connected via a communication network 46 such as the Internet.

The above morphological analysis processing unit 12, compound word analysis processing unit 14, abstraction processing unit 16, anaphora analysis processing unit 18, sentence type determination processing unit 20, case slot filling processing unit 22, event extraction processing unit 24, stock price affected company The extraction unit 26 is realized by the CPU of the computer executing necessary processing according to the OS and application programs.
Further, the dictionary storage unit 28, compound word analysis rule storage unit 30, abstraction rule storage unit 32, anaphora analysis rule storage unit 34, extracted frame definition storage unit 36, event information storage unit 38, inference rule storage unit 40, The ontology storage unit 42 is provided in the hard disk of the computer.
In the dictionary storage unit 28, an event dictionary, an event predicate dictionary, a regional dictionary, a company name dictionary, an activity object dictionary, a synonym dictionary, and the like are stored.

    FIG. 2 exemplifies the registered contents of the event dictionary. For “natural disaster” which is a high-level conceptual character string (abstracted character string) indicating the type of event, “earthquake”, “flood” Specific event names (expression character strings) such as “drought” and “tsunami” are associated in advance. In addition, a character string such as “terrorism”, “civil war”, “armed collision”, “coup” is associated with “conflict”, which is a high-level conceptual character string (abstracted character string), in advance. Yes.

  FIG. 3 exemplifies the registered contents of the event predicate dictionary. For “disaster occurrence”, which is a high-level conceptual character string (abstracted character string) indicating the occurrence fact of an event, “occurrence”, “ Specific predicates (expression character strings) such as “observation”, “occurs”, and “trick” are associated in advance. Moreover, specific predicates (expression strings) such as “occurrence”, “breakout”, “occurs”, “occurs”, etc., for “conflict occurrence” which is a high-level conceptual character string (abstracted character string) Are associated in advance. Similarly, specific predicates such as “bankruptcy”, “bankruptcy”, “increased profit”, “decrease profit”, “disclosure”, etc. for “corporate situation” which is a high-level conceptual character string (abstracted character string) Expression character string) is associated in advance.

  FIG. 4 exemplifies the registered contents of the regional dictionary. “Region”, which is a character string (abstracted character string) indicating a higher conceptual region, is compared with “North America”, “US”, “Canada”. ..., “Tokyo”, “Chuo Ward”, “Minato Ward”, “Kanagawa Prefecture”, etc., are associated with specific area names (expression character strings) in advance.

Although illustration is omitted, in the company name dictionary, a number of specific company names (formal names and abbreviations) to be the subject are registered in association with the abstract character string “company name”.
Similarly, although not shown in the figure, the activity object dictionary includes “liquid crystal”, “required” for abstract character strings such as “production object”, “sales object”, and “development object”. Specific expression character strings that should be objects of corporate activities, such as “liquid crystal television”, “liquid crystal panel”, and “liquid crystal monitor”, are associated in advance.

  In the inference rule storage unit 40, inference rule data defining the influence (±) on the stock price is stored for each combination pattern of the attribute item of each event and the attribute item of the affected company. 5 and 6 show an example of inference rule data stored in the inference rule storage unit 40, and each record has “event class”, “subclass”, and “change” as event attribute items. The data items of “Headquarters location”, “Sales region”, “Production region”, “Sales product”, “Raw materials”, “Affiliated company” are set as attribute items of affected companies. The data item “influence on stock price (±)” is set for each.

In the “event class”, “currency”, “material price”, “natural disaster”, “conflict”, “business situation”, etc., which are high-level conceptual character strings indicating the types of events, are described.
The “subclass” is an item in which a subordinate character string indicating the type of event is stored, and “Euro / yen”, “dollar / yen”, etc. are set as a subclass of “currency”. . In addition, “Petroleum Price” and “Steel Price” are subclasses of “Material Price”, “Earthquake” and “Eruption” are subclasses of “Natural Disaster”, and “Terrorism” and “Civil War” are subclasses of conflict. However, “failure”, “income”, “decrease”, “public”, etc. are set as subclasses of “corporate situation”.

The “change” is an item for setting a change in the time series in the positive direction or the negative direction for the event set in the subclass. For example, for “Euro / yen”, “change from the previous day” Records set with “+” and records set with “comparison with previous day-” are prepared.
In addition, for “oil price” and “steel material price”, records each set with “compared to the previous day +” are registered. Records in which “comparison with the previous day −” is set for “oil price” and “steel price” are also registered in the inference rule storage unit 40.
On the other hand, events such as “earthquake”, “eruption”, “terrorism”, “civil war”, “failure”, “increased profit”, “decreased profit”, and “open” are, in their nature, “positive or negative” The item “change” is blank because it is not familiar with the concept of “time-series change”.

The above-mentioned event attribute items are subdivided as “event class-subclass-change” for easy understanding of the contents, but this classification method is an example. For example, “material price (event class)” Instead of “oil price (subclass) -day change + (change)”, it is also possible to define the event type as one item, such as “oil price increase (day change)”.
Or, conversely, by adding attribute items such as “region” and “time” in addition to “event class-subclass-change”, the event attribute items can be further subdivided.

Each item included in the attribute item of the affected company stores information for identifying a company that affects the stock price due to the occurrence of the event.
For example, in the case of a record of “currency−euro / yen−y / y +”, “Eurozone” is set in the “Sales area” item, and “+” is set in the “Effect on stock price” item. As a result, the appreciation of the euro shows a positive impact on the share price of exporting companies engaged in selling products in the euro area.

  In addition, in the case of the record of “material price−oil price−y / y +”, “oil” is set in the “raw material” item, and “−” is set in the “effect on stock price” item. This shows a negative impact on the stock price of companies that use oil as a raw material for their products.

  In the case of the record “Natural Disaster-Earthquake-Blank”, “Production Area” is set as “Occurrence Area” and “Effect on Stock Price” is set as “−”. This shows that there is a negative impact on the stock price of companies engaged in production activities in the area where the earthquake occurred.

  In addition, in the case of a record of “Corporate situation-bankruptcy-blank”, “Partner” is set in the “Related company” item, and “-” is set in the “Effect on stock price” item. This shows that when a corporate failure occurs, the stock price of a company that has this business partner is negatively affected.

The ontology storage unit 42 stores three types of ontology information: (1) company ontology, (2) resource ontology, and (3) regional ontology.
Here, ontology information means information having a data structure in which the positional relationship (vertical relationship) and the significance between a plurality of components are clearly defined.

FIG. 7 is a conceptual diagram showing the configuration of a company ontology. Each “Corporation” has “Activity”, “Relation”, and “Resource” as subordinate components. It shows that it has.
This figure also includes “Sales”, “Purchase”, “Production”, etc. as specific contents of “Activity”, and “Business Ally ( "Business alliance", "Parent Company (parent company)", "Costomer (customer company)", etc., and the specific contents of "Resource" include "Brand (brand)", "Product (product)", "Service ( Service) ”and the like.
This figure also shows that “Corporation” is associated with “Name (company name)”, “Stock Ticker Number (stock code)”, “URL”, and “Address (head office address)” as specific values. Show.

  FIG. 8A shows the detailed structure of “Activity”, and each “Activity” has a “Resource” class as “what” and “where”. The “Location” class is provided as a component as “where / activity area”.

  FIG. 8 (b) shows the detailed structure of “Relation”. Each “Relation” has “Corporation” class as “who” and “what”. "," String (specific value) "as a component.

  FIG. 8C shows the detailed structure of “Resource”, and each “Resource” has “String (specific value)” as “category” and “name”. "," String (specific value) "as a component.

For example, the information of the meaning meaning “Toyo Motors sells the hybrid car“ Mevius ”in the UK” is expressed as corporate ontology information having the data structure shown in FIG. 9 and stored in the ontology storage unit 42. .
Further, the information having the meaning of “Bull PC and Ojima home appliances are in business alliance for personal computer sales” is expressed as corporate ontology information having the data structure shown in FIG. 10 and stored in the ontology storage unit 42.
These corporate ontology information is actually described by RDF (Resource Description Framework).

FIG. 11 is a conceptual diagram illustrating the structure of resource ontology information, and the specific content of “Resource” of a company is shown in a hierarchical structure.
For example, “automobile” is set as a concept below “Product”, and it is specified that this “automobile” includes “truck” and “hybrid car”.
In addition, it is specified that “gene recombination”, “semiconductor memory”, and the like are included as concepts under “Technology”.
Since only a limited number of resources are illustrated in FIG. 11 due to limited space, in reality, a large number of resources that are assumed as corporate resources are defined in detail across multiple layers. .

  In this way, by storing the positioning (hierarchical structure) between specific elements included in the “Resource” of the company in the ontology storage unit 42 as resource ontology information, the character string “track” becomes “Product It is clarified that it is a subordinate concept of “automobile” included in “”.

FIG. 12 is a conceptual diagram illustrating the structure of regional ontology information, and the specific contents of “Location (region)” are shown in a hierarchical structure.
For example, “United States” and “Canada” are set as one concept below “North America”, and “New York” and “California” are included as subordinate concepts of “United States”. Yes.
It also stipulates that “Japan” and “China” are included as concepts under “Asia”, and “Tokyo” and “Kanagawa Prefecture” are included as subordinate concepts of “Japan”. .
Due to the limited space, only a small number of resources are illustrated in FIG. 12, but in reality, a number of place names that are assumed as corporate activity areas are defined in detail across multiple layers. Yes.

  In this way, by storing the position (hierarchical structure) between specific elements included in “Location” in the ontology storage unit 42 as regional ontology information, for example, “United Kingdom” as a subordinate concept of “Euro area” ”,“ France ”,“ Germany ”, etc. are clarified.

  The event information storage unit 38 stores a large number of structured event information in chronological order. This event information is broadly classified into quantitative event information that has been shaped in advance into a necessary data structure and transmitted from the information distribution server 48, and qualitative event information extracted from text data by the system 10.

FIG. 13 shows an example of quantitative event information, which includes data items of subclass, time, and day-to-day ratio.
These quantitative event information is obtained by performing necessary conversion processing (extraction and shaping processing of necessary data) on the quantitative data transmitted from time to time by exchange information providers and commodity price information providers. Generated.

FIG. 14 shows an example of qualitative event information.
In other words, the qualitative event information in (a) indicates that an earthquake occurred in Niigata Prefecture, and “2009/11/14 10:29”, which is the date and time information when this qualitative event information was generated, is associated. ing.
The qualitative event information in (b) indicates that a terrorist attack has occurred in Dhaka, and the date and time information at the time this qualitative event information was generated is associated with “2009/11/18 16:07”. Yes.
Furthermore, the qualitative event information in (c) indicates that Company A has gone bankrupt, and is associated with “2009/11/20 17:19” which is date and time information at the time when this qualitative event information was generated. .

As mentioned above, when information distribution services such as exchange information and product price information already exist and information with a certain format is provided from there, the meaning of each data and the mutual Since the relationship is clear, it can be used as event information by performing a relatively simple data conversion process.
On the other hand, the first report that tells about the occurrence of natural disasters and conflicts is generally distributed in the form of unstructured text information. Of course, it is not impossible for humans to interpret such irregular text information described in a natural language one by one and convert it into a necessary structure and store it in the event DB 20.
However, it is inefficient to reserve a large number of people for an event that does not know when it will occur, resulting in lack of completeness and immediacy.
Therefore, the system 10 has a function of automatically generating qualitative event information from a large amount of text data collected regularly or irregularly from a plurality of news sites on the Internet. Hereinafter, this function will be described in detail.

  First, the morpheme analysis processing unit 12 performs morpheme analysis on the text data 49 input from the outside. Here, “morpheme analysis” refers to a process of decomposing a sentence written in a natural language into morphemes, which are the smallest meaningful language units, and specifying each part of speech.

For example, when a sentence “Somy has gone bankrupt due to the application of the corporate reorganization law” is given, as shown in FIG. 15, the morphological analysis processing unit 12 converts this into “Somy / Noun”, “Ga / It decomposes into “participant”, “company rehabilitation / noun”, “no / particle”, “application / noun”, “ha / particle”, “receiver / verb”, etc., and identifies each part of speech.
This morpheme analysis itself is a known technique. For example, the following free software can be used as a morpheme analysis engine.
(1) MeCab (http://mecab.sourceforge.net/)
(2) ChaSen (http://chasen.naist.jp/hiki/ChaSen/)

Next, the morpheme analysis processing unit 12 refers to an event dictionary, an event predicate dictionary, a regional dictionary, a company name dictionary, and an activity object dictionary stored in the dictionary DB 26, and an abstract tag (abstract) corresponding to a part of speech of a specific morpheme. Supplementary character string).
For example, since “Somy” has a registered example in the company name dictionary, an abstract tag “<company name>” is added to the part of speech item. In addition, since there is a registered example in the event predicate dictionary regarding “bankruptcy”, an abstract tag “<company situation>” is added to the part of speech item.

Next, the compound word analysis processing unit 14 is activated, and by referring to the compound word analysis rule stored in the compound word analysis rule storage unit 30, the morpheme analysis processing unit 12 selects from the sentences decomposed into morpheme units. Authorize compound words.
As shown in FIG. 16 (a), the compound word analysis rule includes data items of a part of speech connection pattern and a part of speech determination criterion, and the compound word analysis processing unit 14 includes a morpheme that matches the part of speech connection pattern in the sentence. When the sequence is found, these morphemes are connected as compound words, and the parts of speech of the compound words are identified according to the corresponding part of speech determination criteria.

  For example, as shown in FIG. 16 (b), three morphemes of “natural (noun-adjective verb stem)”, “language (noun-general)”, and “processing (noun-sa-variant connection)” were consecutive in the sentence. In this case, the compound word analysis processing unit 14 recognizes the compound word of “natural language processing” because the connection pattern of each part of speech matches the compound word analysis rule (1), and then, based on the part of speech determination criteria of (1). The part of speech is recognized as “noun-general”.

  Also, as shown in FIG. 16 (c), when two morphemes of “high class (noun-adjective verb stem)” and “cosmetics (noun-general)” are consecutive in the sentence, the compound word analysis processing unit 14 Since the connection pattern of each part of speech matches the compound word analysis rule (2), it is recognized as a compound word of `` high-end cosmetics '', and then the part of speech is defined as `` noun-general '' based on the part of speech determination criteria of (2). Authorize.

  Furthermore, as shown in FIG. 16 (d), when two morphemes of “production (noun-sa-variant connection)” and “quantity (noun-suffix)” are consecutive in the sentence, the compound word analysis processing unit 14 Because the connection pattern of part-of-speech matches the compound word parsing rule (3), the part-of-speech is recognized as a compound word of “production volume” and then the part-of-speech is recognized as “noun-general” based on the part-of-speech determination criteria To do.

Next, the abstraction processing unit 16 is activated and associates an abstract tag such as a company name or a region with a morpheme in the sentence.
As described above, the morphological analysis processing unit 12 refers to the dictionary storage unit 26 in advance, and a corresponding abstract tag is assigned to the company name or region recorded in the dictionary. Are naturally limited, and a dictionary-based abstraction process alone may cause a leak.
For this reason, the abstraction processing unit 20 executes abstraction processing based on regular expression rules stored in the abstraction rule storage unit 32, and the corresponding abstract tag for company names and regions not recorded in the dictionary. It has a function to associate

  FIG. 17A shows an example of an abstraction rule. “<Country> of <company_size> (<feature: noun> +)” is “company_size (a character string representing a company size)” + “ It is defined that a noun immediately following “+ country (a character string representing a country)” is recognized as a company name. Further, “first, major, middle-ranking” is defined as an alias expression (alias) of “company_size”, and “US, UK, Europe” is defined as an alias expression of “country_size”.

  Here, as shown in FIG. 17 (b), when a sentence “Retail major US AAA Mart has announced a staff reduction plan” is given, the abstraction processing unit 20 converts this to FIG. 17 (c). ) And replace it with the noun unit OR expression, extract the `` retail major US AAA mart '' that matches the rule, then use the regular expression `` backward reference '' to extract `` AAA mart '' Authorize.

Next, the anaphora analysis processing unit 18 is activated, and the antecedents (pronouns and the like) in the sentence are supplemented by referring to the anaphora analysis rules stored in the anaphora analysis rule storage unit 34.
As shown in FIG. 18 (a), this anaphoric analysis rule includes data items of anaphoric and antecedent determination criteria. When the anaphoric analysis processing unit 18 finds the defined anaphor in the sentence, The antecedent of the anaphor is identified according to the corresponding antecedent determination criteria.

For example, as shown in Fig. 18 (b), when there is a sentence "The company will sell the product from 14th", the anaphora analysis processing unit 18 first sets the anaphora analysis rule (2). Detects that it falls under the antecedent, recognizes “company B” with the latest <company name> tag as an antecedent according to the antecedent determination criteria, and replaces it with “company” in the sentence.
Next, the anaphora analysis processing unit 18 detects that “the same product” in the sentence corresponds to the antecedent of the anaphora analysis rule (3), and attaches the latest <production object> tag according to the antecedent determination criterion. Recognize “New PC” as antecedent and replace it with “Same Product”.

Next, the sentence type determination processing unit 20 is activated, and identification information indicating that it is a title sentence is given to each sentence corresponding to the “title sentence”. Whether a given sentence is a title sentence or a normal sentence (body) is determined by tag information described in a Web file from which the text data 49 is collected. A specific example of the title sentence will be described later.
The sentence to which the identification information of “title sentence” is given is treated differently from the normal sentence to which these identification information is not given in the case slot filling process in the next stage.

Next, the case slot filling processing unit 22 is activated to execute a word (morpheme or compound word) filling process for the case slot set in the memory.
FIG. 19 shows an example of a case slot, which includes items of “particle” and “corresponding word”. In addition, necessary particles (ie, particle particles, case particles) such as (ha), (ga), (), (ni)... Are set in advance in the item of particles.

  Here, as shown in FIG. 20, when the sentence “Somy has gone bankrupt due to the application of the corporate rehabilitation law” is given, the case slot filling processing unit 22 begins the word at the corresponding part of the case slot. Fill in order. For example, since the phrase “Somy is” includes the particle “GA”, “SOMMY”, which is an independent word in the phrase, is filled in the corresponding word storage column of (GA). Similarly, since the phrase “company rehabilitation method” includes the particle “no”, the word “company rehabilitation method”, which is the immediately preceding word, is filled in the corresponding word storage column of (no). Similarly, since the phrase “apply” includes the particle “”, “apply”, which is the immediately preceding word, is filled in the corresponding word storage column of (). Note that the corresponding word storage column for particles other than (ga), (), and (no) is left blank.

Next, the case slot filling processing unit 22 associates “bankruptcy” which is a statement predicate with the case slot.
Generally speaking, “predicate” means a verb, an adjective, a noun + determinant that describes the behavior, state, and nature of the subject, but the “predicate” extracted from the sentence by the case slot filling processing unit 22 Which should be a “predicate” of event information to be extracted, specifically <disaster occurrence>, <conflict occurrence>, <company situation>, etc. corresponding to abstract character strings in the event predicate dictionary Corresponds to the word with the abstract tag.

  For the case slot filled with the necessary words by the case slot filling processing unit 22, the event extraction processing unit 24 at the next stage extracts a plurality of extraction frames for company situation events stored in the extraction frame definition storage unit 36. Predetermined event information is extracted by applying a definition that matches a condition.

FIG. 21 shows an application example of this extracted frame definition. When the following conditions (1) and (2) are all satisfied, the extracted frame definition 50 includes event information from the case slot. It is specified that it should be extracted.
(1) The case slot (GA) or (HA) is filled with the word with the <company name> abstract tag.
(2) As a predicate of the case slot, a word with the <company situation> abstract tag is associated.

The event extraction processing unit 24 applies the extracted frame definition 50 to the filled case slot 52 passed from the case slot filling processing unit 22 and meets the conditions (1) and (2) above. Generates event information 54 in which the word filled in () or () of the case slot is the “subject” and the word associated with the case slot as a predicate is the “predicate”.
The event information 54 is stored in the event information storage unit 38 by the event extraction processing unit 24.

Next, as shown in FIG. 22, when the sentence “Sommy has just been publicly traded in October last year” is given, the case slot filling processing unit 22 performs the case slot predicate as described above. The filling process is executed.
At this time, the empty case slot shown in FIG. 19 is not used, but the case slot immediately before the filling of the corresponding word is copied, and the corresponding column is overfilled with the word of the next sentence. As a general rule, if a word is filled in the corresponding word storage column of (ha) case or (ga) case in the next sentence, the case slot filling processing unit 20 determines that the topic has been changed and responds. Cancel word inheritance.

Specifically, as shown in FIG. 22 (a), for the case slot that is once inherited from the previous sentence, the result of filling the corresponding word storage column of “ha” in the subsequent sentence with “Somy” in the subsequent sentence. , (So) for (), (Apply) for () and (Company Rehabilitation Law) for () are deleted, and (Stock) for ( “Last October” is newly filled.
For this new case slot, the case slot filling processing unit 22 associates the character string “open” as a predicate.

  Although illustration is omitted, by applying the extraction frame definition 50 shown in FIG. 21 to this case slot, the event extraction processing unit 24 causes the event “(Company name: Sommy) → (Company situation: Public)”. Information is extracted and stored in the event information storage unit 38.

Next, as shown in FIG. 22 (b), the case slot filling processing unit 22 proceeds to the case slot filling processing of the following sentence, “After all, it has failed in just one year”. In this case, the clause itself containing (ha) and (ga) does not exist in the sentence, and the subject is omitted, so the upper case of the case slot inherited from the previous sentence over (ha) or (ga) Since the filling of the book does not occur, the corresponding word of the inherited case slot is not cleared.
Therefore, “Somy” is retained for (ha), “Stock” is retained for (), “October last year” is retained for (ni), and “One year” is newly retained for (de). Filled.
For this case slot, the case slot filling processing unit 22 associates a character string of “failure” as a predicate.

  Although not shown, by applying the extracted frame definition 50 shown in FIG. 21 to this case slot as well, the event extraction processing unit 24 becomes “(company name: Sommy) → (company situation: bankruptcy)”. Event information is extracted and stored in the event information storage unit 38.

Next, as shown in FIG. 23, when the sentence “On the other hand, the competing harp achieved increased sales and profits this term”, the case slot filling processing unit 22 temporarily sets the case slot of the previous sentence. Inherit, but when the corresponding “harp” (<company name>) is overwritten in the corresponding word storage field of (ha), it is determined that the topic has changed and other corresponding words are stored. After clearing the corresponding words filled in the column, “increased sales and profits” is filled in () and “competition” is filled in ().
The case slot filling processing unit 22 associates the predicate “increased sales and profits” with the new case slot.

  Although illustration is omitted, by applying the extraction frame definition 50 shown in FIG. 21 to this case slot, the event extraction processing unit 24 becomes “(company name: harp) → (company situation: increased sales and profits)”. Event information is extracted and stored in the event information storage unit 38.

FIG. 24 shows a specific example of extracting event information related to the occurrence of a disaster. Even when a sentence “A strong earthquake of seismic intensity 6 occurred in Niigata Prefecture” is given, case slot filling processing is performed. In the same manner as described above, the unit 22 fills a corresponding portion of the case slot with a word. In other words, since the phrase “in Niigata Prefecture” includes the particle “de”, the independent word “Niigata Prefecture” in the phrase is filled in the corresponding word storage column of (de). Similarly, since the phrase “seismic intensity 6” includes the particle “no”, the previous word “seismic intensity 6” is filled in the corresponding word storage column of (no). Similarly, since the phrase “earthquake” includes the particle “ga”, the word “earthquake” which is the immediately preceding word is filled in the corresponding word storage column of (ga).
The case slot filling processing unit 22 associates “occurrence” assigned with the <disaster occurrence> abstract tag as a predicate with this case slot.

  The event extraction processing unit 24 applies the extracted frame definition for natural disaster events stored in the extracted frame definition storage unit 36 to the case slot filled with the necessary words by the case slot filling processing unit 22. To extract event information.

FIG. 25 shows an application example of this extracted frame definition. When the extracted frame definition 56 satisfies all the following conditions (1) to (3), the event information from the case slot is shown. It is specified that it should be extracted.
(1) Words with <natural disaster> abstract tag are filled in (g) or (ha) of case slots.
(2) Words with the <disaster occurrence> abstract tag are associated as case slot predicates.
(3) Words with the <region> abstract tag are filled in the case slots (ni), (de), or (de).

The event extraction processing unit 24 applies the extracted frame definition 56 to the filled case slot 58 passed from the case slot filling processing unit 22, and confirms that the conditions (1) to (3) are met. After confirmation, “earthquake” as a natural disaster filled in the case slots, “occurrence” as a predicate, and “Niigata Prefecture” as a region are extracted, and event information 60 is generated.
The event information 60 is stored in the event information storage unit 38 by the event extraction processing unit 24.

FIG. 26 shows a specific example when extracting event information related to the occurrence of a conflict. Even when a sentence “Coup has broken out in country X” is given, the case slot filling processing unit 22 As with, fill the relevant places in case slots with words. That is, since the phrase “in country X” includes the particle “de”, “X country”, which is an independent word in the phrase, is filled in the corresponding word storage field of (de). Similarly, since the phrase “coup d'etre” includes the particle “ga”, the “coup d'etre” that is an independent word in the phrase is filled in the corresponding word storage column of (g).
The case slot filling processing unit 22 associates “breakout” with the <conflict occurrence> abstract tag as a predicate with this case slot.

  For the case slot filled with the necessary words by the case slot filling processing unit 22, the event extraction processing unit 24 applies the extracted frame definition for the conflict event stored in the extracted frame definition storage unit 36. Extract predetermined event information.

FIG. 27 shows an application example of this extracted frame definition. The extracted frame definition 62 includes event information from the case slot 64 when all of the following conditions (1) to (3) are satisfied. It is stipulated that should be extracted.
(1) Words with the <conflict> abstract tag are filled in the case slots (GA) or (HA).
(2) Words with the <conflict occurrence> abstract tag are associated as case slot predicates.
(3) Words with the <region> abstract tag are filled in the case slots (ni), (de), or (de).

The event extraction processing unit 24 applies the extracted frame definition 62 to the filled case slot 64 passed from the case slot filling processing unit 22 and confirms that the conditions (1) to (3) are met. After confirming, the event information 66 is generated by extracting the “coup” as the conflict filled in the case slot, the “breakout” as the predicate, and the “country X” as the region.
The event information 66 is stored in the event information storage unit 38 by the event extraction processing unit 24.

  The case slot filling process described above is an example in the case of a normal natural sentence, but if the identification information indicating that the given sentence is a title sentence is given, the particle is omitted. In many cases, the case slot filling processing unit 22 estimates the omitted particles and then executes word filling processing for the case slots.

For example, as shown in FIG. 28, when the title sentence “Typhoon No.18, 15th Dawn, Okinawa Landing” is given, the case slot filling processing unit 22 is attached with an abstract tag of <natural disaster> For the “typhoon” to be the subject of the information, it is presumed that the particle of (ha) or (ga) is omitted, and (ty) and (ga) of the case slot are filled with “typhoon”.
Next, the case slot filling processing unit 22 has the <region> abstraction tag for “Okinawa”, so it is a particle used together with the word representing the region (ni), (de), ( ) Is assumed to be omitted, and “Okinawa” is filled in (in), (in), and (in) of the case slots.
Next, the case slot filling processing unit 22 recognizes that “landing” is a predicate because the <disaster occurrence> abstract tag is given, and associates “landing” with the case slot as a predicate.

  The event extraction processing unit 24 applies the extracted frame definition for natural disaster events stored in the extracted frame definition storage unit 36 to the case slot filled with the necessary words by the case slot filling processing unit 22. Thus, predetermined event information is extracted.

FIG. 29 shows the event information extraction state. The event extraction processing unit 24 applies the extracted frame definition 56 to the filled case slot 68 passed from the case slot filling processing unit 22. , “Typhoon” as a natural disaster filled in (ha) and (ga) of case slots, “landing” as predicate, and (in), (in) and (in) of case slots The event information 70 is generated by taking out “Okinawa” as a region.
The event information 70 is stored in the event information storage unit 38 by the event extraction processing unit 24.

  In the case of an information extraction method using a conventional parsing technique, it is necessary to define the dependency structure between clauses to be extracted as a syntax pattern. In this way, a method using a syntax pattern that depends on the structure of the sentence is used. However, since it is impossible to extract the target information just by changing the order of words, it is necessary to prepare a syntax pattern in consideration of all variations of sentences.

On the other hand, in the case of this system 10, a typical case slot in which the necessary particle types are set in advance as described above, and an extraction frame definition that specifies the type of information to be extracted are prepared. If there is a corresponding particle in the sentence, the previous independent word is filled in the corresponding word storage column of the particle, and the extracted frame definition is applied to this filled case slot, regardless of the word order. It is possible to reliably extract structured information such as “subject-predicate-region” from a natural sentence. In addition, since the word order is not constrained, variations in the extracted frame definition can be suppressed.
In addition, as a general rule, the case slot of the previous sentence is inherited by the next sentence, so even if there is a subject or region abbreviation in the subsequent sentence, it can be easily changed in the subject or area of the previous sentence. Can be supplemented.

In the above, the example in which the event extraction processing unit 24 immediately executes the event information extraction processing on the filled case slot passed from the case slot filling processing unit 22 has been described, but the present invention is not limited thereto. Is not to be done.
That is, the case slots that have been filled with the necessary words by the case slot filling processing unit 22 are stored in the filled case slot storage unit (not shown), and the event extraction processing unit 24 sequentially extracts the extracted frame definitions. By applying, various event information may be extracted together.

Next, a processing procedure related to detection of a stock price-affected company will be described with reference to the flowchart of FIG.
First, the stock price-affected company extraction unit 26 periodically checks the event information storage unit 38 (S10), and determines whether new event information has been registered between the previous check and the present time.
Here, when new event information is registered (S12 / YES), the stock price-affected company extraction unit 26 refers to the inference rule storage unit 40 and ontology storage unit 42, and a company whose stock price is affected by the event information. Is specified (S14).

  For example, if the newly registered event information conveys the fact that “an earthquake has occurred in Niigata Prefecture”, the stock price affected company extraction unit 26 refers to the table shown in FIG. Recognize that there will be a negative impact on the stock price of companies engaged in production activities.

Next, the stock price-affected company extraction unit 26 refers to the company ontology information stored in the ontology storage unit 42, acquires the names and brand codes of companies that are engaged in production activities in Niigata Prefecture, and each company name and brand name. A stock price affected company list in which codes are arranged in a predetermined order is generated (S16).
Specifically, “a company engaged in production activities in Niigata Prefecture” corresponds to a corporation that satisfies the conditions of “Activity = Production” and “where → Niigata Prefecture” in the company ontology information.

Finally, the stock price affected company extraction unit 26 transmits the stock price affected company list to the Web server 44 (S18).
The Web server 44 distributes a Web file including the stock price affected company list to the client terminal 47.
As a result, a report screen 74 including a stock price minus affected company list 72 is displayed on the Web browser of the client terminal 47 as shown in FIG.
If the stock owned by the company is listed in the stock price negative impact company list 72, the user clicks the order button 76 to send the sell order of the stock to the securities company's host computer, and buys or sells the necessary stock. Can be completed quickly.

  In addition, when the newly registered event information is to convey the fact that the euro / yen has increased (yen appreciation) compared to the previous day, the stock price-affected company extraction unit 26 refers to the table shown in FIG. Recognize that there will be a positive impact on the share price of exporters selling products in the euro area.

Next, the stock price-affected company extraction unit 26 refers to the regional ontology information stored in the ontology storage unit 42, and acquires the names (UK, France, Germany, Italy ...) of individual countries belonging to the euro area.
Next, the stock price-affected company extraction unit 26 refers to the company ontology information stored in the ontology storage unit 42, acquires the name and brand code of the company that sells products in the above countries, A stock price plus affected company list in which names and brand codes are arranged in a predetermined order is generated.
The “company that conducts sales activities in country X” specifically corresponds to a corporation that satisfies the conditions of “Activity = Sales” and “where → Country X” in the company ontology information.

Finally, the stock price affected company extraction unit 26 transmits the stock price plus affected company list to the Web server 44.
The Web server 44 distributes the Web file including the stock price plus affected company list to the client terminal 47.
As a result, a screen including a stock price plus affected company list is displayed on the Web browser of the client terminal 47 (not shown). On the other hand, the user can select one or a plurality of stocks from the stock price plus influence list and click a corresponding order button to issue a stock buying order to the securities company's host computer.

  Also, if the newly registered event information conveys the fact that Company A went bankrupt, the stock price-affected company extraction unit 26 refers to the table shown in FIG. Recognize that there will be a negative impact on stock prices.

Next, the stock price-affected company extraction unit 26 refers to the company ontology information stored in the ontology storage unit 42, acquires the name and brand code of the company whose business is A company, and each company name and brand code A stock price minus influence company list is generated by arranging the items in a predetermined order.
“Company with company A as a business partner” specifically corresponds to a corporation that satisfies the conditions of “Relation = Business Ally” and “who → Company A” in the company ontology information.

Finally, the stock price affected company extraction unit 26 transmits the above stock price minus affected company list to the Web server 44.
The Web server 44 distributes a Web file including the stock price minus affected company list to the client terminal 47.
As a result, a screen including a stock price minus affected company list is displayed on the Web browser of the client terminal 47 (not shown). On the other hand, the user can select one or a plurality of stocks from the stock price negative influence list and click a corresponding order button to issue a stock selling order to the host computer of the securities company.

It is a block diagram which shows the function structure of the stock price influence company detection system which concerns on this invention. It is a chart which illustrates the registration contents of an event dictionary. It is a chart which illustrates the registration contents of an event predicate dictionary. It is a chart which illustrates the registration contents of a regional dictionary. It is a chart which shows an example of inference rule data stored in inference rule DB. It is a chart which shows an example of inference rule data stored in inference rule DB. It is a conceptual diagram which shows the structure of company ontology information. It is a conceptual diagram which shows the structure of company ontology information. It is a conceptual diagram which shows the specific example of company ontology information. It is a conceptual diagram which shows the specific example of company ontology information. It is a conceptual diagram which shows the structure of resource ontology information. It is a conceptual diagram which shows the structure of regional ontology information. It is a figure which shows an example of fixed quantity event information. It is a figure which shows an example of qualitative event information. It is a graph which shows the result of a morphological analysis. It is explanatory drawing which shows a compound word analysis rule and its application example. It is explanatory drawing which shows the abstraction process by an abstract rule. It is explanatory drawing which shows an anaphora analysis rule and its application example. It is a conceptual diagram which shows the structure of a case slot. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows an extraction frame definition and its application example. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows an extraction frame definition and its application example. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows an extraction frame definition and its application example. It is explanatory drawing which shows the example of filling of the word with respect to a case slot. It is explanatory drawing which shows an extraction frame definition and its application example. It is a flowchart which shows the process sequence which concerns on the detection of a stock price influence company. It is a block diagram of the report screen containing a stock price minus influence company list.

10 Stock price impact company detection system
12 Morphological analysis processor
14 Compound word analysis processing section
16 Abstraction processing section
18 Anaphoric analysis processing section
20 sentence type determination processing section
24 Event extraction processing section
26 Stock Price Impact Company Extraction Department
28 Dictionary storage
30 Compound word analysis rule storage
32 Abstraction rule storage
34 Anaphoric analysis rule storage
36 Extracted frame definition storage
38 Event information storage
40 Inference rule memory
42 Ontology storage
44 Web server
46 Communication network
47 Client terminal
48 Information distribution server
49 Text data
50 Extraction frame definition
52 filled slot
54 Event information
56 Extraction frame definition
58 filled slot
60 Event information
62 Extraction frame definition
64 filled slot
66 Event information
68 filled slot
70 Event information
72 List of companies with negative impact on stock prices
74 Report screen
76 Order button

Claims (7)

Event information storage means for storing event information having specific attributes that affect the stock price;
Company information storage means for registering attribute information of each company;
An inference rule storage means for storing an inference rule that defines whether the influence on the stock price is positive or negative for each combination pattern of the attribute of the event and the attribute of the company affected by the stock price by the event;
Means for applying the inference rules to each event information stored in the event information storage means, and identifying a company attribute that affects the stock price by the event;
A stock price-affected company extracting means for referring to the company information storage means and extracting a company having the attribute as a stock price-affected company;
Means for generating and outputting a list of the above-mentioned stock price-affected companies,
The attribute of the event information consists of a combination of a character string representing the type of event and a character string representing the region where the event occurred,
The attribute information of the company includes specific types of regional information related to the company,
The above inference rules define whether the impact on stock prices is positive or negative for each combination pattern with the type of event and the specific type of region related to the company.
The stock price-affected company extraction means is a stock price-affected company detection system that extracts, as a stock price-affected company, a company in which the event information occurrence area is registered as a specific type of area related to the company defined in the inference rule. ,
In addition, an event dictionary that defines the correspondence between abstract tags indicating the types of higher-level conceptual events and character strings that represent specific event names,
An area dictionary that defines the correspondence between abstract tags that indicate higher conceptual areas and character strings that represent specific area names,
An event predicate dictionary that defines the correspondence between an abstract tag that indicates the occurrence of a high-level conceptual event and a character string that represents the specific event occurrence;
Means for decomposing sentences in text data into morpheme units and identifying parts of speech of each morpheme;
Referring to each of the above dictionaries, each of the morphemes corresponds to a character string representing the name of the event, a character string representing the region name, and a character string representing the occurrence of the event. A means of associating an abstract tag indicating the type, an abstract tag indicating the region, an abstract tag indicating the event occurrence,
At least for each particle attached to the subject and each particle attached to the region, a case slot provided with a storage column for the corresponding word is filled with the corresponding word in the sentence, and an abstract tag indicating the occurrence of the event is given. Case slot filling means for associating a morpheme as a predicate with the case slot;
A condition for specifying an abstract tag of a subject to be extracted and a particle attached to the subject, a condition for specifying an abstract tag of a predicate to be extracted, an abstract tag of an area to be extracted, and a particle attached to the area Extraction frame definition storage means for storing a plurality of extracted frame definitions in which at least a condition for specifying is specified;
Information extracting means for extracting at least information elements corresponding to the subject, predicate, and region of the sentence by applying the extracted frame definition to the case slots filled with corresponding words, and storing the information elements in the event information storage means; ,
The stock price-affected company detection system, wherein the case slot filling means executes the following processing.
(1) The case slot in which the filling of the corresponding word with respect to the preceding predicate is completed is inherited for the subsequent predicate.
(2) Overwrite the corresponding word for the following predicate in the corresponding word storage field of the case slot.
(3) When the corresponding term storage column of the particle representing the subject of the sentence is overwritten with a word representing the subject related to the subsequent predicate, the corresponding word filled with respect to the preceding predicate is deleted. A means for preliminarily providing identification information indicating a title sentence with respect to a sentence corresponding to the title in the above sentence,
For the words to which the case tag filling means has attached an abstract tag indicating the type of event in the title sentence to which this identification information is assigned, the corresponding word storage column of the particle attached to the subject regardless of the presence or absence of the particle 2. The stock price effect according to claim 1 , wherein words filled with an abstract tag indicating a region are filled in a corresponding word storage column of a particle attached to the region name regardless of the presence or absence of the particle. Company detection system. Event information storage means for storing event information having specific attributes that affect the stock price;
Company information storage means for registering attribute information of each company;
An inference rule storage means for storing an inference rule that defines whether the influence on the stock price is positive or negative for each combination pattern of the attribute of the event and the attribute of the company affected by the stock price by the event;
Means for applying the inference rules to each event information stored in the event information storage means, and identifying a company attribute that affects the stock price by the event;
A stock price-affected company extracting means for referring to the company information storage means and extracting a company having the attribute as a stock price-affected company;
Means for generating and outputting a list of the above-mentioned stock price-affected companies,
The event information attribute consists of a combination of a character string that identifies the company name that is the subject of the event and a character string that represents the type of event related to the company situation,
The attribute information of the company includes specific types of related company information related to the company,
The above inference rules define whether the impact on the stock price is positive or negative for each combination pattern with the type of event related to the corporate situation and the type of related company.
The above-mentioned stock price-affected company extraction means extracts a stock price-affected company as a stock price-affected company that has registered the company name that is the subject of the event information related to the corporate situation as a specific type of related company defined in the above inference rules. A detection system,
In addition, a company name dictionary that defines the correspondence between an abstract tag that indicates a generic company name and a character string that represents a specific company name,
An event predicate dictionary that defines the correspondence between an abstract tag that indicates the occurrence of a higher-level conceptual event related to the corporate situation and a character string that indicates the actual occurrence of an event related to the corporate situation;
Means for decomposing sentences in text data into morpheme units and identifying parts of speech of each morpheme;
Referring to each of the above dictionaries, each morpheme must be a company name for a character string that represents a specific company name and a character string that represents the occurrence of an event related to the company situation. Means for associating an abstract tag that indicates the fact that an event related to the corporate situation has occurred,
At least a case slot with a corresponding word storage field for each particle attached to the subject is filled with the corresponding word in the sentence, and a morpheme with an abstract tag indicating the occurrence of the event is used as a predicate. A case slot filling means for associating the above descriptive word with the slot;
An extraction frame that stores a plurality of extraction frame definitions that specify at least a condition for specifying an abstract tag of a subject to be extracted and a particle attached to the subject and a condition for specifying an abstract tag of a predicate to be extracted. Definition storage means;
By applying the extracted frame definition to the case slot filled with the corresponding word, the information extraction means extracts at least the information element corresponding to the subject of the sentence and the predicate, and stores it in the event information storage means,
The stock price-affected company detection system, wherein the case slot filling means executes the following processing.
(1) The case slot in which the filling of the corresponding word with respect to the preceding predicate is completed is inherited for the subsequent predicate.
(2) Overwrite the corresponding word for the following predicate in the corresponding word storage field of the case slot.
(3) When the corresponding term storage column of the particle representing the subject of the sentence is overwritten with a word representing the subject related to the subsequent predicate, the corresponding word filled with respect to the preceding predicate is deleted. A means for preliminarily providing identification information indicating a title sentence with respect to a sentence corresponding to the title in the above sentence,
For the words with the abstract tag indicating that the case slot filling means is the company name in the title sentence to which this identification information is given, the corresponding word storage of the particle attached to the subject regardless of the presence or absence of the particle The stock price-affected company detection system according to claim 3 , wherein the space is filled in a column. Compound word analysis rule storage means for storing a compound word analysis rule in which a criterion for determining the part of speech of the compound word is defined for each connection pattern of a plurality of parts of speech to be a compound word;
By referring to this compound word analysis rule, if there is a combination of morphemes corresponding to the part-of-speech concatenation pattern specified in the compound word analysis rule in the sentence, compound word analysis that recognizes these morphemes as compound words Means and
The stock price according to any one of claims 1 to 4 , wherein the case slot filling means executes a filling process into a case slot in units of compound words for a combination of morphemes recognized as compound words. Affected company detection system. Abstraction rule storage means for storing abstraction rules for estimating the type of morpheme;
Means for applying the above abstraction rule to a sentence and associating an abstract tag indicating the type with a morpheme that matches the abstraction rule;
The stock price-affected company detection system according to any one of claims 1 to 5 , further comprising: An anaphora analysis rule storage means for storing an anaphora analysis rule that defines a criterion for determining the antecedent for each anaphor;
The anaphoric analysis means is provided that refers to the anaphoric analysis rule, determines an antecedent corresponding to the anaphor in the sentence, and replaces the anaphor with the antecedent . 6. The stock price-affected company detection system described in any one of 6 above.

Images