WO2011155736A2 - Procédé de production dynamique de termes supplémentaires pour chaque sens de chaque expression en langage naturel ; gestionnaire de dictionnaire, dispositif de production de documents, annotateur de termes, système de recherche et dispositif de construction d'un système d'informations sur des documents basé sur le procédé - Google Patents

Procédé de production dynamique de termes supplémentaires pour chaque sens de chaque expression en langage naturel ; gestionnaire de dictionnaire, dispositif de production de documents, annotateur de termes, système de recherche et dispositif de construction d'un système d'informations sur des documents basé sur le procédé Download PDF

Info

Publication number
WO2011155736A2
WO2011155736A2 PCT/KR2011/004113 KR2011004113W WO2011155736A2 WO 2011155736 A2 WO2011155736 A2 WO 2011155736A2 KR 2011004113 W KR2011004113 W KR 2011004113W WO 2011155736 A2 WO2011155736 A2 WO 2011155736A2
Authority
WO
WIPO (PCT)
Prior art keywords
semantic unit
term
document
search
natural language
Prior art date
Application number
PCT/KR2011/004113
Other languages
English (en)
Korean (ko)
Other versions
WO2011155736A3 (fr
WO2011155736A9 (fr
Inventor
박동민
Original Assignee
Park Dong Min
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Park Dong Min filed Critical Park Dong Min
Publication of WO2011155736A2 publication Critical patent/WO2011155736A2/fr
Publication of WO2011155736A3 publication Critical patent/WO2011155736A3/fr
Publication of WO2011155736A9 publication Critical patent/WO2011155736A9/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/31—Indexing; Data structures therefor; Storage structures
    • G06F16/313—Selection or weighting of terms for indexing

Definitions

  • the present invention relates to a term dictionary, a document writer, and an information retrieval involved in generating information, collecting, indexing, searching, and using information, and a term commenter, a document information system construction device, and a semantic web for making them based on semantic terms. (Semantic Web) is included.
  • the technical field to which the present invention belongs is the field of information retrieval. Since the present invention relates to semantic-based information retrieval, the semantic web field is related to the information retrieval field.
  • the Semantic Web represents the information about resources (web documents, various files, services, etc.) and the relations between resources in a distributed environment such as the Internet.
  • the meaning information (Semanteme) is expressed in an ontology that a machine (computer) can process. It is a framework technology that allows automated machines (computers) to process them.
  • Ontology is a formal specification of shared conceptualization of domains and expresses semantic information of domain vocabulary. Ontology is a kind of knowledge representation, and the computer can understand the concept represented by the ontology and process the knowledge. Ontology's axioms and rules are used for inference and proofing, and a separate rule language is used for rule expression.
  • the technical field to which the present invention belongs is the field of information retrieval.
  • the current level of search technology can be clearly seen through major search engines.
  • the major search technologies are natural language-based search technology. Since information is accumulated using unclear natural language and natural language search query is used, it has low search accuracy rate based on meaning.
  • Hong Gil-Dong (a pseudonym). Searching for information about the inventor under the name of Hong Gil-dong includes unnecessary information about 640 people other than the inventor. In this case, the search accuracy rate for the keyword Hong Gil-dong of the natural language search system is 1/641 on average. Indeed, current search engines display a lot of material, but often they don't really want it.
  • the total number of word types and the total number of meanings are proportional to the amount of overall meaning-based efforts.
  • the present invention finds several meanings in a specific natural language sorted content of a search system, generates a term, annotates this new term in the index, and eventually converts the entire Internet documents into semantic unit terms based on the index. It has the same effect as one.
  • Semantic term-based indexing can be used to convert all documents to semantic terminology.
  • this method unlike the ontology dictionary, this method generates terms through the simple task of dividing the natural language into semantic units, so that the general public can easily participate in the task of generating terms and converting documents / indexes into semantic unit terms. When general users generate only a few terms that they are interested in and have knowledge of, search for the corresponding natural language on the Internet and comment out the newly created semantic unit term, it is possible to convert the semantic unit term based on the whole Internet.
  • the present invention improves the meaning-based accuracy rate by several times, tens of times to hundreds of times, in some cases, compared to existing search engines.
  • a single term often has various meanings, and numerous proper nouns such as human names, store names, place names, etc. are invaded to general nouns, verbs, and adjectives.
  • the present invention uses the semantic unit terminology to improve the level of accuracy of the expression unit to the correct rate of the semantic unit by complementing the unclear natural language.
  • the present invention does not merely suggest a new model but includes a method in which the new model can be well established.
  • 1 is a block diagram of a semantic unit term-based information system
  • Figure 2 is the construction of a two-step semantic unit term-based information system
  • 9 is a flowchart illustrating generation of semantic unit terms
  • FIG. 14 is a flowchart of generating, commenting, and searching a semantic unit term segment.
  • 16 is a flowchart illustrating generation and use of semantic unit terminology group.
  • 17 is a block diagram of the center of an independent tin machine
  • 19 is an example of default values corresponding to a specific user.
  • 20 is a flowchart illustrating the determination of semantic unit default values for natural language expressions.
  • 21 is a conceptual structure of an annotation knowledge table
  • 25 is a flowchart in which a knowledge base annotation unit is performed on a document or a query word.
  • 26 is a flowchart for performing annotation knowledge on an index.
  • 27 is a document annotation of an index-based document annotation portion.
  • Fig. 28 is a flowchart for annotating semantic unit terms to specific natural language expressions in an indexed document.
  • 29 is a scale of a semantic unit term (unique ID +) based information system.
  • 32 is a manual annotation type document builder and an automatic annotation type document builder
  • 35 is a minimum configuration of a search system
  • 36 is a diagram illustrating a search commenter added to a search system minimum configuration.
  • 40 is an operation flowchart of a search system using basic functions and a search annotation function.
  • Fig. 41 is a flowchart showing the operation of the search system using basic functions and annotation knowledge functions.
  • 49 shows the difference between a word search comment and a document search comment
  • Fig. 52 is a comparison of the importance of each of the annotation devices.
  • Fig. 53 is a block diagram created around the search commenter
  • 54 is a flowchart of a search comment.
  • 55 is a flowchart illustrating annotations on indexes of search result words.
  • 56 is a diagram illustrating the structure of a searcher
  • 57 is a search query word
  • 59 is a flow chart of query term based on semantic unit terms.
  • 61 is a flowchart for searching for words and displaying items in word units.
  • 62 is a search flow chart listing and displaying word search results by word for each document.
  • 63 is a flowchart for generating and utilizing search knowledge.
  • 64 is a diagram illustrating the construction of a document information system builder
  • 65 is a natural language document information system and a unique ID + document information system.
  • 66 is to construct a document information system using dictionary, index and annotation knowledge.
  • 67 is to build a semantic unit term-based document information system using a dictionary and an index
  • 68 is a document information system construction using dictionary and annotation knowledge.
  • 69 is a flowchart illustrating the construction of a document information system using an index.
  • 70 is a flow chart of document information system construction using annotation knowledge.
  • 71 is a flow chart of document information system using search system index and annotation knowledge.
  • 73 is a flow chart of storing and using after merging a search target document source with additional information
  • 02-01 is a natural language document information system
  • 02-02 is a natural language based search system
  • 02-03 is a semantic unit term based device 1
  • 02-04 is a first-level semantic unit term-based information system
  • 02-05 is a semantic unit term based search system
  • 02-06 is a semantic unit term based document information system builder
  • 02-07 is a semantic unit term based information system
  • 03-01 is the semantic unit term-based document creation step
  • 03-03 is a semantic unit term-based index step
  • 03-04 is a semantic unit term-based index step
  • 03-07 is an annotation knowledge generation step
  • 03-08 is a knowledge base annotation step
  • 03-09 is the establishment stage of document information system based on semantic unit terminology
  • 05-02 is the semantic unit term generation in the word search process
  • 06-01 is the table above and shows that natural language has various meanings.
  • 06-02 shows the following table and shows that unique IDs are assigned to various meanings of natural language.
  • 09-01 is the semantic unit term information acquisition step
  • 11-02 is the semantic unit term search classification stage
  • 11-03 is the semantic unit term classification stratification step.
  • 11-05 is a step to adjust the semantic unit term classification dissent.
  • 16-02 is the search term using the term group
  • 20-01 is the step of determining the default value of semantic unit term by group
  • Step 20-04 applies the semantic unit term internet default
  • 23-02 is the step of receiving annotation knowledge creation request
  • 25-01 is the stage of receiving knowledge base annotation requests
  • 25-02 is annotated knowledge search phase
  • 26-01 is the request to perform index target annotation knowledge step
  • 26-02 is annotated knowledge transformation stage
  • 26-03 is annotated index search step
  • 32-01 is a manual annotated document writer
  • 33-01 is a natural language document creation step
  • 39-01 is the semantic unit term-based document collection stage
  • 39-02 is a semantic unit term-based index step
  • 39-03 is a semantic unit term-based search step
  • 40-01 is the semantic unit term-based document collection step
  • 40-02 is a semantic unit term-based index step
  • 40-04 is the semantic unit term based search step
  • 41-01 is the semantic unit term-based document collection
  • 41-02 is a semantic unit term-based index step
  • 41-04 is the semantic unit term based search step
  • 43-01 is a conceptual index for 43-02 documents
  • 44-01 is a conceptual index for 44-02 documents
  • 44-02 is a search for natural language instructions
  • 44-03 is an arrow pointing to a table showing the values to be placed in the semantic unit term field of the index.
  • 45-02 is the semantic unit term-based indexing stage
  • 46-01 is the semantic unit term query comment section (in the semantic term based search commenter).
  • 46-02 is the semantic unit term query part (which is included in the semantic term based searcher).
  • 49-01 shows a word search annotation method that annotates all words in a document.
  • 49-02 does not record the location within a document, so it shows a document retrieval annotation that records only one thing with the same natural language and the same meaning.
  • 51-01 is a new document document creator type 1
  • 51-03 is the default document commenter.
  • 54-02 is a document retrieval comment request receipt step
  • 55-01 is the semantic unit term based word search step
  • 57-01 is a natural search query
  • 57-02 is a unique ID + search query
  • 59-01 is the natural language query stage
  • 61-01 is the step of receiving a word search request
  • 62-02 is a word search result document-by-word display step
  • 63-01 is a search query review step
  • 63-02 is the creation of search knowledge
  • 63-03 is the stage of receiving a search knowledge disclosure request
  • 65-01 is a natural language document information system
  • 65-02 is a unique ID + document information system
  • 70-01 is the document information system document collection stage
  • 70-02 is the application of annotation knowledge documents
  • 70-03 is the application stage of annotation knowledge document information system.
  • 71-01 is the document information system document collection stage
  • 71-05 shows the application of annotation knowledge documents.
  • 71-06 shows the application of annotation knowledge document information system.
  • Semantic term-In the natural language the same natural language expression may have several meanings. On the contrary, a single meaning may be expressed in various ways. A semantic unit term generates one term for each meaning. When a natural language expression has various meanings, the term is subdivided by a semantic serial number. On the contrary, when the expressions have various expressions, natural language representative expressions are used to have the same meaning. However, as an exception, even if the meaning is the same, if the languages are different, separate semantic unit terms are created.
  • the natural language exists in the form of "natural language + meaning unit term" with the semantic unit terms are annotated to clarify the meaning.
  • the semantic unit term is used in two meanings. It may mean “natural language + meaning unit term”, and it may mean only “mean unit term” regardless of the natural language.
  • the term "natural language + semantic term” means a semantic term.
  • a semantic unit term document means a form in which a semantic unit term is annotated in a natural language.
  • the semantic unit term index is also an index containing both natural language information and semantic unit terms. The terms used to clarify this meaning are Unique ID and Unique ID +.
  • Unique ID-A representative semantic unit term proposed by the present invention It is made by linking a semantic serial number to a natural language representation. One language is created for each language.
  • Annotation-Annotation is used here to clarify meaning by adding semantic unit terms to natural language expressions.
  • convert means to convert a natural language expression into (natural language expression, semantic term) pair. After all, comments and conversions mean the same thing.
  • GUID-Globally Unique Identifier is a pseudo-random number used in application software. While there is no guarantee that a unique value will always be created when generating a GUID, it is very unlikely that the same number will be generated twice if there is an appropriate algorithm. Therefore, the system does not need to maintain serial numbers. However, its length is inconvenient to use.
  • the present invention is a semantic unit term based information system centered on a retrieval system.
  • the basic components of the natural language retrieval system are document collectors, indexers, and searchers.
  • Natural language document writers and natural language search systems use natural language dictionaries.
  • the natural language information system centered on the retrieval system consists of 5 devices: 1) dictionary, 2) document writer, 3) collector, 4) indexer, and 5) searcher.
  • the semantic unit term-based information system includes all the devices of the natural language information system.
  • the basic framework is the same. Devices added because they are semantic terms are 1) dictionary of semantic terms, 2) commenter of meaning unit, 3) search commenter of meaning unit, and 4) builder of document information system based on semantic unit.
  • the actual diagram consists of eight devices except the natural language dictionary. This is because the natural language dictionary is conceptually included in the semantic unit term dictionary. Of the four devices that have been added, the need for a semantic dictionary of terms is too obvious.
  • the other three devices (commenters, search commentators, and document information system builders) are the devices needed to convert information made from natural language into information made from semantic unit terms.
  • semantic unit terminology is not the language that users use in real life, but the number of words is much longer and its length is longer. Therefore, we need special help because we can't remember and write the document. There is a need for devices that help users easily use semantic terms.
  • Annotator is a device that converts natural language into semantic unit term.
  • a document writer, retrieval system, and document information system builder are used as an independent device outside the retrieval system.
  • Search commentators are internal devices in the search system that convert the contents of an index from natural to semantic terms.
  • the document information system builder is a device that converts all documents into semantic unit terms based on knowledge and information accumulated in the state of making semantic unit terms based on the retrieval system.
  • 1 is a configuration diagram including all devices.
  • a semantic unit term based information system is a semantic unit term based information system including dictionary manager, commenter, document writer, retrieval system, and document information system builder.
  • the semantic term dictionary manager is a device that creates semantic unit terms and adds descriptions to them to create dictionaries and manage them. It is a basic device used by all devices of A. semantic term-based information system. Abbreviation is a dictionary manager.
  • Meaning unit term generation unit is a device that generates a dictionary by generating a unique ID, a meaning expression ID, or a semantic based GUID that is a semantic unit term, and adds a description to it.
  • Abbreviation is term generating unit.
  • the semantic unit term management unit is a device that manages the modified semantic unit term.
  • dictionary search unit is a dictionary finder. When a user searches a dictionary by inputting natural language, corresponding semantic unit terms are listed and the user selects one of them. It is similar to the function of inputting Hangul and converting to Hanja, but Hanja conversion is replaced with Hanja, but the dictionary search unit is commented after natural language rather than replacing. Abbreviated name is dictionary search
  • C. Meaning unit term commenter is a device that annotates semantic unit term in natural language expression and is used by D. Meaning unit based document writer, E. Meaning unit based search system, and J. Meaning unit based document information system builder. do. Abbreviation is a tin group. It is very difficult to convert all natural words into semantic units using dictionaries.
  • Annotators are devices that automatically comment or help using annotation knowledge or defaults. It is a device used for comments on natural language in documents, comments on search system indexes, and comments on search query words. It is used to annotate existing documents as well as to comment on natural languages as new documents are created. It can be done by command, or it can be done automatically on a regular basis like an agent. It is also used for annotating bulk documents and for individual documentation.
  • Annotation knowledge is the knowledge that "in any 1) condition, 2) natural language expression is 3) meaning.” This is usually done using a search commenter that finds objects by searching and annotates specific semantic unit terms in certain natural language expressions, and then registers them as annotation knowledge if the results are satisfactory.
  • 1) condition is query term used in search
  • 2) natural language expression is specific natural language expression used in search
  • 3) meaning is semantic unit term used to comment in search.
  • Default management unit is a device that manages default values.
  • the default value is the semantic unit term for a specific natural language most frequently used on the individual, in a particular organization, in a particular field or on the Internet. In situations where multiple default values are applied, they usually have priority in order of individual, specific group, sector, and the Internet, and the user can specify the priority or default value. If there is no comment knowledge and a specific natural language cannot be annotated as a semantic unit term, the default value of the highest priority is applied.
  • C3.Knowledge-based comment section (document / index / query comment) is usually marked as C3.Knowledge-based comment section. It is a device that annotates or helps semantic unit terms in natural language by using comment knowledge or default value. It can be called and used, or it can be run regularly like an agent. It is a device that can be used for all annotations, including bulk documentation.
  • Index-based document commenting unit is an apparatus that annotates the contents of the document by extracting the information of the index while the index is converted based on semantic terms.
  • the fact that the index has already been based on semantic unit terminology means that the semantic unit term-based information system is completed.
  • C5.Annotation management unit is a device that shows all the comments and reviews the contents so that the comment errors can be corrected. Comments added by the user's comment knowledge, comments added by the user's search comment, etc. can be viewed in the order of the comment date.
  • the term-based document composer can write a document in semantic unit terms, but conceptually, it creates a document in natural language and finds the corresponding semantic unit term using natural language, and then goes through the two-step process of commenting on the natural language. Create a term-based document.
  • the abbreviation is a document writer.
  • Natural language writing unit is the same as the existing natural language-based document generator.
  • Meaning unit term document comment section is a device that annotates documents written in natural language in semantic unit terminology. Annotation is a difficult task only with the help of the semantic unit terminology dictionary, but it can be done without any difficulty with the help of C3.
  • the semantic term based search system is a device for indexing and searching the collected documents based on semantic unit terms.
  • Internal devices include 1) document collector, 2) indexer, 3) search commenter, and 4) searcher.
  • F. Document Collector is a device that collects documents to be searched.
  • Semantic term-based indexer is a device for creating semantic term-based indexes from retrieved documents. Abbreviation is indexer
  • Semantic term-based search commenter is a device that combines search and annotation functions to annotate indexes.
  • Abbreviation is a search commenter
  • Document search comment section is a device that annotates a specific semantic unit term to a specific natural language contained in all or part of the documents found by a search.
  • the abbreviation is a document search comment.
  • Word search comment section is a device that writes and annotates a specific semantic unit term for all or part of the words found by the search. Abbreviations are word search comments.
  • a semantic unit term based searcher is a searcher that searches a query made of semantic unit terms for an index created based on a semantic unit term.
  • Abbreviation is a searcher
  • the document search unit is a list of documents whose search results are the same as in the existing search system. For example, if a word search result is 4 words in 2 documents, 2 items are listed. The resulting items may be subject to document processing.
  • the word search unit is a word list of the search results. For example, if the word search result is 4 words in 2 documents of 2 documents, 4 items are listed. Result items may be subject to word processing.
  • Search Knowledge Management Unit is a device that creates and manages search knowledge. If the user determines that the search query is meaningful, the user may register it as search knowledge. Existing natural language search was so low in accuracy that it was less likely to continue to be used as knowledge. On the other hand, the semantic unit term-based search can pursue 100% accuracy. The knowledge of low accuracy rate increases the error rate by operation, but the semantic unit term base can be used in combination.
  • Semantic term-based document information system builder extracts semantic unit term information from semantic unit term-based index to make documents in document information system based on semantic unit term or convert documents to semantic unit term using annotation knowledge.
  • Device. Abbreviation is document information system builder.
  • Index-based document information system construction unit is an apparatus that makes the documents in the document information system based on semantic unit terms using index information.
  • J2.Annotation knowledge-based document information system construction unit is a device that makes the documents in the document information system based on semantic unit terms using annotation knowledge.
  • the proposed method is to sort the document information system by words and to annotate the whole word. Fortunately, there are devices that sort by word. This is a search system. In the retrieval system, the contents of all searched documents are sorted by words. The proposed method is to make the index of the retrieval system based on the semantic term instead of changing the document information system to the semantic term. Changing the index to semantic terminology is the same as changing the document information system to semantic terminology.
  • a semantic unit term-based index can make a natural language document information system into a semantic unit term-based document information system.
  • the proposed first-level semantic unit term-based information system (02-04) is made by introducing a search system (02-02) into the natural language document information system (02-01). After sorting by word by the index of the retrieval system, the index should be changed based on semantic terms.
  • the semantic unit term-based device 1 is a semantic unit term dictionary, a default DB and an annotation knowledge DB and three devices (a semantic unit term dictionary manager, a semantic unit term commenter, and a semantic unit term based search annotator). These devices make natural language-based indexes into semantic unit-based indexes.
  • the purpose of the first-level semantic unit term-based information system is to build a semantic unit terminology dictionary and a semantic unit term-based index. It can be said that the semantic unit term-based index and dictionary were completed in the first stage, but it is still based on natural language in terms of document information system and retrieval system. Also, step 1 has no role in terms of new documents, rather than existing documents indexed in the retrieval system.
  • the semantic unit term-based processing apparatuses for new documents are added, the search apparatuses are changed to the semantic unit term based, and the semantic unit term based retrieval system apparatus is changed to change the document information system to the semantic unit term based (02-05). ) And semantic unit term-based document information system builder (02-06) are added.
  • the core devices of the semantic unit term-based information system are contained in the first-level information system. If the first stage succeeds, there is no obstacle to the completion of the semantic unit term-based information system. This is because the second stage is not a task performed by a large number of users, but a task performed by the operator / developer and the user simply uses the result.
  • FIG. 3 is a flowchart in which a semantic unit term based information system centered on a search system operates.
  • the first four steps (document creation step (03-01), document collection step (03-02), indexing step (03-03) and search step (03-04)) are typical features of the search system. It is not based on semantic unit terminology. If documents are written as semantic terminology documents from the beginning, they can be treated the same as natural language-based information systems, and no special procedures need to be introduced. However, since the dictionary of semantic unit terminology is insufficient at the beginning, it is difficult for the document itself to be written based on semantic unit terminology. Almost all of them are collected and indexed as natural language documents, and the actual step of making indexes based on semantic terms is the next step. From now on, the semantic terminology-based procedure begins.
  • semantic unit term dictionary manager creates a term using natural language expressions and creates a dictionary entry for the term by pairing the generated term with a description (03-05).
  • the user must divide specific natural words by meaning and display the meaning unit by indexes sorted by words.
  • the user searches using a query to find a specific meaning of a specific natural language and annotates the semantic unit term in the index to the corresponding natural language expression included in the found document (03-06).
  • Conventional natural language indexes index document positions and document names in natural language fields, whereas semantic term-based indexes index document positions and document names in natural language / semi units.
  • the work of creating indexes on a semantic basis based on search annotations can be done. But a more sophisticated approach can be applied here. Rather than performing the search knowledge once and forgetting it, storing this information can be used for other purposes.
  • the most representative example is the application to new documents.
  • the search system index adds content as new documents continue to be added. It is inconvenient for the user to regularly perform search knowledge on newly indexed documents.
  • the search query word used in the search comment, the natural language expression to be commented, and the semantic unit term to be commented out become comment knowledge when stored.
  • annotation knowledge is later performed to perform the same tasks as existing search annotations.
  • Annotation knowledge is usually done on a different target than previous search annotations. New documents that are newly created and included in the search system index can be performed regularly. Annotation knowledge can be performed in the form of an agent by setting time and period (03-08). Repeated search annotations and knowledge base annotations build up many semantic term annotations in the index.
  • semantic unit term annotation information for each document from the semantic unit term-based index and applying it to the corresponding document, make the document into semantic unit term-based document, and make the document information system based on the semantic unit term.
  • Document information systems can be based on semantic units (03-09).
  • semantic unit term dictionary is completed, semantic unit term based index is completed, and semantic unit term based document information system is completed.
  • FIG. 4 is a diagram illustrating the configuration of a pre-manager.
  • Meaning unit term generation unit is implemented by selecting one of four methods, unique ID, semantic expression ID, semantic unit GUID, and semantic expression GUID, but it does not mean that several methods are applied at the same time.
  • the term merging is used for merging one of the two or merging by making a third term when two semantic unit terms have the same meaning.
  • Terminology classification is the same as classifying Obama as "man,” president. Classifications do not have to be entered at term generation and can specify multiple values.
  • Terminology aliases can be created for semantic terminology that is used frequently. Long semantic unit terms are term aliases because they are inconvenient for users to enter and difficult to remember. This term alias is translated into the corresponding semantic unit term before being used by the actual device.
  • division function is a function for dividing, dividing and searching a term in detail when a term is frequently used.
  • semantic unit terms there are only a few cases and hundreds of millions of cases. If hundreds of millions of cases are found, the terminology split will be used.
  • a term group is a group of several terms, and the group search shows the combined results of each of the terms in the group.
  • Meaning unit term dictionary search unit is a dictionary finder. When a user searches a dictionary by inputting natural language, corresponding semantic unit terms are listed and one of them is selected. It is similar to the function of inputting Hangul and converting to Hanja, but Hanja conversion is replaced with Hanja, but the dictionary search unit is commented after natural language rather than replacing.
  • the retrieval system is the best system to show related information easily while making sense of semantic terminology.
  • the retrieval system makes it easy to generate semantic terminology and to create means for annotating semantic unit terms in the index.
  • the retrieval system is the best tool for transforming natural language based information system into semantic based.
  • FIG. 6 shows how ambiguous a natural language is and why a semantic unit term is necessary.
  • the upper part of FIG. 6 shows the cause of the invention (06-01).
  • Natural language has many meanings. This causes the general search engines to have a low accuracy rate based on semantic unit terms. In the case of Hong Gil-dong (a pseudonym, the inventor's name), the accuracy rate is 1/641. A myriad of proper nouns invade common nouns and verb adjectives, making the meaning of words unclear.
  • the lower part of FIG. 6 shows that a semantic unit term is generated for each meaning of the natural language expression (06-02).
  • the unique ID is a representative semantic unit term used in the present invention and is made by adding a natural language representative expression and a semantic serial number. Unique ID is created separately for each meaning. Looking for Hong Gil-dong on a particular social network service (SNS), there are 641 people with the same name. In Hong Gil-dong_1, 1 is the semantic serial number. After that, if a new Honggil-dong is found, it will be Honggil-dong_642 using the largest meaning serial number. If the semantic unit term is used instead of the natural language, it is 100% at the search accuracy rate of 1/641 in Hong Gil-dong.
  • SNS social network service
  • Unique ID + is a concept that includes a natural language expression for the user in addition to the unique ID for clear expression.
  • the unique ID table contains a representative expression and a unique ID value, and contains a one-line description and a detailed description of the meaning of the unique ID.
  • the one-line description is used when many unique IDs are listed at the same time, and the description is used when there is enough space to see only one unique ID.
  • natural language and unique ID are one-to-many relationship, but there can be many expressions for one entity. In this case, other expressions that are not representative expressions are entered in other natural language expressions.
  • semantic unit term 9 is a flowchart illustrating generating a semantic unit term.
  • the generation of semantic unit terms is all parts of speech in all languages of the world. The number is at least 10 billion because all proper nouns, including personal names, are included.
  • semantic unit term with the desired meaning in the dictionary search and the natural language expression is not the same, but it is included in other expressions, the semantic unit term can be used. When it is necessary to create a semantic unit term, there is no semantic unit term with a desired meaning.
  • a natural language expression and a description of a specific meaning of the natural language expression must be input (09-01).
  • a new semantic unit term is generated by connecting the semantic serial number of the natural language expression to the input natural language expression.
  • a unique ID that is a semantic unit term defined in the present invention is generated (09-02).
  • the semantic unit term dictionary item is generated by pairing the generated semantic unit term and the obtained description (09-03).
  • semantic unit terms are embodiments of the present invention. These terms are very easy to define and very easy to use compared to ontology dictionaries, which can be called conventional semantic dictionaries. Therefore, general users who do not have expertise can participate in generating semantic unit terms of interest and build new document information system using these terms. For example, if the natural language AAA has three meanings, the effort to create three terms AAA_1, AAA_2, and AAA_3 to create a unique ID, and write a description for each one is completed. .
  • the four semantic unit terms may have different shapes, but basically, the knowledge required by the user or the information to be input is similar. Because it is created in the natural language system, it does not require the effort and knowledge to create a completely new language.
  • the semantic unit term generation method can be considered when there are only two natural languages in the world and each has two meanings.
  • AAA_1, AAA_2, BBB_1, BBB_2 is a unique ID method and the system must maintain a semantic serial number for each natural language.
  • Unique ID is a method of maintaining and using serial numbers for each natural language. This is the best way to read and remember the user.
  • Unique ID is a representative semantic unit term proposed by the present invention. The process of dividing natural language expressions into semantic units is easy to understand. On the other hand, making various expressions as one semantic unit term may be a little inconvenient for general users because the concept of natural language representation should be introduced. For example, in many news, President Obama is represented as Barack Obama, but there are also cases where it is expressed as Barack Hussein Obama, Barack Hussein Obama II, Barack, and Obama. Creating a term for each of these expressions results in a semantic expression ID. Since the semantic expression ID is not a semantic unit term, it is necessary to merge the semantic unit into a semantic unit term.
  • semantic merge ID The merge of semantic expression ID into semantic unit is called semantic merge ID.
  • the semantic merge ID corresponds to a unique ID
  • the semantic expression ID corresponds to a unique ID +. Comparing the unique ID method and the semantic expression ID method, the semantic expression ID requires several times the term generation effort. It makes the term dictionary large and the user uncomfortable by writing explanations without the need for expression units rather than semantic units.
  • the unique ID + does not have a separate term description can confirm the efficiency of the unique ID method.
  • the unique ID is the most recommended semantic unit terminology in that the term generation effort is the smallest among the proposed semantic unit terms and is easy to remember and use.
  • the natural language representation is Barack Obama, while Barack Hussein Obama, Barack Hussein Obama II, Barack, and Obama are other expressions.
  • the generated unique ID becomes Barack_Obama_1 assuming that the semantic serial number of the corresponding natural language expression possessed by the system is 1.
  • the part enclosed in square brackets is a unique ID + and corresponds to a semantic expression ID.
  • Unique ID of the present invention has the following meaning. Unique ID was created to remove the ambiguity of natural language, and terms are created for each of the various meanings of natural language. It is the most representative semantic unit term, and it is divided into semantic units including all proper nouns such as names, place names, etc., which are confused with other words. The global set of 6 billion people, including all languages and all parts-of-speech, must be a separate, unique ID item, at least 10 billion.
  • Teen ID can easily create based on natural language, so general users can create and annotate terms. Unique ID is a precise language with a rich dictionary. The prerequisite for the new term to actually be established and empowered is to be able to annotate all existing documents with unique IDs. It is not worth it without the annotation method.
  • unique ID + will be the basis for search engines, language translation, semantic web, AI, and classification.
  • the unique ID maintains the generation method that depends on the natural language even when creating a concept that does not exist in the existing natural language expression. Create a natural language representation for the new concept and create a unique ID based on the generated natural language representation.
  • a detailed description of the semantic unit term is described using a unique ID. Since the implementation of semantic expression ID and semantic unit GUID is very different from the implementation of unique ID, a separate explanation is not necessary unless a separate explanation is necessary.
  • the classification of semantic unit terms means that the object of classification is a semantic unit term.
  • the semantic unit term is also used for the classification name to which the semantic unit term belongs.
  • Classification names can be natural, semantic, or mixed forms of natural and semantic terms.
  • the semantic unit term may have a classification name of 0 or more, and the classification name of the semantic unit term may be added or deleted at any time, and the classification name does not need to be defined before use in the term, and when the term is created or the term is changed. If you enter a classification name that has not existed before, a new classification name is automatically registered, and one classification name belongs to more than 0 classifications and hierarchies. If there is disagreement, the classification and hierarchical structure of terms can be refined through group intelligence such as discussion. Intuitive semantic term classification method.
  • classification name is entered in the classification field of a term while generating or changing a semantic unit term, the term belongs to the classification name (11-01).
  • the term classification can proceed in bulk through search.
  • the semantic unit term dictionary is searched and the selected terms belong to a specific classification (11-02).
  • Classification can have a hierarchical structure.
  • the hierarchical structure is created by selecting two classification names and setting up a hierarchical relationship. This hierarchical relationship setting has a complicated hierarchical structure when it is repeated (11-03).
  • This semantic unit term classification can be changed if a change such as an error is found (11-04).
  • the classification of semantic unit terms proceeds with the participation of many people as natural language develops. Procedures for setting up, discussing, and voting are provided so that the classification of semantic unit terms can be developed by many people (11-05).
  • FIG. 12 illustrates a method of using a semantic unit term term alias to create and use a term alias that can be used when a semantic unit term is long and difficult to remember.
  • Terms apply to semantic terminology, and term aliases are created and used for individuals, specific groups, or the Internet.
  • the term alias is created using three pieces of information: applied group, term alias, and semantic unit term (12-01). To use a terminology of a group, the group's term aliases are listed in the individual's terminology list (12-02).
  • the actual query term is executed or translated into the corresponding semantic unit term before the document is stored (12-03).
  • FIG. 14 illustrates a method of managing specific semantic unit terms by dividing them into term segments when necessary to subdivide semantic unit terms, and using them to annotate and search like semantic unit terms subdivided using semantic unit term terms. .
  • the semantic unit term term division is performed (14-01).
  • the terminology division may consist of several layers, not just one.
  • the terminology of the lower hierarchy can be created (14-02). Once a term split has been created, it can be used to annotate a document or search system index (14-03) and search for the term using the term split (14-04).
  • semantic unit terminology group If you define a term group, you can create a search query using the term group name. In the example shown in the figure, the search term “2010 Korea High School Grade 1 _Grp” shows the list of the results found with “Hong Gil Dong_1” and the results found with “Kim Gil Dong_1”. Semantic unit terminology
  • the term group unlike terminology, has no use for annotating documents or search system indexes, and semantic unit terms are more precise language than natural language. Thus, if you search in semantic terms, only a small number of documents can be searched in. Groups can be used to increase concepts or search results at a reasonable size. A list of graduates should be found and each one searched, and this term group function provides a convenient way to perform two-step tasks at once.
  • Figure 16 shows how to create a semantic unit term group and use it. After inputting a semantic unit term or group list to be grouped, a group name to be created, and a group description, and requesting to create a semantic unit terminology group, a term group is generated using the input items (16-01). The created term group can be used in search queries. The term group included in the search query is converted into a semantic unit term query and the search is performed (16-02). Natural language is not an object that can accumulate knowledge in search because its meaning is unclear. This is because the error is widened as it is used in various ways. Semantic unit terminology can be used in various ways because it is precise and close to 100% of search accuracy.
  • the semantic unit term-based information system includes a device for annotating document builders, retrieval systems, and document information system construction units.
  • the independent commenter is mainly described, and the overall commenter is comprehensively described in the section describing the search commenter in the search system.
  • Semantic unit term commenter provides comment function to all devices (document writer, retrieval system, system builder).
  • Mean unit term commenter is a device for annotating semantic unit term in natural language expression.It is C1.Annotation knowledge management unit, C2.Default management unit, C3.Knowledge-based annotation unit, C4.Index-based document annotation unit and C5.Annotation unit. It is composed and used with the semantic unit dictionary manager.
  • Commentators are called independent commenters, meaning they can be used without being dependent on a particular device.
  • the search commenter is a powerful commenting device, but is separate from this independent commenter because it depends on the searcher.
  • Independent commentators are called on different devices and used in a variety of ways.
  • Annotation knowledge consists of 1) comment conditions, 2) natural language expressions to be commented, and 3) semantic unit terms to be commented on.
  • C1.Annotation Knowledge Management Unit is responsible for creating, modifying and deleting this annotation knowledge.
  • Meaning unit default value management department creates and manages default value for individual or group.
  • the default value is the semantic unit term that a particular person or group uses the most for a particular natural language expression.
  • the individual's default is the highest priority, groups such as companies or sectors come first, and everyone's Internet has the lowest priority. Individuals using default values decide which default values to apply.
  • C3.Knowledge-based comment section is a device that annotates semantic unit terms in natural language expression using annotation knowledge and default value.
  • Knowledge base annotations are performed on documents, indexes and queries. That is, it is used for annotation in all parts of natural language input. It can be called from where natural language is input or used in the form of an agent that is executed regularly. It can be done in the form of automatic annotation.
  • annotation knowledge When annotation knowledge is accumulated enough, all annotations can be automatically performed.
  • Knowledge base annotations apply annotation knowledge and default values when run. Whether or not to accept defaults in the absence of annotation knowledge is determined by the configuration. The default value means the highest frequency of use and does not mean that the accuracy is above the standard.
  • Index-based document Annotation unit is a device that converts a document into semantic term based on information in the index.
  • the target document In order to use the information in the index, the target document must already be included in the search system index. If the document is based on semantic unit terminology, the relevant part of the index can be changed to semantic unit term base. Conversely, if the information in the index is based on semantic unit terminology, the document can be based on semantic unit terminology.
  • This device can be said to be a device for type conversion of existing information.
  • C5.Annotation management unit is a device that shows all the comments and reviews the contents so that the comment errors can be corrected. My comment manager can view comments added by the comment knowledge that you created, comments added by your search comment, etc. in the order of comment date.
  • the heading part is a natural language, and the content below the heading indicates various meanings of the natural language (various meanings mean unique IDs).
  • the colored unique ID is the default semantic unit term for the natural language of a specific person.
  • the default value specifies a specific value among several meanings of natural language.
  • the default value of natural language Hong-gil-dong is set to Hong-gil-dong_1 (inventor Hong-gil-dong), operation is set to operation_3 (operation), and eyes are set to eye_1 (Eye). If there is a setting to apply the default value, the system automatically annotates the unique ID value, which is the default value when the user enters the above natural language according to the contents of this default DB.
  • 19 is an example of default values corresponding to a specific user.
  • Their priorities are individuals> groups> the Internet. Usually, the default value for the entire Internet is the lowest priority, and smaller groups usually have higher priority. Therefore, the individual's default has the highest priority.
  • the number and priority of groups they belong to can be determined by each user or set by the system. If you set the document field in advance while creating the document, the default value of the field is applied. In general, higher priority has a default value for some natural language and lower priority has a default value for many natural language.
  • the Internet has defaults for all natural languages.
  • the final default value is that of the highest priority individual. In order for the lowest Internet default to be the final default, all other group defaults must not exist. In the picture above, in the case of natural language Hong Gil-dong, there are several default values, but the highest priority personal default value is the comprehensive default value. In the case of natural language operation, the default values of the group and the Internet exist. In the case of natural language eyes, only the Internet has a default value, which is the final default value.
  • each group records the frequency of use of semantic unit terms by natural language expression and sets the semantic unit term with the highest frequency of use as the semantic unit term default value of the natural language expression (20-01). If a person is known because a search query is being made or the owner of a document is specified, the semantic unit term for a specific natural language expression is applied as the person's default (20-02). If the default value does not exist and the group (field) of the document is specified in the application of personal default value, the semantic unit term for the natural language expression is applied as the default value of the group. Apply priority to groups (20-03). If the corresponding default value does not exist in the group default application step, the semantic unit term for the natural language expression is applied as the default value of the Internet (20-04).
  • 21 shows a conceptual structure of an annotation knowledge table.
  • the comment condition refers to the search query.
  • This commentary knowledge is explained as follows.
  • this annotation knowledge acts like a search annotation.
  • the search engine searches for “President Obama” and annotates the unique ID barack_obama_1 in the index for the found documents.
  • this annotation knowledge is performed on a document, it finds a "President Obama” in the document and converts the Obama to Obama: barack_obama_1.
  • the search query can contain not only natural language but also a lot of information used in advanced search such as unique ID +, target site, field, date range and so on.
  • the search is performed by acquiring a search query using phrases allowed by the search query grammar, such as a natural language / meaning term expression, an operator, a period, a site, a field, a category (23-01).
  • phrases allowed by the search query grammar such as a natural language / meaning term expression, an operator, a period, a site, a field, a category (23-01).
  • Annotated knowledge and annotated knowledge ID are created that contain the verified search query word, the natural language to be annotated, the semantic unit term to be commented, and annotated knowledge item is created by combining the annotated knowledge, annotated knowledge ID, and description (23-03). .
  • Annotation knowledge is information that is applied when knowledge base annotations are performed.
  • the default value is applied only if there is a setting to apply.
  • the default value is inaccurate information compared to annotation knowledge. Therefore, whether the knowledge base comment is left uncommented or the default is applied is determined by the configuration.
  • the order of application is annotation knowledge> personal default> group default> Internet default. If there is a higher priority semantic unit term, it is used in the semantic unit term annotation of the natural language. If not, the semantic unit term of the next rank is used. If the semantic unit term used in the annotation processing is not correct, the user must correct it.
  • FIG. 25 illustrates a process of annotating by performing a knowledge base annotation on a document or query word. Indexing works with the help of the search system, but in the case of documents or queries, the search system is not involved. Thus the procedure is very different.
  • First select a natural language expression to be commented and make a knowledge-based comment request (25-01).
  • annotation knowledge is typically generated from search system queries by default. Therefore, not all annotation knowledge can be used for annotation in natural language expressions.
  • the annotation knowledge is indicated by a function that checks whether it is applicable in the absence of a search system, so the applicability can be confirmed in advance. If the corresponding annotation knowledge is not one but multiple, which one is to be performed first is the annotation knowledge itself. In general, priority has priority because it is determined that a small number of results is accurate when a search is performed.
  • annotation knowledge is the annotation knowledge that performs the search comment and stores the content of the search comment.
  • annotation knowledge is a duplication of what you've done in previous search annotations. But search system indexes are always changing. Adding new documents is the biggest reason. It is very inconvenient for a person to perform a search annotation each time new documents are added, but if you save the contents at the time of the search annotation, it can be automatically performed regularly.
  • annotation knowledge you can modify some of the content of the previous annotation knowledge in order to change the length of time or reenactment.
  • a comment knowledge request for indexing is entered by inputting a comment knowledge ID and a change element (26-01).
  • the requested comment knowledge is modified to reflect the change elements before execution (26-02).
  • (26-03) Annotate the semantic unit terms included in the annotation knowledge (26-04).
  • FIG. 27 shows that an index-based document annotation unit annotates a document using only index information.
  • the search commenter or commenter accumulates semantic information in the index, while the index-based document commenter is used to extract information from the index and apply it to natural language documents. It is a device that works backwards with semantic unit term indexer.
  • knowledge base annotations are typically used, using annotation knowledge and default values.
  • Index-based document annotations use information accumulated in the index, not annotation knowledge.
  • semantic unit term-based index semantic unit term annotations are accumulated by search commenter or commenter. The information stored in the index may be more than what can be obtained from annotation knowledge.
  • index-based document annotation unit is called and used mainly by the index-based document information system builder. It can also be called and used by the document writer.
  • Fig. 28 shows a procedure for annotating semantic unit terms to specific natural language expressions in documents indexed to a retrieval system. Documents are included in the index, but they do not necessarily have semantic term comments for specific natural languages in the document. This figure shows the procedure for annotating semantic terminology to a specific natural language expression using all available information, such as information in the search system index, annotation knowledge, and default values. For documents included in the index, the richest and most accurate information is the annotation information from the index.
  • the semantic unit term annotation is extracted by extracting information on the natural language expression in the document from the index (28-01). If the information is not obtained from the index, the annotation knowledge DB is searched to find the annotation knowledge of the natural language expression, and the semantic unit term is annotated in the natural language expression (28-02). If there is no information corresponding to the annotation knowledge and the default value is set, the default semantic unit term for the natural language expression is applied (28-03).
  • semantic unit term unique ID +
  • the only way to build a system is to decompose it into individual units so that individuals can decompose the Internet and do as much work as they need. But even when working at the individual level, it should not be a way of unevenly burdening individuals.
  • an individual comments on the entire word of his or her document it is difficult to proceed normally. Many words are used in one document. It takes a lot of effort to process many words regardless of the total number of comments. In fact, the number of comments is not proportional to the effort of the individual, but is proportional to the number of unique IDs used.
  • the unique ID unit annotation method has 23,000,000 times higher productivity than the document unit annotation method.
  • the annotation requirements for the entire information system are constant. Therefore, tin productivity is the most important measure of new system construction.
  • Unique ID unit annotation is a key device that enables the construction of a new system. Normally this is generated by the agent and performed regularly for new documents.
  • 32 shows a manual annotation type document builder and an automatic annotation type document writer.
  • a document writer can basically create a semantic term-based document with only a semantic term dictionary manager. You can create a document in natural language and search the semantic unit term dictionary to select the desired semantic unit term by referring to the description of each semantic unit term. However, it is unlikely that a manual document writer will actually be used. This is because document authors are not parties to semantic confusion in natural language, and manual commenting is inconvenient (32-01). The document composer will become the semantic terminology comment in the form of automatic commenting from the time when sufficient comment knowledge is accumulated, and the document composer will review and partially revise the comment content. Documents written in natural language in the autocomment format are autocommented using annotation knowledge and default values. After automatic commenting, the document writer displays a dictionary description of the semantic terminology that was commented out, and which commentary knowledge or default value was commented on (32-02).
  • Annotation knowledge is not a device that helps annotate with only one word entered. Although the default value can suggest recommended semantic terms even when there is only one word, it is normal to start a comment after completing a natural language document because it prevents the use of highly accurate annotation knowledge (33-01).
  • FIG. 34 shows only a search system in FIG. 1 and simplifies other parts.
  • the J. semantic term-based document information system builder is a device that uses only the results of the retrieval system and is not related to the performance of the retrieval system.
  • the plot consists of all the annotation devices that populate the content of the semantic term-based index.
  • FIG. 36 is a block diagram of a semantic unit term-based search commenter added to a basic semantic based search system. Only the commenter is missing among the devices that help the comment. Except for the problem that the commenter cannot repeat the annotation knowledge, it can be said that it is completed from the point of view of the search system. If you do not repeat the contents of previous search annotations on new documents that are newly added to the index, such as agents, it may be inconvenient for people to repeatedly perform search annotations. Therefore, semantic unit term annotations may be incomplete. If these features are included in the search commenter itself, then the search system is complete. However, the absence of a structure that utilizes search annotation knowledge beyond the search system can be a major obstacle to creating a complete semantic term-based information system.
  • 39 illustrates a method of operating a semantic unit term based search system having only basic functions.
  • This method has only basic functions, and the semantic terminology information of the index is obtained from the semantic terminology based document. Other than this, it does not provide a means to add semantic terminology information of the index.
  • the search system collects documents included in the search target, and whether the collected documents sufficiently include semantic unit term information determines the semantic unit term base level of the search system (39-01). Index the collected documents against natural and semantic terms (39-02). Searching for natural words and semantic unit terms stored in the index using query terms including semantic unit terms and natural language expressions (39-03).
  • 40 illustrates a method of operating a semantic unit term based search system in which semantic unit term information is obtained from collected documents and search annotations.
  • the search system collects documents included in the search object (40-01). Index the collected documents against natural and semantic terminology (40-02). Receives a search annotation request along with a query to find an annotation object, a natural language expression to be commented, and a semantic unit term to be annotated, and annotates the semantic unit term on the search system index to the natural language expression included in the search result of the query. -03). Search the natural language and semantic unit terms stored in the index by query words including semantic unit terms and natural language expressions (40-04).
  • FIG. 41 illustrates a method of operating a semantic unit term-based retrieval system for obtaining semantic unit term information from collected documents and annotation knowledge.
  • the search system collects documents included in the search object (41-01). Index the collected documents against natural and semantic terms (41-02). Annotated semantic terms in natural language expressions are annotated using annotation knowledge that has information that certain natural language expressions have meaning under certain conditions (41-03). Search for natural words and semantic unit terms stored in the index using query terms including semantic unit terms and natural language expressions (41-04).
  • Fig. 42 is a configuration diagram created around the indexer. Parts other than the indexer are simplified.
  • the indexer is responsible for indexing the collected documents. Semantic term-based indexes have a semantic term field added to the index. Semantic term comments in a semantic term-based document are recorded in the added field. The search commenter also records the semantic terms in this field. If the indexer fails to fill this part, the search commenter or commenter fills this part to base the semantic unit term. If the natural language has only one meaning, it is not necessary to comment. Natural language itself can also play a role as a semantic unit term.
  • This figure is the index (43-01) of the second Hong Gil-dong of a specific document (43-02) found by searching for "Hong Gil-dong".
  • the unique ID + value is formed. After all, this index is the document location index for the unique ID + value.
  • the indexing device creates a search system index (45-01) with a semantic unit term field blank for each word included in the collected document. If a semantic unit term annotation is included in the word, the semantic unit term is recorded in the semantic unit term field of the word index item (45-02).
  • 46 shows all annotation devices belonging to various devices. In the previous section on semantic unit terminology commenters, the independent commenter section is described, but all commenter devices are described here. 46 is different from FIG. 1.
  • the semantic unit term query term comment unit is included in the search commenter 46-01 and the searcher 46-02.
  • search To search, a query term must be prepared, and the query term is also the target of semantic term term annotation. Because query words are very short sentences, they are less important in terms of comments. It is usually treated as part of the document comment. In the case of search commenters, comments are made after the search.
  • the search portion of the search commenter uses much of the same functionality as the searcher. Therefore, the query is used in the search commenter, and the query word in the search commenter is the target of semantic unit term annotation like the query word in the searcher.
  • Annotation devices often contain the word document.
  • a document should understand exactly what it means in many ways.
  • Documents are sometimes used to mean “document search comments.” The opposite concept is a "word search comment”.
  • Documents also mean the subject of comments. The opposite of what it means to comment on a document is the record of the index.
  • Semantic Unit Term Document in the comments section means that the document is annotated rather than an index.
  • Documents in the document retrieval comments are document-level records. The target of all search comments is the index.
  • FIG. 47 briefly describes annotation devices that form the basis of a semantic unit term-based information system as part of the description of FIG. 46.
  • semantic unit term-based information system making natural language information based on semantic unit term is the core task.
  • the function of adding semantic unit term to natural language is simply called annotation function.
  • Annotation targets are places where comments are made. It is divided into document comment, index comment, and search query comment (47-01).
  • the target document is already indexed to the retrieval system and indicates whether it is annotated using the functionality of the retrieval system or an annotation method that does not use the retrieval system. This means that new documents are not included in the index and are processed regardless of the search system (47-02).
  • the splitting of search annotations occurs because existing search results are listed as documents. An incomplete way to comment on what a word in a document means is what is meant by a document search comment. Word search comments are more precise (47-03).
  • the C4 index-based document annotation unit, the J1 index-based document information system building unit, and the J2 annotation knowledge-based document information system building unit are functions that are performed secondarily after the first-level semantic unit term-based information system is already completed. It is therefore of no early importance (47-04).
  • the document information system and the index can be easily based on the semantic unit terminology when one is based on the semantic unit term.
  • the first thing to be based on semantic unit terminology is index, not document information system. This is because the semantic unit term base of indexes is much easier.
  • D2. Semantic terminology Document annotations are not a secondary device, but are not of great importance initially, in that they are not devices that annotate indexes.
  • the semantic unit term query term comment is not important because the amount of comments is extremely small.
  • the C3 knowledge-based commentary, the H1. Document search commentary, and the H2. Word search commentary are the initial critical devices (47-05).
  • Index comment is applied to the word search comment method.
  • 49 shows a difference between a word search comment and a document search comment.
  • Word search comments are a way to record all occurrences and are natural. Annotate each word in the document. This is the correct comment. Record up to each occurrence of each word in the document. This method is difficult to apply to existing search systems. A new search device made for this processing is the word search section (49-01). Document retrieval comments are inaccurate and the original comment should be done at every word level, and the problem is caused by the inability to obtain the desired information because the search is not a specific word, but a device to find a specific document. It is an annotation method that may disappear in the long run. Compared to the tin method per generation, only one Hong-gil-dong and two seas are recorded. The position of words should not be recorded (49-02).
  • New and old documents have different processing environments. Since new documents are not included in the search system index, they cannot be processed for the index. New document comments annotate the document itself. Existing documents are commented on the index (51-03). Existing document annotations are annotated with the retrieval system and new document annotations are annotations that proceed regardless of the retrieval system.
  • the new Document Builder-2 writes directly to the search system index, but means that it has a built-in indexer, which is done without any intervention from the search system. Storing the results directly in the search system's index does not mean using the traditional document annotation method. In the case of document writer-1, the document writer creates a semantic unit term-based document, and the collector collects the semantic unit term-based index (51-01).
  • the document writer does not pass the semantic terminology to the collector and then directly indexes it (51-02).
  • the indexing method can be conveniently used in situations where it is difficult to store and keep the annotated documents separately. Normally, you cannot save a changed document to its original location unless you are the owner of the document. In this situation, the changed contents are stored directly in the index without storing the changed documents.
  • the information stored in the index can be used at any time to convert a natural language document into a document annotated with semantic terms.
  • Existing document commenters comment on the index with the documents included in the index.
  • New documents can also be commented using existing document commenters if they are included in the index without any semantic term annotation work until the document is written. This is because annotating with indexes is more efficient.
  • the document retrieval section is a forced part because the existing retrieval system has a structure for searching a document.
  • the word search feature is added, the document search comment is not a necessary device. This is because comments are added to certain words rather than added to the document.
  • 55 shows a procedure of annotating a specific semantic unit term in an index to a specific natural language expression for words found through a search.
  • This method specifies that a natural language representation of a location in a document is performed and is performed in a structure of searching for words unlike a conventional search function.
  • the words are searched by obtaining a query including natural and semantic unit terms (55-01).
  • a search annotation request is received together with a list of all the search results words or some selected words, a natural language expression to be annotated, and information about semantic unit terms to be annotated (55-02).
  • the corresponding semantic unit terms are commented on the search system index for the natural language expression, and the position in the document of the natural language expression is clearly recorded (55-03).
  • Semantic term-based searcher includes I1. Document search unit, I2. Word search unit and I3. Search knowledge management unit, and there is a natural language query unit for creating a search query and a semantic unit term query term comment unit. Search comments do not comment the document, but comment the found words. Therefore, to help the search commenter's role, the searcher has been enhanced with the ability to find words rather than documents. Compared to a document search for a document, a word search has been added to clarify which words within the found document are desired to be listed. In the existing natural language search, the search method was not called knowledge.
  • the semantic unit term-based search can be 100% accurate and can be registered as a search knowledge and used in combination.
  • Search knowledge is created by registering the experience of search as knowledge. Both the search commenter and the searcher need a search query, and the query is the target of the semantic term term annotation. Therefore, the searcher has a natural language query unit and a semantic unit term query term comment unit. In the representative diagram (FIG. 1), the query-related part is not exposed as a component.
  • 57 shows a search query.
  • Query terms are used in search systems and search commenters in search systems.
  • a natural language search query is composed of one or more natural words and various operators such as and / or, specific time periods, specific sites, specific classifications, etc. (57-01).
  • the unique ID + search query consists of one or more unique ID + and various operators such as and / or, a specific time period, a specific site, a specific classification, etc. (57-02).
  • 59 shows a method of creating a semantic unit term-based query word.
  • Semantic unit terminology is difficult to remember and use, so input natural language and convert it to semantic unit term by dictionary search. Similar to the existing query method, a natural language is obtained to prepare a query (59-01). A natural language expression to be annotated in the query is selected and a dictionary search request is made (59-02). Obtain the selected item from the list of semantic unit terms listed and annotate the natural language (59-03). For the query words annotated with the semantic unit term, the natural / mean unit pair is changed to the pure semantic unit term (59-04).
  • a retrieval system is a device for retrieving a document and thus lists the document items (60-01). This method of document listing makes it difficult to process certain words within a particular document. If the natural language in a document is always used in the same sense, it is not a big obstacle to commenting. In practice, document-level commenting is not a major obstacle because you can comment on the meaning of each specific natural language in a document. In particular, the accuracy of the initial semantic unit term-based retrieval system is not a big obstacle. In general, since the natural language retrieval rate is very low and shows a superior accuracy rate, it is not a big problem to reduce the accuracy rate slightly based on the semantic unit term.
  • Word item listing eliminates the problem of document-level comments. It can be clearly expressed as a semantic unit term of a natural language expression at a specific position in a specific document. (60-02) This is a feature that existing search systems should add. However, this can be inconvenient if you need to use the traditional document listing method.
  • the document / word item listing method combines the document listing method and the word listing method (60-03). Word commentary does not necessarily mean that only one word is processed. Search for “President Obama” to support President_1 comment on President and comment barack_obama_1 on Obama.
  • the number of search result items is the same as the number of words searched for, and can be used for word-by-word processing.
  • the word search query can find the words you want, display the results in word units, and the number of items listed is the same as the number of words searched.
  • a search query for finding a document and a term (natural language expression or semantic unit terminology) information to be searched for in the searched document are received (61-01).
  • the words searched by the word search query are listed and displayed (61-02).
  • search procedure 62 shows a search procedure for searching for words and listing and displaying the results by word for each document.
  • the search results are organized by word by document, and the results can be used for document-by-document and word-by-word processing.
  • the search query finds the words you want within the desired document, displays the document as one item, and displays each word unit for each document.
  • the results are displayed in the same way as the number of items listed, plus the number of documents and terms.
  • a search query for finding a document and a document / word search request are received with information on a term (natural language expression or semantic unit term) to be searched for in the searched document (62-01).
  • the words searched by the word search query are listed and displayed by word of each document (62-02).
  • 63 shows a procedure of generating and utilizing a search knowledge.
  • Existing natural language search was so low in accuracy that it was less likely to continue to be used as knowledge.
  • the semantic unit term-based search can pursue 100% accuracy rate.
  • the knowledge of low accuracy rate increases the error rate by operation, but the semantic unit term base can be used in combination.
  • This procedure provides a means to perform search queries to review the results and to register and use meaningful search queries as search knowledge.
  • Perform and review the semantic unit term-based search query (63-01). Receives a search knowledge generation request along with a search query and its description, generates a search knowledge ID, and turns the knowledge search ID, search query and description into search knowledge (63-02). 63-03) Reveal search knowledge (63-04).
  • FIG. 64 is a diagram illustrating the construction of a document information system builder. Parts other than the document information system builder are simplified.
  • the document information system builder plays a role in building the document information system using information stored in the index or annotation knowledge.
  • 65 shows a natural language document information system and a unique ID + document information system.
  • the document information system is an entire document, including documents of various types such as Internet documents, companies, and personal documents.
  • the natural language document information system is a document information system based on the natural language dictionary (65-01), and the unique ID + document information system (65-02) is created based on the unique ID dictionary.
  • Creating a semantic term-based document information system is a huge task.
  • the value of changing the document information system is the same as the value of the index of the retrieval system that contains all of these documents based on semantic terms. Perfect commentary knowledge is of the highest value. This is because annotation knowledge has the added value of being able to base many parts of semantic terms on future documents. Annotation knowledge cannot be made right away. Making indexes based on semantic terms is the best way to base document information systems on semantic terms and is the best way to create annotation knowledge.
  • FIG. 66 illustrates the construction of a semantic unit term-based document information system using a semantic unit term dictionary, index, and annotation knowledge.
  • the semantic unit term dictionary is mandatory. Without this, neither the semantic term index nor the annotation knowledge can be created.
  • the semantic unit term index contains information about which natural language representation of a document is meant. Therefore, if the semantic unit term index has enough information, the semantic unit term document information system can be created.
  • Annotation knowledge is the knowledge that "under certain conditions, what natural language means what.” Therefore, if there is sufficient comment knowledge, semantic terminology document information system can be made.
  • 67 shows that a semantic unit term based document information system is constructed using a semantic unit term dictionary and an index. If the semantic terminology index has enough information, a semantic terminology document information system can be constructed. However, semantic terminology gives no information about newly created documents.
  • FIG. 68 illustrates the construction of a semantic unit term-based document information system using a semantic unit term dictionary and annotation knowledge. If there is sufficient annotation knowledge, it is possible to construct semantic unit term-based document information system using only annotation knowledge. Therefore, it is possible to construct a semantic unit term-based document information system without the help of a retrieval system. However, it requires more computing power than semantic-based using index information. In general, index information is larger than the semantic unit term information of annotation knowledge.
  • FIG. 69 illustrates a procedure for constructing a document information system such as the Internet based on a semantic unit term using a search system index in which information for annotating natural language expressions included in each document is accumulated.
  • the method of using index can be applied only to the documents included in the search target of the search system.
  • the semantic unit term annotation information accumulated in the index of the search system is classified by document location and the semantic unit term annotation information of each document is classified.
  • Each document collected by the retrieval system includes new semantic terminology annotation information for the document (69-02).
  • Documents created by including semantic unit terms are stored in a separate storage location of the retrieval system including the existing document location information (69-03).
  • 69 is a procedure of extracting information from a search system index and constructing a semantic unit term-based document information system.
  • FIG. 70 shows a procedure for constructing a document information system such as the Internet based on semantic unit terminology using annotation knowledge accumulated in annotating natural language expressions as semantic unit terminology.
  • Annotation knowledge can be applied without being dependent on a specific search system. Therefore, it is applicable to new documents of a specific search system.
  • Collect documents in the document information system It does not use a retrieval system and performs document collection directly (70-01).
  • the semantic unit term is annotated for all natural language expressions in the document. ).
  • Document information such as the Internet
  • search system index for documents that are included in the search system and having sufficient semantic unit term information accumulated in the index, and the annotation knowledge for new documents or documents outside the search system that do not have information in the index. It is a procedure to build a system based on semantic unit terminology.
  • the semantic unit term annotation information accumulated in the index of the search system is classified for each document position for the documents included in the search system to generate semantic unit term annotation information for each document (71-02).
  • Each document included in the retrieval system contains new semantic terminology annotation information for that document (71-03).
  • Documents created by including semantic unit terms are stored in a separate storage location of the retrieval system including the existing document location information (71-04).
  • the corresponding annotation knowledge is searched for the natural language expression contained within each document, and the applied annotation knowledge is applied to the corresponding natural language expression. Comment on the semantic unit term (71-05). After commenting is completed for each document, repeating the steps of storing the existing document location information in a separate storage location makes the semantic unit term-based document for all documents not included in the search system (71-06).
  • FIG. 72 is a flowchart illustrating a procedure for managing disagreements about the contents of a semantic unit term dictionary item, comment contents, annotation knowledge, default value, and search knowledge by using collective intelligence.
  • a user with disagreement about the semantic unit term dictionary entry's content, comment content, comment knowledge, default value, and search knowledge requests a discussion creation along with the discussion topic to create a discussion item on the topic (72-01).
  • 73 is a view illustrating a storing and using procedure after merging a search target document original with additional information. It is a method of storing and using the changed document contents in the situation where the contents of the search target document of the search system need to be supplemented or changed and the original document cannot be directly modified.
  • the target document is stored in a separate place along with the document address (73-01).
  • Change documents stored in separate places (73-02). Upon receiving a request for change to the address of the original document, the changed document is found and provided using the stored original document address (73-03).

Abstract

La présente invention concerne la modification d'un système d'informations comprenant des expressions en langage naturel en un système d'informations basé sur des expressions de sens individuelles, qui s'accompagne de modifications fonctionnelles d'un système de recherche d'informations, d'un dictionnaire des termes, d'un dispositif de production de documents et d'un convertisseur de termes. La précision des systèmes de recherche actuels est très faible. Cela s'explique par le fait qu'un langage naturel exprime de nombreuses significations avec peu de mots. Les expressions devenant plus longues et plus difficiles à mémoriser à mesure que le nombre de termes augmente, les gens utilisent moins de termes mais ils les emploient d'une manière répétitive. Lorsqu'on introduit des expressions de sens individuelles dans lesquelles un terme correspond à un sens, la précision d'un système de recherche peut approcher 100 %. La présente invention concerne également un procédé permettant de produire facilement des expressions de sens individuelles et un procédé permettant d'appliquer efficacement les expressions de sens individuelles produites à des documents provenant du monde entier. Le procédé de création d'expressions de sens individuelles est en fait une technique qui consiste à décomposer chaque terme d'un langage naturel jusqu'à obtenir le nombre de ses significations respectives. Puisqu'il s'agit d'une simple décomposition des termes, n'importe qui peut produire des expressions. La tâche consistant à appliquer des termes produits à des documents provenant du monde entier est considérable. Pour cette tâche, d'après la présente invention, au lieu de modifier chaque mot qui est utilisé répétitivement, un alignement est exécuté pour chaque mot et certains groupes de mots alignés sont traités simultanément. Même si un mot a été utilisé plusieurs centaines de milliards de fois dans des documents du monde entier, il n'est pas nécessaire de procéder plusieurs centaines de milliards de fois à des conversions des termes. Si le mot en question a plusieurs sens, la tâche de conversion peut être exécutée simplement au moyen de plusieurs commandes de tri. Même si l'utilisation répétitive de termes ne représente pas une grande charge lors d'une conversion des termes, puisque le nombre d'expressions de sens individuelles est en lui-même gigantesque, la conversion des termes est difficile. La tâche consistant à traiter près de 10 milliards d'expressions de sens individuelles est immense. Un procédé permettant de résoudre cette difficulté consiste à répartir uniformément la tâche sur un certain nombre d'utilisateurs. Le principal facteur d'ambiguïté d'un langage naturel réside dans la présence d'un nombre incalculable de noms propres. Ils ont une influence sur les domaines des noms, des adjectifs, des verbes et de toutes les autres parties du discours, ce qui provoque une confusion sémantique. Dans la mesure où la population mondiale compte plus de 6 milliards d'individus, les noms propres qui se rapportent à des personnes, et ils ne sont pas limités à cet usage, représentent à eux seuls plus de 10 milliards de termes. La présente invention concerne une configuration dans laquelle cette immense tâche est confiée de manière uniforme à un nombre incalculable d'utilisateurs. Pour répondre à leurs besoins, les utilisateurs peuvent exécuter des tâches leur permettant de répondre à leurs exigences et de bénéficier des fruits de leur travail. Si les utilisateurs considèrent qu'une conversion de termes est nécessaire, ils peuvent effectuer des tâches de production et de conversion de termes de manière à pouvoir entretenir en permanence un état satisfaisant pour les utilisateurs. La présente invention concerne : 1) un gestionnaire de dictionnaire d'expressions de sens individuelles qui peut facilement produire des expressions de sens individuelles ; et 2) un annotateur de recherche, autrement dit un moyen permettant de catégoriser et de convertir (annoter) des mots appartenant à un groupe de mots dans des expressions de sens individuelles. L'annotateur fonctionne comme une partie d'un système de recherche. L'alignement et la recherche de mots utilisent des fonctions existantes du système de recherche. La présente invention concerne en outre 3) un convertisseur (annotateur) des expressions de sens individuelles mettant en œuvre une fonction similaire à celle de l'annotateur de recherche. La tâche consistant à construire un système d'informations d'envergure mondiale basé sur des expressions de sens individuelles est une entreprise de grande ampleur. Pourtant, le manque de clarté des significations dans le langage naturel constitue un obstacle majeur au développement de nombreux domaines. La présente invention propose une base permettant de réaliser des progrès considérables dans les domaines du Web sémantique, des systèmes de recherche, de la traduction des langues et de l'intelligence artificielle en les faisant bénéficier d'un langage clair.
PCT/KR2011/004113 2010-06-07 2011-06-06 Procédé de production dynamique de termes supplémentaires pour chaque sens de chaque expression en langage naturel ; gestionnaire de dictionnaire, dispositif de production de documents, annotateur de termes, système de recherche et dispositif de construction d'un système d'informations sur des documents basé sur le procédé WO2011155736A2 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020100053571A KR20110133909A (ko) 2010-06-07 2010-06-07 모든 자연어 표현의 각각의 의미마다 별도의 용어를 동적으로 생성하는 방법 및 이를 기반으로 하는 사전 관리기,문서작성기, 용어 주석기, 검색 시스템 및 문서정보체계 구축장치
KR10-2010-0053571 2010-06-07

Publications (3)

Publication Number Publication Date
WO2011155736A2 true WO2011155736A2 (fr) 2011-12-15
WO2011155736A3 WO2011155736A3 (fr) 2012-04-19
WO2011155736A9 WO2011155736A9 (fr) 2012-06-21

Family

ID=45098509

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2011/004113 WO2011155736A2 (fr) 2010-06-07 2011-06-06 Procédé de production dynamique de termes supplémentaires pour chaque sens de chaque expression en langage naturel ; gestionnaire de dictionnaire, dispositif de production de documents, annotateur de termes, système de recherche et dispositif de construction d'un système d'informations sur des documents basé sur le procédé

Country Status (2)

Country Link
KR (1) KR20110133909A (fr)
WO (1) WO2011155736A2 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014104944A1 (fr) * 2012-12-27 2014-07-03 Abbyy Development Llc Procédé de marquage de dictionnaire
US9946762B2 (en) 2014-09-16 2018-04-17 International Business Machines Corporation Building a domain knowledge and term identity using crowd sourcing
CN111008594A (zh) * 2019-12-04 2020-04-14 科大讯飞股份有限公司 改错题评阅方法、相关设备及可读存储介质

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101602342B1 (ko) * 2014-07-10 2016-03-11 네이버 주식회사 의미 태깅된 자연어 질의의 의도에 부합하는 정보 추출 및 제공 방법 및 시스템
KR101661198B1 (ko) * 2014-07-10 2016-10-04 네이버 주식회사 단문／복문 구조의 자연어 질의에 대한 검색 및 정보 제공 방법 및 시스템
CN110059167A (zh) * 2019-01-23 2019-07-26 艾肯特公司 自然智能的自然表达处理方法、回应方法、设备及系统
CN117709375A (zh) * 2024-02-01 2024-03-15 成都帆点创想科技有限公司 文本翻译方法及装置

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR19980029011A (ko) * 1996-10-25 1998-07-15 구자홍 어휘 자동 분류 장치 및 방법
KR20020036059A (ko) * 2000-11-07 2002-05-16 옥철영 사전 뜻풀이말에서 추출된 의미정보를 기반으로하는 의미중의성 해결 방법
KR20070059881A (ko) * 2005-12-06 2007-06-12 한국전자통신연구원 의미공기정보 지식베이스 구축 방법 및 장치

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR19980029011A (ko) * 1996-10-25 1998-07-15 구자홍 어휘 자동 분류 장치 및 방법
KR20020036059A (ko) * 2000-11-07 2002-05-16 옥철영 사전 뜻풀이말에서 추출된 의미정보를 기반으로하는 의미중의성 해결 방법
KR20070059881A (ko) * 2005-12-06 2007-06-12 한국전자통신연구원 의미공기정보 지식베이스 구축 방법 및 장치

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2014104944A1 (fr) * 2012-12-27 2014-07-03 Abbyy Development Llc Procédé de marquage de dictionnaire
US9946762B2 (en) 2014-09-16 2018-04-17 International Business Machines Corporation Building a domain knowledge and term identity using crowd sourcing
CN111008594A (zh) * 2019-12-04 2020-04-14 科大讯飞股份有限公司 改错题评阅方法、相关设备及可读存储介质
CN111008594B (zh) * 2019-12-04 2024-04-19 科大讯飞股份有限公司 改错题评阅方法、相关设备及可读存储介质

Also Published As

Publication number Publication date
WO2011155736A3 (fr) 2012-04-19
KR20110133909A (ko) 2011-12-14
WO2011155736A9 (fr) 2012-06-21

Similar Documents

Publication Publication Date Title
WO2011155736A2 (fr) Procédé de production dynamique de termes supplémentaires pour chaque sens de chaque expression en langage naturel ; gestionnaire de dictionnaire, dispositif de production de documents, annotateur de termes, système de recherche et dispositif de construction d'un système d'informations sur des documents basé sur le procédé
WO2012074338A2 (fr) Procédé de traitement de langage naturel et de formule mathématique et dispositif associé
WO2010068068A2 (fr) Procédé de recherche d'informations et procédé de fourniture d'informations fondés sur les intentions de l'utilisateur
WO2020009297A1 (fr) Appareil et procédé d'amélioration des performances de compréhension d'un langage sur la base d'une extraction de domaine
WO2019177182A1 (fr) Appareil de recherche de contenu multimédia et procédé de recherche utilisant une analyse d'informations d'attributs
WO2011137724A1 (fr) Dispositif de conversation homme-machine en langage quasi naturel basé sur la logique sémantique
WO2010036012A2 (fr) Système de recherche d'opinion fondé sur internet, recherche d'opinion, système et procédé de service publicitaire associé
WO2012091360A2 (fr) Procédé et système de fourniture de contenu personnalisé par l'utilisateur
WO2011007935A1 (fr) Système et procédé de fourniture d'un service consolidé destiné à une page d'accueil
WO2018034426A1 (fr) Procédé de correction automatique d'erreurs dans un corpus balisé à l'aide de règles pdr de noyau
WO2017209564A1 (fr) Procédé de fourniture d'une liste d'applications et dispositif associé
WO2013168860A1 (fr) Procédé pour afficher un texte associé à un fichier audio, et dispositif électronique
WO2010021527A2 (fr) Système et procédé d'indexation d'objet dans une image
WO2013176365A1 (fr) Procédé et dispositif électronique permettant de rechercher facilement un enregistrement vocal
WO2013176366A1 (fr) Procédé et dispositif électronique pour une recherche facile durant un enregistrement vocal
WO2017146437A1 (fr) Dispositif électronique et son procédé de fonctionnement
WO2012130145A1 (fr) Procédé et dispositif d'acquisition et de recherche d'informations de connaissance pertinentes
WO2014010975A1 (fr) Appareil à interface utilisateur et procédé pour terminal utilisateur
WO2020197257A1 (fr) Procédé de traduction utilisant des éléments représentés visuellement, et dispositif associé
WO2012097701A1 (fr) Procédé, système et support de stockage informatique pour pré-lecture de données de réseau
WO2020082766A1 (fr) Procédé et appareil d'association pour un procédé d'entrée, dispositif et support d'informations lisible
WO2014010819A1 (fr) Procédé d'implémentation de données structurées et non structurées dans un document xml
WO2013119007A1 (fr) Appareil de création de données sur la base d'une roue d'idées
WO2021246642A1 (fr) Procédé de recommandation de police de caractères et dispositif destiné à le mettre en œuvre
WO2023080276A1 (fr) Système d'apprentissage profond distribué à liaison de base de données basé sur des interrogations, et procédé associé

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 11792647

Country of ref document: EP

Kind code of ref document: A2

NENP Non-entry into the national phase in:

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 11792647

Country of ref document: EP

Kind code of ref document: A2