EP1386255A1 - Procede permettant de trouver des reponses a des questions - Google Patents

Procede permettant de trouver des reponses a des questions

Info

Publication number
EP1386255A1
EP1386255A1 EP02708919A EP02708919A EP1386255A1 EP 1386255 A1 EP1386255 A1 EP 1386255A1 EP 02708919 A EP02708919 A EP 02708919A EP 02708919 A EP02708919 A EP 02708919A EP 1386255 A1 EP1386255 A1 EP 1386255A1
Authority
EP
European Patent Office
Prior art keywords
clause
question
natural language
conditions
syntactic
Prior art date
Legal status (The legal status 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 status listed.)
Ceased
Application number
EP02708919A
Other languages
German (de)
English (en)
Inventor
Eva Ingegerd Ejerhed
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
OPENAMPLIFY LIMITED
Original Assignee
Hapax Ltd
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
Priority claimed from SE0101127A external-priority patent/SE0101127D0/xx
Application filed by Hapax Ltd filed Critical Hapax Ltd
Publication of EP1386255A1 publication Critical patent/EP1386255A1/fr
Ceased legal-status Critical Current

Links

Definitions

  • the present invention relates to the field of information retrieval from unrestricted text in different languages. More specifically, the present invention relates to a method, and a corresponding system, for automatically finding answers to a natural language question in a natural language text database.
  • the performance of an information retrieval system is usually measured in terms of its recall and its precision.
  • recall has a standard definition as the ratio of the number of relevant documents retrieved for a given query over the total number of relevant documents for that query.
  • precision measures the exhaustiveness of the search results.
  • precision measures the quality of the search results. Due to the many documents retrieved when using the above type of search methods, it has been realized within the art that there is a need to reduce the number of retrieved documents to the most relevant ones.
  • Questions are a specific type of query.
  • systems range from delivering answers to simple questions to presenting complex results compiled from different sources of information. How well a question is answered is typically judged by human standards. Differently expressed, how would a well informed human being respond to a question with respect to correctness and exhaustiveness of the answer (if there is more than one answer) , with respect to the succinctness of the answer to the question posed, and with respect to delivering answers quickly.
  • a basic difficulty for question answering systems is that, as opposed to general information retrieval systems, the inquired fact is often very specific. Thus, the need for precision becomes even more acute.
  • Many prior art question answering systems suffer from being dependent on knowledge specific to a domain, to a line of business or a special trade. World knowledge optimal for one domain is of little value to another and thus hard to port . To update world knowledge for a domain specific question answering system automatically is not technically feasible and such systems do not scale well.
  • An object of the present invention is to provide an improved method, and a corresponding system, for automatically finding answers to a natural language question by means of a computer stored natural language text database, that are not subject to the foregoing disadvantages of existing methods for this task, i.e. that are not domain specific, that deliver answers to questions with high precision and that do not require complex methods of language analysis.
  • This object is achieved by a method and a system according to the accompanying claims.
  • the present invention is based on the insight that the relationship between the constituents and their respective syntactic functions in a question clause within a natural language question and the constituents and their respective syntactic functions in a clause that constitutes an answer to the natural language question can be used successfully in order to find answers to a natural language question in a natural language text database .
  • constituents refers to the basic units of text, such as word tokens, phrases etc. An important property of these units is that they can be found using finite state methods that recognize a strict hierarchy of constituents. Using finite state methods for syntactic analysis is well known within the art. However, the finite state method referred to here is a method of finding so-called initial clauses. Such a method is described in further detailed in the Swedish patent application SE 0002034-7 “METHOD FOR SEGMENTATION OF TEXT" and US patent application US 09/584 135 "METHOD FOR SEGMENTATION OF TEXT” .
  • Initial clauses have the property of being non-recursive, i.e. no initial clause includes another initial clause. Whenever the term clause is used in the following, it should be interpreted as initial clause .
  • a method for automatically finding an answer to a natural language question in a computer stored natural language text database.
  • the natural language text database has been analyzed with respect to syntactic functions of constituents, lexical meaning of word tokens, and clause boundaries, i.e. these are known to the system performing the method.
  • the natural language question comprises a question clause, which is the clause that conveys the content of the information need.
  • the method comprises an analysis step, where a computer readable representation of said question clause is analyzed with respect to the syntactic functions of its constituents and the lexical meaning of its word tokens.
  • a set of conditions for a clause in the natural language text database to constitute an answer to the question clause is defined.
  • the conditions relate to the syntactic functions of constituents and the lexical meaning of word tokens in the clause. Clauses that satisfy the conditions are identified in the natural language text database, and one or more answers to the question clause are returned by means of the identified clauses that satisfy said conditions.
  • the conditions that are defined according to the invention are based on the relationship between the constituents and their respective syntactic functions in a question clause and the constituents and their respective syntactic functions in a clause that answers the question clause. More specifically, one or more of the constituents in the question clause, or constituents that are equivalent in terms of lexical meaning, occur in a clause that answers the question, and the syntactic functions in the clause that answers the question of each of the constituents, or constituents that are equivalent in terms of lexical meaning, can be determined from the syntactic functions of the constituents of the question clause.
  • an advantage of a method of the invention is that it can be performed without the need of a large database with world knowledge which will decrease the amount of data to store. Moreover, the precision of such a method is high.
  • the use of relations for several different type of constituents also permits several answers to one question, and answers that do not necessarily identify objects by name but that still convey significant information to a user.
  • the invention identifies a limitation in prior art, where question answering systems have been considered to relate only to the answering of questions that have unique answers. In most cases this is not the case and such prior art methods thus have a limited applicability for a large set of questions (user information needs) .
  • the proposed method enables the finding of relations between persons or objects.
  • lexical meaning should be interpreted broadly. For example, in addition to word tokens that have the same lemma and word tokens that are synonyms, it is in some cases fruitful to consider word tokens that belong to the same broad semantic class to be considered as having equivalent lexical meanings.
  • names, definite descriptions and personal pronouns may be interpreted as having an equivalent lexical meaning, such as the name Jim Jarmusch, the definite description the director of Down by law, and the personal pronoun he .
  • One condition in the set of matching conditions is preferably a condition relating to a lexically headed constituent having the syntactic function of main verb in the question clause.
  • This condition stipulates that the lexically headed constituent having the syntactic function of main verb in the question clause has to have a corresponding constituent in a matching clause, i.e. a lexically headed constituent having the syntactic function of main verb and having an equivalent lexical meaning, in order for that clause to constitute an answer to the question clause.
  • This condition introduces the use of a condition that relates to a verb in the questions clause, which in prior art has not been considered to convey any significant information regarding the queried information.
  • Another condition in the set of conditions is preferably a condition relating to a lexically headed constituent having the syntactic function of subject in the question clause.
  • This condition stipulates that the lexically headed constituent having the syntactic function of subject in the question clause has to have a corresponding constituent in a clause, i.e. a lexically headed constituent having the syntactic function of subject and having an equivalent lexical meaning, in order for that clause to constitute an answer to the question clause.
  • Yet another condition in the set of conditions is preferably a condition relating to a lexically headed constituent having the syntactic function of object in the question clause.
  • This condition stipulates that the constituent having the syntactic function of object in the question clause has to have a corresponding constituent in the clause, i.e. a constituent bearing the syntactic function of object and having an equivalent lexical meaning, in order for that clause to constitute an answer to the question clause.
  • conditions on other constituents in clauses may be added to the set of conditions in order to increase the precision further.
  • Such conditions are for example conditions relating to constituents having the syntactic functions of manner adverb, place adverb, time adverb, and causal adverb, respectively, of the question clause, or conditions relating to constituents bearing any other syntactic function. Also these condition are preferably used in combination with one or more of the other conditions.
  • the conditions above may be used separately, but they are preferably combined in such a way that they jointly state necessary and sufficient conditions for a database clause to constitute an answer to a given question clause. This increases the precision of the method even further.
  • a condition regarding the constituents in the question clause may be defined stipulating that the constituents of the question clause, or constituents that have equivalent lexical meanings, should occur in a clause of the natural language text database in order for that clause to constitute an answer to the question clause.
  • conditions referring to a sequence of two or more clauses in the natural language text database are also envisaged.
  • One embodiment of the invention is directed to constituent questions (wh-questions) comprising an interrogative pronoun, such as what, who, which etc.
  • an interrogative pronoun such as what, who, which etc.
  • the syntactic function of the queried constituent of the question clause is determined not only in response to the analysis step, but also in response to the interrogative pronoun.
  • conditions can be defined that increase the precision of the method even further. This is due to the fact that the interrogative pronoun itself carries information of respective semantic classes of constituents of a clause that answers the question clause.
  • the syntactic function of the queried constituent is the same syntactic function as the interrogative pronoun has.
  • the syntactic function of the queried constituent will be another syntactic function than the interrogative pronoun has, but it will be possible to determine the syntactic function of the queried constituent based on the identified interrogative pronoun and the analysis in the analysis step.
  • the interrogative pronoun can also be used in order to determine the broad semantic class of the queried constituent. For example, the presence of the interrogative pronoun who in a natural language question indicates that the queried constituent is a noun phrase denoting a person.
  • Another embodiment concerns yes/no questions. These questions do not comprise any interrogative pronoun.
  • each constituents of a question clause in a yes/no question has a corresponding constituent, i.e. a constituent that has the same lexical meaning and the same syntactic function, in a clause that constitutes an answer to the question clause.
  • a yes/no question can be distinguished from a statement will differ depending on the language. For example in some language it can be determined from the word order.
  • the answer to the question may be found in a clause that satisfies the conditions.
  • the answer to the question clause will be evident to the user.
  • a yes or no answer can alternatively be derived automatically from the clauses that satisfy the conditions, and then presented to the user.
  • a system for automatically finding an answer to a natural language question by means of a computer stored natural language text database.
  • the system comprises storage means for storing the natural language text database.
  • the natural language text database has been analyzed with respect to syntactic functions of constituents, lexical meaning of word tokens, and clause boundaries.
  • the system also comprises analyzing means for analyzing a computer readable representation of a question clause of a natural language question with respect to syntactic functions of its constituents and lexical meaning of its word tokens, and defining means for defining, in response to an analysis performed by the analyzing means, a set of conditions for a clause in the natural language text database to constitute an answer to the question clause.
  • the conditions relate to syntactic functions of constituents and lexical meaning of word tokens in the clause.
  • the defining means are operatively connected to the analyzing means.
  • the system comprises answer finding means for identifying in the natural language text database clause that satisfy the conditions and for returning an answer to the question clause by means of the clauses that satisfy the conditions.
  • the answer finding means are operatively connected to the defining means and to the storage means.
  • figure 1 is a flowchart of a method according to an embodiment of the invention
  • figure 2 is an illustration of an example of an analyzed natural language question
  • figure 3A-B are illustrations of portions of text that constitute answers to the natural language question of figure 2
  • figure 4 is an illustration of another example of an analyzed natural language question
  • figure 5A-D are illustrations of portions of text that constitute answers to the natural language question of figure 4
  • figure 6 is a schematic diagram of a system according to an embodiment of the invention.
  • FIG 1 a flow chart of an embodiment of the invention is shown.
  • one or more answers to a natural language question are found in a natural language text database.
  • a natural language text database is a subset of the text information found in web servers connected to the Internet .
  • the natural language text database has been analyzed in an antecedent process thereby enabling the use of linguistic properties of the text database in order to find answers to a natural language question.
  • the analysis comprises the determination of a morpho-syntactic description for each word token of the natural language text database, a classification of the broad semantic class for each word token, the location of phrases in the natural language text database, the determination of a phrase type for each of the phrases, and the location of clauses in the natural language text database.
  • the morpho-syntactic description comprises a part-of-speech and an inflectional form
  • the phrase types comprise different types according to the syntactic functions of the phrases and the part of speech of their heads.
  • the syntactic functions comprise subject, object, main verb, adverbs etc.
  • a clause can be defined as a unit of information that roughly corresponds to a simple proposition, or fact.
  • the natural language text database has also been indexed and stored. The spaces between each word token are numbered consecutively, whereby the location of each word token is uniquely defined by the numbers of the two spaces it is located between in the natural language text database. The interval defined by these two numbers form a unique word token location identifier.
  • each word token is associated with a word type, it is sufficient to store all of the word types of the natural language text database and then, for each of the stored word types, store the word token location identifier of each word token associated with this word type. Furthermore, the location of a phrase is uniquely defined by the number of the space preceding the first word token of the phrase and the number of the space succeeding the last word token of the phrase. These two numbers form a phrase location identifier. Thus, each phrase type is stored and the phrase location identifier of each of the phrases of this phrase type is stored.
  • phrase location identifier due to the way the phrase location identifier is defined, it is easy to find out whether a word token occurs in a phrase of a certain type by determining whether the word token location identifier is included in a phrase of this type.
  • the location of a clause is uniquely defined by the number of the space preceding the first word token and the number of the space succeeding the last word token of the clause. These two numbers form a clause location identifier.
  • Each of the clause location identifiers is stored. Location identifiers for sentences, paragraphs, and documents are formed in an equivalent manner and each of them is stored.
  • a natural language question that is to be answered in this embodiment has been classified in a prior process either as a constituent question or a yes/no question. Furthermore, the question clause of the natural language question has been identified in a prior process as well.
  • the question clause is the clause of the natural language question that conveys the content of the information need. In a direct question, the question clause is the main clause, and in an indirect question the question clause is a subordinate clause.
  • a question clause is analyzed in the same way that the natural language text database has been analyzed, i.e. the syntactic function of its constituents and the lexical meaning of its word tokens are determined.
  • a set of conditions for a clause in the natural language text database to constitute an answer to the question clause are defined in step 104.
  • the conditions are that at least one of the constituents in the question clause should have corresponding constituents in the clause, i.e. constituents that each has the same syntactic function and an equivalent lexical meaning as the corresponding constituent in the question clause.
  • clauses that satisfy the conditions are identified in the natural language text database in step 106 of figure 1.
  • the word type of the natural language text database that correspond to a word token in the question clause, and that have a lexical meaning equivalent to the word tokens in the question clause are identified.
  • the word token location identifiers associated with the identified word types are identified in the index.
  • the identified word token location identifiers are then used to identify the word tokens in the natural language text database that are included in a phrase of the same type as the word token in the question clause is included in, i.e. a phrase that has the same syntactic function. This is done by searching the phrase location identifiers associated with the phrase type that the word token in the question clause is included in, and determining which of the identified word token location identifiers are included in one of these phrase location identifiers.
  • This comparison is done for each of a subset of the word tokens in the question clause, and in addition to determining if the word token is included in the same phrase type, it is determined whether the word tokens are included in the same clause. This can be done easily by determining whether the word token location identifiers are included in the same clause location identifier.
  • portions of text that each comprises one of the clauses that satisfy the set of conditions are extracted in step 108 of figure 1. These portions of text may then be presented to a user as an answer to the natural language question, or be further processed.
  • Figure 2 illustrates an example of an analyzed natural language question.
  • the question is: Who is expelling diplomats ? .
  • the question only includes one clause and the clause also constitutes a sentence.
  • the question clause of the question is the entire question.
  • the question clause has been analyzed with respect to a morpho-syntactic description for each word token, a lexical description (not shown) comprising lemma, a broad semantic class for each word token and synoyms, the location of phrases, a phrase type for each of the phrases, and the location of clauses.
  • the morpho-syntactic code is indicated, and for each space between the word tokens the number of the space is indicated.
  • the location of phrases and their respective type is also indicated.
  • a set of conditions is defined for a clause in an analyzed natural language text database to constitute an answer.
  • the natural language text database has been analyzed with respect to a morpho-syntactic description for each word token, lemma and a broad semantic class and a synonym set for each word token, the location of phrases, a phrase type for each of the phrases, the location of clauses, and the location of sentences.
  • expelling is the main verb
  • diplomats is the object noun phrase of the question clause.
  • FIG 3A-B portions of text that constitute answers to the natural language question of figure 2 are illustrated.
  • the answers have been extracted from the analyzed natural language text database .
  • a sentence is illustrated that includes an answer clause.
  • the first clause of the sentence has the main verb expelling, the object noun phrase Russian diplomats and the subject noun phrase the US.
  • the clause satisfies the conditions above.
  • the entire sentence that the clause is included in is extracted and presented as an answer.
  • FIG 3B a sentence is illustrated including only one clause.
  • the clause has the main verb expelling, the object noun phrase a matching number of US diplomats and the subject noun phrase Russia .
  • the clause satisfies the conditions above, and the clause is extracted and presented as an answer.
  • Figure 4 illustrates an example of an analyzed natural language question.
  • the question is: What did the ECB do? .
  • the question clause of the question is the entire question.
  • the question clause has been analyzed with respect to a morpho-syntactic description for each word token, lemma and a broad semantic class for each word token (not shown) , the location of phrases, a phrase type for each of the phrases, and the location of clauses.
  • the morpho-syntactic code is indicated, and for each space between the word tokens the number of the space is indicated.
  • the location of phrases and their respective type is also indicated.
  • the natural language text database has been analyzed with respect to a morpho-syntactic description for each word token, a broad semantic class for each word token, the location of phrases, a phrase type for each of the phrases, the location of clauses, and the location of sentences.
  • the ECB is the subject noun phrase
  • do is the main verb of the question clause.
  • the fact that the ECB is the subject noun phrase will give the condition that the head of the subject noun phrase in a clause should be the ECB in order for the clause to constitute an answer to the question.
  • the interrogative pronoun what together with the main verb do i.e. do_what, indicates that the queried constituent is an active verb phrase.
  • a further condition is that a clause should include an active verb phrase in order for the clause to constitute an answer to the question.
  • figure 5A-D portions of text that constitute answers to the natural language question of figure 4 are illustrated.
  • a sentence including clause boundaries within the sentence illustrates one answer to the question in figure 4.
  • the first clause of the sentence has the subject noun phrase the ECB and an active verb phrase has made mistakes .
  • the clause satisfies the conditions described with reference to figure 4.
  • the entire sentence that the clause is included in is extracted and presented as an answer.
  • a sentence including clause boundaries within the sentence illustrates a second answer to the same question.
  • the second clause of the sentence has the subject noun phrase the ECB and an active verb phrase imposed.
  • the clause satisfies the conditions described with reference to figure 4.
  • a sentence including clause boundaries within the sentence illustrates a third answer to the same question.
  • the first clause of the sentence has the subject noun phrase the ECB and an active verb phrase has never pursued a pure policy of minimising the rate of inflation .
  • the clause satisfies the conditions above.
  • the second clause also comprises an active verb phrase has taken a much more practical approach of maximising the rate of growth, but it does not include a subject noun phrase including the ECB and thus it does not satisfy the conditions.
  • the entire sentence that the first clause is included in has been extracted and presented as an answer.
  • the relation between the active verb phrase in the second clause and the ECB in the first clause will be apparent to a user.
  • a sentence including only one clause illustrates a fourth answer to the same question.
  • the clause has the subject noun phrase the ECB and an active verb phrase has performed almost spectacular well .
  • the clause satisfies the conditions above, and the clause is extracted and presented as an answer.
  • the system comprises analyzing means 602 for analyzing a computer readable representation of a clause, storage means 604 for storing an analyzed natural language text database, a question manager 606, defining means 610 for defining conditions for a clause to constitute an answer to a question clause, answer finding means 612 for finding clauses in a text database that constitutes answers to a question clause, and result managing means 620.
  • the text analyzing unit 602 is arranged to analyze a natural language text input , such as a natural language question or a natural language text database.
  • the analysis includes the determination of a morpho-syntactic description for each word token of the natural language input, a classification of the broad semantic class for each word token, the location of phrases in the natural language input, the determination of a phrase type for each of the phrases, and the location of clauses in the natural language input.
  • the morpho-syntactic description comprises a part-of-speech and an inflectional form
  • the lexical description of a word type comprises lemma, semantic class, and synonyms
  • the phrase types comprises different types denoting the syntactic functions of the phrases, such as subject noun phrase, object noun phrase, other noun phrases and prepositional phrases.
  • the memory means 604 operatively connected to the text analysis unit 602, are arranged to store a natural language text database that has been analyzed by the text analysis unit 602.
  • the natural language text database is stored in an index in the storage means 604.
  • the indexing is based on a numbering scheme where the spaces between each word token are numbered consecutively.
  • An alternative numbering scheme where each word token is consecutively number is also within the scope of the invention.
  • Each word token is then defined by its word type and the numbers of the two spaces it is located between in the natural language text database .
  • the two numbers of the spaces between which a word token is located form a word token location identifier for this word token.
  • a phrase is uniquely defined by its phrase type and the number of the space preceding the first word token of the phrase and the number of the space succeeding the last word token of the phrase.
  • the number of the space preceding the first word token of a phrase and the number of the space succeeding the last word token of the phrase form a phrase location identifier for this phrase.
  • a clause, a sentence, a paragraph and a document location identifier, respectively is defined as the number of the space preceding the its first word token and the number of the space succeeding its last word token.
  • the word types, word token location identifiers, phrase types, phrase location identifiers, clause location identifiers, paragraph location identifiers, sentence location identifiers and document location identifiers are stored in the index that is operatively connected to the indexer.
  • the logical and hierarchical structure of the index is shown in the table below:
  • the question manager 606 in figure 6 is operatively connected to the text analysis unit 602 and comprises defining means 610 for defining conditions for a clause in the natural language text database to constitute an answer to a question clause that has been analyzed in the text analysis unit 602.
  • the conditions are that a subset of the constituents in the question clause, should have corresponding constituents in the clause, i.e. constituents that each has the same syntactic function and an equivalent lexical meaning as the corresponding constituent in the question clause.
  • the question manager 806 comprises answer finding means 812 for finding clauses in the natural language text database that constitutes answers to the question clause.
  • the answer finding means 612 use the structure of the index in order to do identify clauses that satisfy the condition defined by the defining means 610.
  • the word type of a word token in a question clause By determining the word type of a word token in a question clause, the corresponding word type in the index, and other word types in the index that have an equivalent lexical meaning give the word token location identifiers since these are stored in the index. Furthermore, since the phrase type that the word token of the question clause is included in, and the phrase type that the word tokens of the natural language text database are included in has been determined in the text analysis unit, it can be determined which of the identified word token location identifiers are included in a phrase of the same type as the word token in the surface variant, i.e. that has the same syntactic function.
  • the system comprises a result manager 612, operatively connected to the storage means 604, for extracting each portion of text comprising a clause that satisfied the conditions that are defined by the defining means.
  • the portion of text to be extracted can be chosen as the clause satisfying the conditions, the sentence that the clause is included in, or the paragraph that the clause is included in, or the document that the clause is included in.
  • the extraction means use the index to find the desired units (clause, sentence, paragraph or document) by consulting the respective location identifiers in the index.

Landscapes

  • Machine Translation (AREA)

Abstract

L'invention concerne un procédé et un système permettant de trouver automatiquement une ou plusieurs réponse(s) à une question en langage naturel dans une base de données textuelles en langage naturel stockée dans un ordinateur. La base de données textuelles est analysée en termes de fonctions syntaxiques de constituents, de sens lexical d'unités lexicales et de limites de propositions, et la question en langage naturel comporte une proposition interrogative. Une représentation lisible par ordinateur de la proposition interrogative est analysée en termes de fonctions syntaxiques de ses constituents et de sens lexical de ses unités lexicales. En réponse à cette analyse, un ensemble de conditions selon lesquelles une proposition issue de la base de données textuelles en langage naturel peut constituer une réponse à la proposition interrogative est défini. Lesdites conditions concernent les fonctions syntaxiques des constituents et le sens lexical des unités lexicales de la proposition. Ensuite, les propositions qui remplissent lesdites conditions sont identifiées dans la base de données textuelles en langage naturel et la /les réponse(s) est/sont produite(s) au moyen desdites propositions.
EP02708919A 2001-03-30 2002-03-25 Procede permettant de trouver des reponses a des questions Ceased EP1386255A1 (fr)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
SE0101127A SE0101127D0 (sv) 2001-03-30 2001-03-30 Method of finding answers to questions
SE0101127 2001-03-30
PCT/SE2002/000585 WO2002080036A1 (fr) 2001-03-30 2002-03-25 Procede permettant de trouver des reponses a des questions

Publications (1)

Publication Number Publication Date
EP1386255A1 true EP1386255A1 (fr) 2004-02-04

Family

ID=30113476

Family Applications (1)

Application Number Title Priority Date Filing Date
EP02708919A Ceased EP1386255A1 (fr) 2001-03-30 2002-03-25 Procede permettant de trouver des reponses a des questions

Country Status (1)

Country Link
EP (1) EP1386255A1 (fr)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0631244A2 (fr) * 1993-06-24 1994-12-28 Xerox Corporation Un procédé et système pour le recouvrement d'informations
WO2001020500A2 (fr) * 1999-09-17 2001-03-22 Sri International Recuperation d'informations au moyen d'interrogations formulees en langage naturel

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0631244A2 (fr) * 1993-06-24 1994-12-28 Xerox Corporation Un procédé et système pour le recouvrement d'informations
WO2001020500A2 (fr) * 1999-09-17 2001-03-22 Sri International Recuperation d'informations au moyen d'interrogations formulees en langage naturel

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
See also references of WO02080036A1 *
VELARDI P; PAZIENZA M T; DE'GIOVANETTI M: "CONCEPTUAL GRAPHS FOR THE ANALYSIS AND GENERATION OF SENTENCES", BM JOURNAL OF RESEARCH AND DEVELOPMENT, INTERNATIONAL BUSINESS MACHINES CORPORATION, vol. 32, no. 2, 1 March 1998 (1998-03-01), NEW YORK, NY, US, pages 251 - 267, XP000022626, ISSN: 0018-8646 *

Similar Documents

Publication Publication Date Title
US7707023B2 (en) Method of finding answers to questions
Moldovan et al. Using wordnet and lexical operators to improve internet searches
Witten Text Mining.
AU2002364593B2 (en) System, method and software for automatic hyperlinking of persons' names in documents to professional directories
Kowalski Information retrieval systems: theory and implementation
CA2551803C (fr) Procede et systeme pour une recherche amelioree de donnees
US6584470B2 (en) Multi-layered semiotic mechanism for answering natural language questions using document retrieval combined with information extraction
US10296584B2 (en) Semantic textual analysis
US6678677B2 (en) Apparatus and method for information retrieval using self-appending semantic lattice
US20040122846A1 (en) Fact verification system
WO2001084376A2 (fr) Systeme pour repondre a des questions formulees en langage naturel
Bdour et al. Development of yes/no arabic question answering system
KR20010108845A (ko) 정보검색에서 질의어 처리를 위한 단어 클러스터 관리장치 및 그 방법
Kosinov Evaluation of N-grams Conflation Approach in Text-Based Information Retrieval.
Vossen et al. Meaningful results for Information Retrieval in the MEANING project
JP2004133564A (ja) 文書検索装置
Pai Text summarizer using abstractive and extractive method
Al-Lahham Index term selection heuristics for Arabic text retrieval
Ermakova et al. IRIT at INEX: question answering task
EP1386255A1 (fr) Procede permettant de trouver des reponses a des questions
Milić-Frayling Text processing and information retrieval
Liu Intelligent search techniques for large software systems.
JPH11195041A (ja) 文書検索装置、方法及び記録媒体
Murali Krishna et al. Extractive summarization technique based on fuzzy membership calculation using rough sets
Al-Lahham Arabic document indexing for improved text retrieval

Legal Events

Date Code Title Description
PUAI Public reference made under article 153(3) epc to a published international application that has entered the european phase

Free format text: ORIGINAL CODE: 0009012

17P Request for examination filed

Effective date: 20031027

AK Designated contracting states

Kind code of ref document: A1

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LI LU MC NL PT SE TR

AX Request for extension of the european patent

Extension state: AL LT LV MK RO SI

17Q First examination report despatched

Effective date: 20070214

RAP1 Party data changed (applicant data changed or rights of an application transferred)

Owner name: OPENAMPLIFY LIMITED

REG Reference to a national code

Ref country code: DE

Ref legal event code: R003

STAA Information on the status of an ep patent application or granted ep patent

Free format text: STATUS: THE APPLICATION HAS BEEN REFUSED

18R Application refused

Effective date: 20170305