CN104794109A - Intelligent answering system for learning machine - Google Patents
Intelligent answering system for learning machine Download PDFInfo
- Publication number
- CN104794109A CN104794109A CN201510167090.8A CN201510167090A CN104794109A CN 104794109 A CN104794109 A CN 104794109A CN 201510167090 A CN201510167090 A CN 201510167090A CN 104794109 A CN104794109 A CN 104794109A
- Authority
- CN
- China
- Prior art keywords
- semantic
- text
- answer
- frame elements
- learning machine
- 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.)
- Granted
Links
Landscapes
- Machine Translation (AREA)
Abstract
The invention belongs to the field of natural language processing research and particularly discloses an intelligent answering system for a learning machine. The intelligent answering system comprises a semantic analysis unit, a second relation establishing unit, a rational analysis unit and an answer extracting unit, wherein multiple obtained candidate answers are evaluated, the candidate answer with the highest evaluated value is used as an answer, and then a system result is provided. The intelligent answering system can be effectively combined with the learning machine of the prior art, will not be influenced by the number of questions of a question bank of the existing learning machine, can conduct analysis and judgment automatically and intelligently provide answers, and effectively fills up the blank in the aspect. The intelligent answering system has the advantages that corresponding relations between semantic roles are established in authentic Chinese materials, and an induction path is formed between the questions and the answers; the granularity of reading comprehension questions and answers reaches the chuck level, and more natural and more appropriate answers are provided.
Description
Technical field
The invention belongs to natural language processing research field, be specifically related to a kind of intelligence answer system being applied to learning machine.
Background technology
The mode that early stage learning machine adopts is all solidify on a memory, education resources such as a large amount of dictionaries, grammer, data for buyer provides some assisted learning functions.But its function is also limited by very large, consumer buys a learning machine, and the content being also only limitted to this learning machine and providing is assisted in the study obtained.Its function is comparatively single, cannot be expanded.Began from 2005, learning machine has developed into the audio-visual education programme product supporting that different study form and diversified subject are supported.Be mainly reflected in the aspects such as classroom is synchronously taught, multinational speech study, the application of standard terminological dictionary.
The appearance of computer learning machine, closely bound up with the development evolvement of mobile phone.Smart mobile phone is a kind of mobile phone installing corresponding open operating system, but the saying of intelligence is mainly for function, does not mean and itself has many " intelligence ".The birth of computer learning machine, is applied on mobile phone based on open operating system exactly, then by this type of technology, is applied to learning machine.What current most smart mobile phone adopted is Android (Android) intelligent operating system, and computer learning machine is also like this.
The computer learning machine in early stage, is mainly reflected in and the functional module of traditional learning machine is implanted in system with the form of APK.Along with the development of technology, learning machine manufacturer is many in android system, develops the learning system of oneself, is integrated by the education resource of system, and relies on mobile communication network if WIFI function is to realize download and the renewal of education resource.
Numerous product in study is released in prior art, as the software etc. of learning machine, numerous study aspect, all directly provide after answer is stored, especially in Text Inference and reading comprehension two, this mode of learning is more single, in the systems such as learning machine, the volume of exam pool is all limited, and there have study to be numerous unfavorable.
Summary of the invention
The present invention overcomes the deficiency that prior art exists, and aims to provide a kind of intelligent answer method and is applied in learning machine reading comprehension question and answer, and answer portrayed in language block rank.
In order to solve the problems of the technologies described above, the technical solution used in the present invention is: a kind of intelligence answer system being applied to learning machine, and concrete steps are as follows:
One, semantic analysis unit: after input text, first Chinese frame net semantic resources is utilized to carry out semantic character labeling to one section of text, follow Chinese sentence frame elements Marking Guidelines, use Chinese frame net semantic resources to carry out frame elements mark to the language block in all sentences, enter relation tectonic element;
Two, relation tectonic element: for the text through semantic analysis cell processing, first utilizes frame elements title semantic information to the syntax semantic association in text, namely judges that whether two frame elements titles are identical, if the same tectonic relationship; Otherwise utilize frame elements hyponymy semantic information to the syntax semantic association in text, namely judge whether two frame elements are correlated with, if relevant, tectonic relationship; Otherwise utilize frame elements semantic type semantic information to the syntax semantic association in text, namely judge that whether two frame elements semantic types are consistent, if the same tectonic relationship; Iteration is not stopped until build all relations for each sentence in text, forms question and answer inference graph, enter rational analysis unit;
Three, rational analysis unit: for the unknown semantics composition through marking in problem, can carry out rational analysis by the question and answer inference graph built, obtain multiple corresponding known semantic role, enter answer extracting unit after finding;
Four, answer extracting unit: the multiple candidate answers obtained are assessed, provides system results using the highest for assessed value as answer.
Wherein, the problem proposed for text is treated when composition notebook, is configured to semantic association, forms question and answer inference graph.
Wherein, described question and answer inference graph is on the basis of text, add a problem.Suppose the scene involved by problem, associated scenario must be had in the text corresponding, relevant semantic component also must be had in the text corresponding, after being marked by text for the unknown semantics role involved by problem, be fused in corresponding Text Inference, form question and answer inference graph.
Said process can be reduced to: first utilize Chinese frame net semantic resources to carry out semantic analysis to one section of text, and use the semantic informations such as frame elements title, frame elements hyponymy and frame elements semantic type to each syntax semantic association in text, simultaneously by some problems of proposing for text also when composition notebook is treated, structure textual association, forms question and answer inference graph; On this basis for the Unknown Component in problem, figure finds Induction matrix by inference; Every paths has one or more known semantic role corresponding to Unknown Component, so this known semantic role just can be used as a candidate answers, and the granularity level of this answer is in language block rank; Finally, the multiple candidate answers obtained are assessed, realizes reading comprehension question and answer using the highest for assessed value one as answer.
The present invention compared with prior art has following beneficial effect.
One, native system effectively can combine with learning machine of the prior art, not by the impact of the volume of the exam pool such as existing learning machine, can automatic analysis, judgement, provide answer intelligently, effectively filled up the blank of this aspect.
Two, the present invention is not when building inference rule, to utilize in Chinese frame net relation between distinctive frame elements, finding the corresponding relation of semantic role, realizing reasoning, present approach reduces reasoning cost, closer to the inference ideas of the mankind by analyzing language material.Method proposed by the invention takes full advantage of Chinese frame net semantic resources, semantic role analysis is carried out to text, a kind of text inference method is proposed, Text Inference is applied in reading comprehension question and answer, and feature the answer of a language block rank, for reading comprehension problem provides more lively proper answer.
Three, technique effect of the present invention mainly contains two, and one is the relation correspondence built under Chinese real corpus between semantic role, between problem and answer, define Induction matrix; Two is that the granularity of reading comprehension question and answer has been portrayed language block rank, defines more natural, more proper answer.
Four, the text inference method based on Chinese frame net is applied in reading comprehension question and answer, solves the question and answer problem in natural language processing, to the enlightening effect of natural language processing correlative study.The answer results of language block rank provides more facility also to the research of question answering system.The inventive method thinking structure is clear, successful, and extensibility is strong.
Accompanying drawing explanation
Below in conjunction with accompanying drawing, the present invention will be further described in detail.
Fig. 1 is system flowchart of the present invention.
Fig. 2 is relation tectonic element particular flow sheet in the present invention.
Embodiment
In conjunction with the accompanying drawings, the present invention is further detailed explanation, and accompanying drawing is the schematic diagram simplified, and only basic structure of the present invention is described in a schematic way, therefore it only shows the formation relevant with the present invention.
Chinese frame semantics net in the present embodiment be one with the frame semantics of Fillmore be theoretical foundation, with English FrameNet be with reference to, true with Chinese data be the Chinese vocabulary semantic database for computing machine of foundation.In frame semantics theory, the meaning of word describes and must be associated with semantic frame, and semantic frame usually describes event, sight, action or a state by using the framework metadatas such as participant, object, background.
Chinese frame net mainly comprises three partial contents---framework storehouse, lemma storehouse and example sentence storehouse.With regard to framework storehouse, framework wherein contains the description to various scene, i.e. the definition of various framework, also contains and explains various semantic role, i.e. various core frame element and non-core frame elements, wherein the displaying of core frame element to semantic scene is indispensable.After a semantic frame is activated, corresponding core frame element just has embodying of explicit or implicit expression.And for lemma storehouse and example sentence storehouse, specific lemma in lemma storehouse can activate a specific semantic scene, i.e. semantic frame, and just can must be achieved by example sentence corresponding with the certain sense of this lemma in example sentence storehouse when this specific lemma evokes corresponding frame.
In natural language processing field, for each sentence, each section of text, if use Chinese frame net to carry out semantic analysis to it, can be regarded as scene one by one in the theory of frame semantics mainly through frame relations to describe the relation between these scenes.In the data of FrameNet external disclosure, there are eight kinds of frame relations, altogether 1676 frame relations pair are constructed to 1190 semantic frames, two frameworks are had each relation centering, wherein relatively independent, more abstract one is referred to as host framework, and what another one was more specific is then referred to as the next framework, for the upper the next framework of often kind of frame relations, FrameNet has got a more specifically name, as table 1.
Table 1 frame relations type
| Relations | Sub | Super | Frame relations Chinese name |
| Inheritance | Child | Parent | Inheritance |
| Perspective_on | Perspectivized | Neutral | Visual angle relation |
| Subframe | Component | Complex | Total score relation |
| Precedes | Later | Earlier | Precedence relationship |
| Inchoative_of | Inchoative | State | Initial relation |
| Causative_of | Causative | Inchative/State | Causa relation |
| Using | Child | Parent | Use relation |
| See_also | Referring Entry | Main Entry | Reference relation |
Inheritance is the strictest a kind of relation of semantic corresponding relation, and the frame elements in father's framework, frame elements semantic type, frame elements close to tie up in subframe must have corresponding composition corresponding.Than father framework more specifically, therefore comprised semantic composition can be more for the semantic scene that certain subframe describes.In addition, if father's framework and some other framework have frame relations, so the next framework also should have corresponding frame relations; Use relation is the hyponymy comparatively more wide in range than the semantic corresponding relation of inheritance, and the next framework only employs a part of semantic information; Total score relation is when analyzing some comparatively complicated scenes, some states related to for this scene and transition state may form independent framework, the scene that each self-described one is relatively independent, just defines total score relation like this between COMPLEX FRAME and these comparatively simple frameworks of its bottom.Such as complicated " employer " scene, just relate to " recruitment ", " employing " and " dismissing " several little scene; Perspective relation is a kind of special use relation, that the different visual angles participating in role to the difference that may relate in a framework is set out the frame relations formed, such as in " work " scene, form respectively " employer " and " employee " scene from the different angles of employer and employee; Cause-effect relationship and initial relation are the frameworks considered morphological change and build, and cause-effect relationship describes the relation between transitive verb and intransitive verb, and initial framework describes the relation between intransitive verb and state adjective; Precedence relationship describes ordinal relation intrinsic between event or state, and " recruitment ", " employing " and " dismissing " that such as " employer " framework relates to just has fixing ordinal relation; Also have some frameworks to seem closely similar due to the similarity of semanteme, but in fact have fine distinction, in order to be distinguished, between the similar framework that these have otherness, establish reference relation.
According to the character of Chinese frame net, under different frames or under same framework different sentence example semantic role between there is corresponding relation.Semantic ambiguousness own is very large, and the present invention not only considers the hyponymy of frame relations middle frame element when carrying out reasoning research, also contemplates the correspondence between identical frames element under different frames, and the similarity etc. of frame elements semantic type.For a text, after utilizing Chinese frame net semantic resources to carry out full text mark, between different semantic roles, just can build corresponding relation, thus form a Text Inference figure, for the semantic role that will be studied, multiple correspondence can be constructed from this inference graph.
And for question and answer reasoning, be on the basis of Text Inference, add a problem.Because problem proposes for text, the present invention's hypothesis, for the scene involved by problem, must have associated scenario corresponding in the text, relevant semantic composition also must be had in the text corresponding for the unknown semantics role involved by problem.After problem sentence marks by the present invention, be fused in corresponding Text Inference figure, the process therefore finding problem answers just becomes the semantic role that in searching problem, unknown semantics role is corresponding in the text.Problem only has one, and unknown semantics role also only has one, but semantic role corresponding with it in the text may have multiple.Finally, the multiple candidate answers obtained are assessed, realizes reading comprehension question and answer using the highest for assessed value one as answer.
Being embodied as on learning machine:
As depicted in figs. 1 and 2, a kind of intelligence answer system being applied to learning machine, concrete steps are as follows:
One, semantic analysis unit: after input text, first Chinese frame net semantic resources is utilized to carry out semantic character labeling to one section of text, follow Chinese sentence frame elements Marking Guidelines, use Chinese frame net semantic resources to carry out frame elements mark to the language block in all sentences, enter relation tectonic element;
Two, relation tectonic element: for the text through semantic analysis cell processing, first utilizes frame elements title semantic information to the syntax semantic association in text, namely judges that whether two frame elements titles are identical, if the same tectonic relationship; Otherwise utilize frame elements hyponymy semantic information to the syntax semantic association in text, namely judge whether two frame elements are correlated with, if relevant, tectonic relationship; Otherwise utilize frame elements semantic type semantic information to the syntax semantic association in text, namely judge that whether two frame elements semantic types are consistent, if the same tectonic relationship; Iteration is not stopped until build all relations for each sentence in text, forms question and answer inference graph, enter rational analysis unit;
Three, rational analysis unit: for the unknown semantics composition through marking in problem, can carry out rational analysis by the question and answer inference graph built, obtain multiple corresponding known semantic role, enter answer extracting unit after finding;
Four, answer extracting unit: assess the multiple candidate answers obtained, one that assessed value is the highest provides system results as answer.
Wherein, the problem proposed for text, also when composition notebook is treated, is configured to semantic association, forms question and answer inference graph.
Wherein, described question and answer inference graph is on the basis of text, add a problem, suppose the scene involved by problem, associated scenario must be had in the text corresponding, also relevant semantic component must be had in the text corresponding for the unknown semantics role involved by problem, after text is marked, be fused in corresponding Text Inference, form question and answer inference graph.
By reference to the accompanying drawings embodiments of the invention are explained in detail above, but the present invention is not limited to above-described embodiment, in the ken that those of ordinary skill in the art possess, can also makes a variety of changes under the prerequisite not departing from present inventive concept.
Claims (3)
1. be applied to an intelligence answer system for learning machine, it is characterized in that: comprise the steps:
One, semantic analysis unit: after input text, first Chinese frame net semantic resources is utilized to carry out semantic character labeling to one section of text, follow Chinese sentence frame elements Marking Guidelines, use Chinese frame net semantic resources to carry out frame elements mark to the language block in all sentences, enter relation tectonic element;
Two, relation tectonic element: for the text through semantic analysis cell processing, first utilizes frame elements title semantic information to the syntax semantic association in text, namely judges that whether two frame elements titles are identical, if the same tectonic relationship; Otherwise utilize frame elements hyponymy semantic information to the syntax semantic association in text, namely judge whether two frame elements are correlated with, if relevant, tectonic relationship; Otherwise utilize frame elements semantic type semantic information to the syntax semantic association in text, namely judge that whether two frame elements semantic types are consistent, if the same tectonic relationship.Iteration is not stopped until build all relations for each sentence in text, forms question and answer inference graph, enter rational analysis unit;
Three, rational analysis unit: for the unknown semantics composition through marking in problem, can carry out rational analysis by the question and answer inference graph built, obtain multiple corresponding known semantic role, enter answer extracting unit after finding;
Four, answer extracting unit: the multiple candidate answers obtained are assessed, provides system results using the highest for assessed value as answer.
2. a kind of intelligence answer system being applied to learning machine according to claim 1, is characterized in that, the problem proposed for text, also when composition notebook is treated, is configured to semantic association, forms question and answer inference graph.
3. a kind of intelligence answer system being applied to learning machine according to claim 1, it is characterized in that, described question and answer inference graph is on the basis of text, add a problem.Suppose the scene involved by problem, associated scenario must be had in the text corresponding, relevant semantic component also must be had in the text corresponding, after being marked by text for the unknown semantics role involved by problem, be fused in corresponding Text Inference, form question and answer inference graph.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510167090.8A CN104794109B (en) | 2015-04-09 | 2015-04-09 | A kind of intelligent answer system for being applied to learning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510167090.8A CN104794109B (en) | 2015-04-09 | 2015-04-09 | A kind of intelligent answer system for being applied to learning machine |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104794109A true CN104794109A (en) | 2015-07-22 |
| CN104794109B CN104794109B (en) | 2017-06-23 |
Family
ID=53558907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510167090.8A Active CN104794109B (en) | 2015-04-09 | 2015-04-09 | A kind of intelligent answer system for being applied to learning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104794109B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107832295A (en) * | 2017-11-08 | 2018-03-23 | 山西大学 | The title system of selection of reading machine people and system |
| CN107967293A (en) * | 2016-10-20 | 2018-04-27 | 卡西欧计算机株式会社 | Learn auxiliary device, study householder method and recording medium |
| WO2021184311A1 (en) * | 2020-03-19 | 2021-09-23 | 中山大学 | Method and apparatus for automatically generating inference questions and answers |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103823794A (en) * | 2014-02-25 | 2014-05-28 | 浙江大学 | Automatic question setting method about query type short answer question of English reading comprehension test |
| US20150026106A1 (en) * | 2012-02-23 | 2015-01-22 | National Institute Of Information And Communcations Technology | Non-factoid question-answering system and computer program |
-
2015
- 2015-04-09 CN CN201510167090.8A patent/CN104794109B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150026106A1 (en) * | 2012-02-23 | 2015-01-22 | National Institute Of Information And Communcations Technology | Non-factoid question-answering system and computer program |
| CN103823794A (en) * | 2014-02-25 | 2014-05-28 | 浙江大学 | Automatic question setting method about query type short answer question of English reading comprehension test |
Non-Patent Citations (1)
| Title |
|---|
| 李茹 等: "基于汉语框架网的旅游信息问答系统设计", 《中文信息学报》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107967293A (en) * | 2016-10-20 | 2018-04-27 | 卡西欧计算机株式会社 | Learn auxiliary device, study householder method and recording medium |
| CN107967293B (en) * | 2016-10-20 | 2021-09-28 | 卡西欧计算机株式会社 | Learning support device, learning support method, and recording medium |
| CN107832295A (en) * | 2017-11-08 | 2018-03-23 | 山西大学 | The title system of selection of reading machine people and system |
| CN107832295B (en) * | 2017-11-08 | 2021-06-04 | 山西大学 | Title selection method and system of reading robot |
| WO2021184311A1 (en) * | 2020-03-19 | 2021-09-23 | 中山大学 | Method and apparatus for automatically generating inference questions and answers |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104794109B (en) | 2017-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Matsuda et al. | World Englishes and the teaching of writing | |
| Marshall et al. | 2B or not 2B plurilingual? Navigating languages literacies, and plurilingual competence in postsecondary education in Canada | |
| Blum et al. | Universals of lexical simplification | |
| Sasaki et al. | Explanatory variables for EFL students' expository writing | |
| Seargeant | Naming and defining in world Englishes | |
| CN101251862B (en) | Content-based problem automatic classifying method and system | |
| Freire et al. | The expropriation of dual language bilingual education: Deconstructing neoliberalism, whitestreaming, and English-hegemony | |
| Savski | Local problems and a global solution: Examining the recontextualization of CEFR in Thai and Malaysian language policies | |
| CN109710261A (en) | A kind of method and electronic equipment that building blocks code is automatically converted to programming code | |
| Jian | The schematic structure of literature review in research articles of applied linguistics. | |
| CN104794109A (en) | Intelligent answering system for learning machine | |
| Starr et al. | Inclusion, exclusion, and racial identity in Singapore's language education system | |
| CN105786971A (en) | International Chinese-teaching oriented grammar point identification method | |
| CN116861242A (en) | Language perception multi-language pre-training and fine tuning method based on language discrimination prompt | |
| Wirag et al. | Applying cognitive linguistics to foreign language teaching and learning: Addressing current research challenges | |
| Zhou | Nation-state building and rising China | |
| Wood | Computer assisted reading in German as a foreign language, developing and testing an NLP-based application | |
| CN106446264B (en) | Document representation method and system | |
| Pakir | The Kachruvian paradigm and its range and depth of influence | |
| López-Sánchez | Re-writing the goals of foreign language teaching: The achievement of multiple literacies and symbolic competence | |
| CN112765359B (en) | Text classification method based on few samples | |
| Ilynska et al. | Application of LSP texts in translator training | |
| Bacha et al. | Tela: Towards environmental learning arabic | |
| CN101989387A (en) | System and method for foreign language sentence pattern loanword training based on student interlanguage management | |
| Harris et al. | Issues in all-Irish education: Strengthening the case for comparative immersion |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |