CN116050420B

CN116050420B - Chinese and French voice semantic recognition method and device based on preposition sentence

Info

Publication number: CN116050420B
Application number: CN202211416411.XA
Authority: CN
Inventors: 袁伟菡
Original assignee: Wuhan University WHU
Current assignee: Wuhan University WHU
Priority date: 2022-11-12
Filing date: 2022-11-12
Publication date: 2023-09-22
Anticipated expiration: 2042-11-12
Also published as: CN116050420A

Abstract

The application relates to a Chinese and French semantic recognition method and equipment based on preposition sentences, when detecting that an input native language contains preposition sentences, the method is executed with the following steps; firstly, analyzing prepositive sentences into target objects and background objects according to a native language grammar; modeling the target object and the background object, establishing a relation in the model, and analyzing the relation; and secondly, in the translation, according to the relation between the target object and the background object, calling or translating the corresponding prepositions, and carrying out post-translation synthesis on the target object and the background object to obtain the translation semantics and then outputting the translation semantics. The application has reasonable design, compact structure and convenient use.

Description

Chinese and French voice semantic recognition method and device based on preposition sentence

Technical Field

The application relates to the field of Chinese and French semantic recognition, in particular to a Chinese and French semantic recognition method and equipment based on preposition sentences in static accommodation orientations.

Background

Space is the basis of various concept categories of human beings, and the process of systematic research thereof is characterized by the essential philosophy discussion of the space in Missioudet, then the psychological experiment analysis of the physical and temporal relationship and the shape of the substance movement, the psychological experiment analysis of the perception of the body-physical and psychological spacial orientation, and then the language is changed. Chinese and french originate from different language systems, and the existing CN1542649a natural language generation system is a language information statistical model for the constituent structure of the ordering in sentence realization. Although the implementation of sentence sorting is provided, in translation, in the middle-method translation process, the voice is generally analyzed, and is subjected to numerical processing to generate native language characters, and then the translation is performed according to phrases from a traversal database to generate translations. However, the applicant finds that the middle method is derived from different language systems, and when phrase recombination is performed, such as transliteration and interpretation is obscure and easy to generate ambiguity, but the existing language semantic analysis does not solve the technical problem, and the applicant finds that the translation accuracy is poor, and the applicant finds that the translation of the middle method can generate huge ambiguity aiming at sentences containing prepositions, so that the translation is inaccurate.

CN101201818A uses HMM to calculate language structure, make word segmentation, machine translation and speech recognition, and adopts hidden markov model to solve long sentence recognition so as to make sentence smooth, but it can not implement automatic recognition, and solves the problem of sentence sequential smooth when Chinese and French speech and semantic are mutually translated.

CN102089805a is a system and method for concept mapping, which adopts a concept mapping method to solve the word translation correctness, but cannot solve the statement passing problem.

The CN106471570A multi-command single-utterance input method only solves the problem that accurate translation cannot be realized when voice input generates correct native language semantics, and the CN109754809A voice recognition method, device, electronic equipment and storage medium mainly aim at a main-meaning object sequence structure, particularly Chinese, cannot realize translation and have no technical significance although voice semantic recognition is provided. CN110083837a is a keyword generating method and device, which is a fast generating method, and does not implement all traversal of speech semantics. The CN110326041A is used for natural language interaction of intelligent assistants, is currently in a substantial examination stage, is unknown whether a scheme can be realized, is different from the technical thought of the application, solves different problems, and is a CN110782870A voice synthesis method, device, electronic equipment and storage medium, and only solves the problem of semantic voice output generated by translation.

How to improve the recognition capability of a translation machine on the voice semantics of a centering method and ensure the translation accuracy becomes a technical problem to be solved urgently.

In addition, when the product scheme falls to the ground, the inventor aims at the problems that the matched product equipment cannot automatically identify and correct positions and the main board is easy to be blocked and damaged, and performs secondary improvement, and of course, the principle structure of the equipment can be expanded to other voice semantic identification inter-translation equipment, and the premise is that the equipment at least comprises an integrated voice semantic hardware module, a shell, an upper cover and the like.

Disclosure of Invention

The inventor finds that the expression of the space in the semantics mainly shows the description of the relative position relation among the entities, and the expression can be divided into a specific entity-to-entity relation and a specific entity-to-abstract entity relation according to the attribute of the entity, wherein the former is a real azimuth relation, and the latter is an abstract azimuth relation; the spatial relationship can be divided into static and dynamic according to the relative positional relationship. The static spatial relationship is often embodied as an azimuthal relationship.

The static spatial orientation relationship is expressed by a structure of 'preposition + background object' or '(preposition+) background object + orientation word', for example, when the 'accommodating' spatial orientation relationship is expressed, the 'preposition dans + background object' is adopted in French, the 'preposition +) background object + orientation word' is adopted in Chinese, and the 'in + background object + position' is adopted in Chinese. The understanding of the differences in expression patterns cannot be achieved solely by one-to-one correspondence of language symbols, and is particularly evident in abstract static bearing relationships. In order to reduce the obstacle of the Chinese-French semantic exchange to the greatest extent, the application takes the theory of the idiom as a guide, and improves the speech semantic recognition efficiency through the analytic design of the use condition of prepositions and azimuth words which mainly express the accommodating space relation.

When detecting that the input native language contains prepositional sentences, executing the following steps of the method;

firstly, analyzing prepositive sentences into target objects and background objects according to a native language grammar;

modeling the target object and the background object, establishing a relation in the model, and analyzing the relation;

and secondly, in the translation, according to the relation between the target object and the background object, calling or translating the corresponding prepositions, and carrying out post-translation synthesis on the target object and the background object to obtain the translation semantics and then outputting the translation semantics.

The system comprises a voice input module, a semantic analysis modeling module and a text processing module, wherein the voice input module converts the native language voice into characters and then carries out semantic analysis modeling and/or directly carries out semantic analysis modeling on the native language voice;

the system comprises a voice output module, wherein the voice output module is used for converting semantic synthesis into translated text and outputting the translated text after converting the translated text into voice and/or outputting the translated text after converting the semantic synthesis into the translated text voice.

Wherein the background object is static and/or dynamic relative to the object;

the context includes a real entity and/or an abstract entity;

the dimension after background modeling is three-dimensional, two-dimensional and/or one-dimensional;

the background object has concave-convex degree and/or flatness;

the relationship includes a supporting relationship and/or a containment relationship, the containment including a penetrating partial containment and/or a surrounding full containment.

When the preposition phrase adopts a static space azimuth relation, setting a French constitution formula to adopt prepositions and background objects, and setting a Chinese constitution formula to adopt prepositions, background objects and azimuth words;

when the background object and the target object are identified as the bearing relation of the accommodation space, the bearing object is set as prepositions dans+the background object in French, and the bearing object is set as prepositions+the background object in Chinese;

when the background object is judged to be a real entity, modeling is directly carried out according to the corresponding object of the mother language, and the modeling is stored in a database or an existing model prestored in the database is called;

when the background object is judged to be an abstract entity, metaphor modeling cognition is carried out, and modeling is stored in a database or an existing model prestored in the database is called; constructing a background structural formula by the object and the background object;

when the background object of the abstract entity is considered to be accommodating in the French, the abstract entity is correspondingly regarded as a container characteristic; when the background object of the abstract entity is Chinese and is not considered to have container characteristics, and French is considered to have accommodative property, introducing an expert database, in French, correspondingly taking the abstract entity as container characteristics, and establishing three-dimensional space dimension of the abstract entity so as to form a background object by entering sentences;

in the process of analyzing the relationship, aiming at the modeled model, the feature part of the relationship and the object is enhanced, the feature part is regarded as the feature of the whole entity, and the rest features or dimensions are discarded.

When the relation of the object in the background object is translated into French in Chinese, when the object is positioned at the high position of the background object and is in direct or indirect contact with the background in the model relation, the relation is smaller than the background object, and the gravity of the object is opposite to the relative motion direction between the object and the background, the preposition sur is called;

when the boundary part or the whole of the target object is contained by the background object, the preposition dans is called;

when the accommodative property, namely the concavity of the modeled container is lower than a set threshold value, aiming at French, the relation between the target object and the background object tends to be related to the supported object, and sur is called; conversely, the relationship between the object and the background object tends to be the relationship between the object and the container, and dans is called;

the priority of the relationship is higher than the priority of the background modeling dimension;

when the background object is modeled as a plane, invoking sur;

when the relation is that the target object penetrates or is put in or goes deep into the background object, invoking dans;

invoking sur when the directionality of the restriction of the background object to the target object is larger than a set threshold;

the relation model of the background object and the target object comprises a protection type model, a limiting type model, a position limiting type model, a sight limiting type model and a relation type model;

when the object is of a protection type or a limiting type, the object is placed in a background object, is wrapped by the background object and is subjected to constraint force in an inward direction, and a corresponding Chinese translation sentence is intersected with a dynamic image pattern on the basis of recognizing and analyzing a container pattern, so that attention is focused on a position limiting feature and a force application-bearing relation, namely, the source direction of the constraint force applied to the content by the container or the specific position and direction of the force generation on a force application surface can be determined on the azimuth word;

From the perspective of metaphorically recognizing the drawings, in modeling, under the dynamic action of the target object on the background object, the spatial position relationship of the inner and outer properties of the target object and the background object occurs;

in Chinese, when the model is considered as one-dimensional and two-dimensional surface call, the three-dimensional model is considered as the body call;

the method is matched with and updated with a Chinese database and a French database, and the Chinese database and the French database are linked through the relation in a model database; the French database establishes a category applicable to a real entity and an abstract entity in the received azimuth constitutive dans+ background scene; finding out the background object + inner or the background object + upper of the category and the Chinese database receiving azimuth structure;

the French database is used for determining a space background category suitable for accommodating the azimuth structure based on an informationized French treasury database, a French assistant online dictionary and/or French modern grammar; in the Chinese database, based on Xinhua dictionary, modern Chinese grammar and/or thesaurus construction, when the object is in the closed space of the background object model container, the azimuth word structure background object is called, and when the background object model container has openness relative to the object, the azimuth word structure background object is called.

The apparatus includes a housing member; a main board component is arranged on the shell piece; the shell piece is buckled with an upper cover piece; the main board of the main board assembly is provided with a process notch; the main board of the main board assembly is provided with a process notch; a processor is arranged on the main board; the processor is electrically connected with a voice input module, a wireless transceiver module, a power supply, a voice output module and a memory; the memory is used for storing data, the database comprises a Chinese database and a French database, and the Chinese database and the French database are linked through the relation in the model database; the French database establishes a category applicable to a real entity and an abstract entity in the received azimuth constitutive dans+ background scene; find out the background object + inner or the background object + upper of the category and the Chinese database receiving direction.

The production line comprises a carrier and a loading transmission group for supporting the carrier and having a mould adjusting station and an opening station for placing and/or detecting the adjusting station; the output end of the feeding conveying group is connected with a direction changing station of the circulating conveying assembly; a take-out station on the endless conveyor assembly; the taking-out station is connected with an assembly line through a transfer device procedure; a plurality of clamping manipulators are distributed on the assembly line.

The production line carries out the following steps that SA, empty carriers are placed on a feeding conveying group and conveyed, and the carriers carrying the workpieces are blocked from moving forward when ascending at each station through an L-shaped lower bracket;

SB, the output carrier of the feeding conveying group is conveyed in a direction changing way through the circulating conveying component; the circulating conveying assembly outputs the empty carrier after the station is taken out, and waits for being sent to the feeding conveying group for the second time after the empty carrier is taken out;

and SC, the main board is sent to an assembly line through a clamping manipulator to be placed in the shell.

Aiming at the mid-method voice semantic recognition, in order to improve the accuracy of translation sentences, the applicant finds that deviation exists in mid-method inter-translation, and when translation deviation errors occur for prepositions sentences, especially when transliteration is adopted, the translation accuracy is poor. The invention improves the accuracy of the inter-translation and effectively reduces the error rate. Wherein, the proper noun and the voice recognition can adopt general technology. The invention processes and recognizes prepositions around the applicant, thereby realizing the improvement of translation accuracy.

Drawings

FIG. 1 is a schematic diagram of a semantic recognition infrastructure. FIG. 2 is a diagram of speech semantic modeling. Fig. 3 is a schematic diagram of a computer semantic recognition process. Fig. 4 is a schematic diagram of a semantic recognition refinement structure. Fig. 5 is a database execution flow chart. FIG. 6 is a flow diagram of speech semantic preference. FIG. 7 is a diagram of a preferred phonetic semantic overall preference. Fig. 8 is a diagram of a preferred database structure. Fig. 9 is a schematic diagram of a preferred portable translator. Fig. 10 is a schematic view of a preferred motherboard configuration. Fig. 11 is a preferred carrier flow architecture diagram. Fig. 12 is a schematic view of a preferred carrier structure. Fig. 13 is a diagram showing a preferred two-aid improvement. Fig. 14 is a schematic view of a preferred inclined support plate structure. Fig. 15 is a modified structural view of a preferred opening mode. Fig. 16 is a preferred modeling diagram.

Detailed Description

The application is described in further detail below with reference to fig. 1-16.

Example 1 physical verification of static accommodation azimuth relationship

Based on the accommodation thinking form, the method is embodied in a language according to the cognition and understanding mode of the real space azimuth relation, the abstract entity is encoded, the spatial attribute is defined, and the azimuth relation expressed as the abstract is decoded, so that the abstraction and quantification are realized. The entity related to the real space azimuth relation structural formula as a reference object or background has space characteristics and can directly provide space reference for a target object; the entity serving as the reference background in the abstract spatial azimuth relation structural formula does not have space, and can form the background structural formula with the target object after being assigned through metaphor modeling and cognition.

The accommodation orientation relationship formed between the real entities being based on the accommodation of one of the entities itself, e.g. Le cadeau est dans la(gift in box), +.>The (box) itself has the property of containing the item. In the accommodation relation formed between abstract entities, accommodation cognition of real entities is projected on the accommodation relation, and the accommodation cognition is conceptualized as abstract accommodation entities. For example, in Qu 'est-ce quis' est pass endas la rue (what happens on a road), rue (road) itself has a supporting property, but does not have a holding property, but french uses prepositions dans expressing holding relationships to match with it, so as to form a "dans+background scene" structure, and the road is given container features through holding cognition, so that the road can serve as a background in abstract entity azimuth relationships. The static inclusion space orientation expression is a structural formula formed by medium word class indicating that an object is in the space range of another object, and the structural formula corresponds to a structural formula of 'dans+background scene' in French, and a structural formula of 'background object+inner' in Chinese. Through the cognitive assignment of the container concept metaphor, the abstract entity in the constitution obtains the three-dimensional space dimension which is not possessed originally, so that the entering sentence forms the background. The static inclusive spatial orientation relationship expressed by using the structural formula of 'dans+background object' in French is treated as 'background object+inner' of the same structural formula in Chinese translation sentence, but the prior art is treated as structural formula opposite translation using the static supportive spatial orientation word 'upper+background object'.

The inventors have found that the choice of an orientation word is related to how to look at the geometric properties of the object represented by the noun. In semantic analysis, a part of an entity and its characteristic are emphasized, and are regarded as the characteristics of the whole entity, and other parts are desalted and discarded. In particular, in an abstract entity containment aspect, the abstract entity is assigned a containment attribute that is only a part of the real-world entity characteristics associated therewith. When the Chinese cognitive main body is used for recognizing the solution language static state to accommodate the azimuth relation, the three-dimensional space features of the entity serving as the background are not necessarily completely utilized, and in many cases, only a certain space dimension can be highlighted. Take the following two sentences as examples:

(1) Il dort dans son lit he sleeps in bed.

(2) De telles occasions sont rares dans l 'histoirede l' humanite. In human history, this opportunity is not so great.

In the example sentence (1), the "lit" (bed) is a real entity having a characteristic that itself can serve as a spatial background. "cas" (case) in the example sentence (2) then belongs to an abstract entity to which a spatial attribute is given via a container metaphor. French uses the static inclusive space preposition structure type "dans+background object" salient bed as the three-dimensional accommodation property of the background, while Chinese sentence translation uses the static supportive space azimuth word structure type "background object+upper" salient bed as the two-dimensional planarity or one-dimensional linearity of the background. The dimension steering that occurs when the same entity acts as a background in different languages is showing the difference in the extraction of spatial attributes of the entity by two language users. Such differences occur not only from interlingual to interlingual, but also in expressions in the same language.

(3) Il est sur son lit he lies in bed.

For the same entity serving as a background, different structural expressions are selected, so that the obvious change of the spatial attribute of the entity can be reflected. On the basis of summarizing and summarizing features of French prepositions dans and sur, in the expression of the spatial position, when a target object is positioned at the high position of a background, is in direct or indirect contact with the background and is smaller than the background, the gravity of the target is opposite to the relative motion direction between the target and the background, the prepositions sur is needed to be used; and dans is used when the boundary of the object is partially or completely contained by the background (whether or not there is contact). When the container is not sufficiently typical in containment, the relationship between the target and the background changes when the concavity is different, resulting in a bias in the choice of the two by the language user. In French, the greater the concavity, the more the relationship between the target and the background tends to be the contents/container relationship, so dans is used; the less concavity the more the relationship between the target and the background tends to be the supported/support relationship, so sur is used. Taking the most typical "armchair/chair" entity as background for example:

le chef comptable est assis dans le the accounting is sitting on armchair.

(5) L' huissier est aussi sur la chair.

Armchairs as a spatial background have a concavity that is significantly greater than that of chairs, so that the choice of prepositions with which they are associated is generally not confused. The inventors have also encountered the following:

l' antenna de t vision est sur le toit. The antenna of the television is on the roof of the house.

(7) L' arbre est dans la tere. Tree (root) is in the ground.

(8) La pomme est sur le apple is on a teadish.

(9) La pomme est dans le the apples are in a tray.

(10) C' est lui le fautif dans cette affaire he is the missing party in this event.

(11) Le ministre exige le silence absolu sur cette affaire.

In the example sentences (6) and (7), the toit (roof) and the tere (earth) serving as the backgrounds are both planes, but the target l 'antenna in the example sentence (6) does not penetrate the background ceiling, so the preposition sur is applicable, and the target l' arbre in the example sentence (7), especially the root thereof penetrates the background labre (earth), enters the underground space, so the preposition dans is used. I.e., whether the object is entered into a planar background, it is determined whether the background remains two-dimensional or is converted into a three-dimensional space, thereby further selecting the prepositions (see fig. 5 and 6). The difference between the example sentences (8) and (9) is that the functions of the le plateau (tea tray) and the le plat (dish) are different, so that prepositions of different space dimensions of the table are matched: the former is intended to hold and the latter is intended to hold, although both approximate a two-dimensional plane in vision, because the dish has a greater concavity than the teadish, the latter is collocated with dans instead of sur. The difference in the spatial context in function also affects the choice of prepositions (see figures 5 and 7). The example sentences (10) and (11) serve as abstract spaces of the background, and from the view of the relation between the target object and the background, the error party's' he 'in the example sentence (10) is the participant of the event, so that the le fautif is placed in or goes deep into the l' affaire, and the relation between the tree root in the example sentence (7) and the ground can be metaphorically modeled. In the example sentence (11), confidentiality is a requirement, and there is a limit to the event, and the limit has obvious directionality, so the preposition sur is collocated.

The different expression chosen in expressing the static receptive bearing is based on consideration of the entity itself acting as a background. A real entity and an abstract entity, and only some spatial characteristics of a source entity are concerned. Whether the accommodability of the entity is typical enough, whether the boundary is obvious, the opening degree is high or low, the concavity of the entity and the like can influence the selection of the expression;

example 2 influence of awareness of accommodation orientation relationships by decoding into a combinatorial collocation of the image patterns and its metaphors

The accommodation space concept is generated based on accommodation thinking, and the corresponding image diagram is a 'container' diagram. When the relationship is analyzed as the 'inner/outer', the relationship includes a protection type, a restriction type, a position restriction type, a sight line restriction type and a relationship type. Taking example sentences (1) and (2) as examples, the corresponding French example sentences are of a protection type or a limiting type, the emphasis is that the content is wrapped by the container due to being placed in the container, the specific source direction of the force is not clear, the specific position of a force application point or surface cannot be determined (see figure 1), the corresponding Chinese translation sentence is intersected with the dynamic image drawing on the basis of recognizing the container drawing, and the attention is focused on the position limiting characteristic and the force application-bearing relation, namely the source direction (three-dimensional, see figure 2) of the constraint force applied by the container or the specific position and direction (two-dimensional or one-dimensional, see figures 3 and 4) of the force generation on the force application surface can be determined through the azimuth word.

From the metaphorically-aware drawings, under the action of power, things can have a spatial positional relationship of the inside-outside nature with the container, and the necessity of the occurrence of the orientation word and its choice will also vary depending on the definition of the container boundary or the representativeness of the container. From the characteristic of the word class, the azimuth word category of the ' face ' spatial mark in Chinese is larger than that of the ' body ' spatial mark azimuth word category, so that the ' face ' replaces the ' body ' in language and can be expressed as ' up ' replaces ' in language. The Chinese direction word "upper" has two semantics of [ + contact ] and [ + attachment ], and when the contact occurs on the background of the container, the contact does not necessarily occur in a plane, but the attention is focused on a certain part of the container, so that the impression of the background, namely the plane, is generated. This can also be used to interpret statements that take a "line" or "point" as a point of attention, but still use the term "up" in azimuth.

Through the combing of static accommodation azimuth relationship body recognition in French and Chinese, the inventor analyzes the recognition cause behind the language expression difference of the entity and the image pattern. On this basis, the difference in two languages of the Fabry-Perot is aimed at for the accommodation space relation.

Firstly, performing lead design on a database, and determining the category applicable to a real entity and an abstract entity in a containing azimuth structure 'dans+background object'; the corresponding part of the categories corresponding to the Chinese accommodation azimuth structure 'background object + inner' or 'background object + upper' is found.

First, the inventors refer to an informationized French Treasury (TLFI) database, a French assistant online dictionary, and a French modern grammar to determine the spatial context category to which the accommodation azimuth "dans+background scene" constructs apply. In Chinese, the difference of attention to the container space body and the face is included between the azimuth word structure of 'background object + inner' and 'background object + upper', the former focuses on the closed space, and the latter is relatively open in space.

3.5 data analysis and discussion

Because the test study test comprises three stages of French learners, the data collection of the inventor is firstly measured item by item according to the learning depth, and then summarized and summarized according to the difference of the spatial background types. This also provides powerful data for the inventors to examine whether the learning depth has an impact on the spatial background metaphor.

The greater the difference in spatial background structure cognition between the two languages, the higher the error rate of the tested judgment. The following questions are given as examples:

(12)Est-ce que tu vas faire du ski()les montagnes

In the example sentence (12), the correct accommodation azimuth is the expression "dans+background object". The wrong choice is focused on the "sur+background" formulation, which the inventors consider to be influenced by the conceptual diagram of the "background+up" azimuth formulation in the chinese sentence "do you go to ski on mountain". In french cognition, montane (mountain) has spatial multidimensional property, and although skiing occurs on the surface of the mountain as a movement event, a skier needs to enter the mountain to meet the condition of the movement event, so that the limitation of a container on a target object is highlighted, and a preposition of 'dans+background object' rather than a preposition of 'sur+background object' is selected. However, what is expressed in chinese language is the supporting effect provided by mountain for skiing, focusing on the upward supporting force exerted by the container in contact with the object. Therefore, the word structure of the "background object+upper" orientation is more suitable. If the concept cognition process is projected into French under the influence of the Chinese structural formula, the wrong judgment is further made.

The application takes two languages of the learning method and the Chinese as investigation objects, and can exert influence factors behind the language expression difference shown by the two languages when the background of the accommodation space of the reality and the abstract entity is identified. Starting from the theoretical angle, under the guidance of a body recognition linguistic thought, the recognition difference of the background of the accommodating space is mainly limited by two factors, on one hand, the entity serving as the background has limited spatial attribute, and under the condition that the accommodating characteristics are not typical enough, different choices of the structural formula use are easily caused; on the other hand, the cognition of space is greatly influenced by the image patterns, and especially when an abstract entity serves as a background of the accommodation space, cross substitution among different patterns is possible for the difference among languages. Based on the method, the method collects data discovery about the problem of structure selection bias possibly generated when two languages of the Fabry-Perot are used for identifying the background of the accommodation space of the real entity and the abstract entity, the identification bias rate of the background of the space of the abstract entity is obviously higher than that of the background of the space of the real entity, and the bias mainly occurs when the dimension of the real entity is changed, the representativeness of the abstract entity container is not high, and the like. Meanwhile, the test also shows that the spatial concept recognition of the medium-law language can influence the computer to a certain extent, in particular to abstract spatial category. However, with the deepening of the learning depth of the foreign language of the computer, training samples are increased continuously, and particularly, the word sequence problem in the Chinese-law simultaneous translation is improved.

And downloading basic translation data packets and preference data packets from the background to the voice semantic device by the translator according to the use frequency and industry, wherein a three-level database can be designed to realize intelligent recognition of quick and accurate translation and unusual sentences and realize background correction. A first-level database for pre-storing translation sentences; the second-level database stores and classifies the prepositions by the entity storage and classification; according to the stored similar phrases, directly calling and replacing similar entities; the third-level database is used for explaining and classifying the entity description; classifying the relation description, reconstructing according to the grammar of the translation language, outputting and reporting to a background for manual detection.

Referring to fig. 1-15, as shown in fig. 1, comprising a housing part 1; a main board assembly 2 is arranged on the shell piece 1; the shell member 1 is buckled with an upper cover member 3; the main board of the main board assembly 2 is provided with a process notch 4; the main board of the main board assembly 2 is provided with a process notch 4; a processor is arranged on the main board; the processor is electrically connected with a voice input module, a wireless transceiver module, a power supply, a voice output module and a memory; the memory is used for storing data, the database comprises a Chinese database and a French database, and the Chinese database and the French database are linked through the relation in the model database; the French database establishes a category applicable to a real entity and an abstract entity in the received azimuth constitutive dans+ background scene; find out the background object + inner or the background object + upper of the category and the Chinese database receiving direction.

The French database is used for determining a space background category suitable for accommodating the azimuth structure based on an informationized French treasury database, a French assistant online dictionary and/or French modern grammar; in a Chinese database, based on a Xinhua dictionary, a modern Chinese grammar and/or a thesaurus construction, when a target object is in a closed space of a background object model container, calling a position word structure background object+ and when the background object model container has openness relative to the target object, calling the position word structure background object+ on the position word structure background object; the main board assembly 2 is transported through a production line of the mesoscopic sentence-based french language semantic recognition device.

The production line comprises a carrier 17 and a loading conveying group 5 for supporting the carrier 17 and provided with a mould adjusting station 6, an opening station placing 7 and/or a detection adjusting station 8; the output end of the feeding conveying group 5 is connected with a turning station 10 of the circulating conveying assembly 9; a take-out station 12 is provided on the endless conveyor assembly 9; the extraction station 12 is connected with an assembly line 15 through a procedure of a transfer device 13; a plurality of clamping manipulators 14 are distributed on the assembly line 15.

An L-shaped lower bracket 16 is arranged at each station of the feeding conveying group 5 in a lifting manner and is used for blocking a carrier 17 carrying a workpiece from advancing;

The carrier 17 comprises two end clamps 19; a middle connecting portion 18 is connected to both end clamping portions 19; a lateral adjusting clamping plate 20 is oppositely arranged on the end clamping part 19, and the lateral adjusting clamping plate 20 is used for clamping the main board; a spring return clamping assembly 21 which returns by a spring is swung on the end clamping portion 19; the spring reset clamping assembly 21 comprises a rotating shaft arranged on the end clamping part 19, and a reset lower swinging plate 25, a pressing plate part 22 and an external force lower pressing plate 23 which are connected in a Y-shaped manner are hinged on the rotating shaft; the front guide plate 24 is provided at one time on the external force lower pressure plate 23; a plurality of isolation supporting frames 26 are arranged on the end clamping part 19;

the pressing plate part 22 and the reset lower swinging plate 25 are arranged in an electromagnetic attraction manner relative to the end clamping part 19, and the pressing plate part 22 is used for pressing down the main board;

at the mould adjusting station 6, there are side baffles for adjusting the lateral position of the carrier 17;

a guide bracket 27 with a guide inclined surface 28 which is higher at the rear and lower at the front is arranged in the opening station 7;

a pressing top column 29 is elastically provided in front of the guiding lower inclined surface 28 for pressing the external force lower pressing plate 23 so that the pressing plate portion 22 swings upward to open for placement of the main plate;

a detection support frame 30 is arranged at the detection adjustment station 8, and a rotary swinging frame 31 is transversely rotated at the front end of the detection support frame 30 and is used for corresponding to the process notch 4;

An upper top sensor 32 and a lower contact sensor 33 are arranged at the rear end of the detection support frame 30;

a knife-type toggle plate 34 is arranged on the assembly line 15; the assembly line 15 is a strip of intermediate gap channels 35;

at the turning station 10, an inclined support plate 39 positioned at the middle gap passage 35 is inclined, the knife-type toggle plate 34 comprises a support frame 36 arranged on the assembly line 15, and an up-and-down swinging frame 37 reset by a spring is arranged on the support frame 36; the knife-type toggle plate 34 is used for changing the carrier 17 to forward for conveying; a lower gap portion 38 is provided between the upper and lower swing frames 37 and the assembly line 15;

at the take-out station 12, an opening manipulator 11 is provided for taking out the main board from the carrier 17 in the open state;

in the opening manipulator 11, a multi-axis controlled manipulator support 40 is provided, and a rotating shaft portion 41 is laterally provided on an end portion U-shaped frame of the manipulator support 40; the root parts of the swinging plates 42 are arranged at the two ends of the rotating shaft part 41; a horizontal lower pressure plate 43 is provided at an end of the swing plate 42.

The method comprises the following steps of SA, placing empty carriers 17 on a feeding and conveying group 5 and conveying, and blocking the carriers 17 carrying workpieces from moving forward in each station by lifting through an L-shaped lower bracket 16;

SA1, adjusting the positioning of a carrier 17 at a mould adjusting station 6 through a corresponding manipulator and a sensor;

SA2, opening a station and putting 7; firstly, under the downward pressing of a guiding lower inclined surface 28 through a guiding bracket 27 with a high back and a low front, the front guiding plate 24 and the external force lower pressing plate 23 are gradually contacted, the external force lower pressing plate 23 is pressed down through a lower pressing top post 29 against the spring force, and the pressing plate part 22 swings upwards to be opened; then, the main board is attached to the intermediate connection portion 18 by a robot;

SA3, opening a detection adjustment station 8; the main board is provided with a process notch 4, and the rotary swinging frame 31 is arranged in a rotary way;

when the process notch 4 does not correspond to the rotary swinging frame 31, the fan plate of the rotary swinging frame 31 presses down or lifts up the main board, so that the rear part of the carrier 17 is tilted up or swung down, and is contacted with the corresponding upper top sensor 32 or lower contact sensor 33, and the corresponding manipulator clamps the carrier 17 to leave;

when the process gap 4 corresponds to the rotary swing frame 31, the carrier 17 is not contacted with the upper top sensor 32 or the lower contact sensor 33;

SB, the carrier 17 for conveying the output of the feeding conveying group 5 in a direction changing manner through the circulating conveying assembly 9; the circulating conveying assembly 9 outputs the empty carrier 17 after passing through the taking-out station 12, and waits for being sent to the feeding conveying group 5 for the second time after taking out;

SB1, at the turning station 10, firstly, the carrier 17 is sent to the inclined support plate 39 and slides down below the up-down swing frame 37, when the carrier 17 falls onto the up-down swing frame 37, falls down by its own weight, the up-down swing frame 37 swings down, and falls down to the lower gap portion 38; then, the carrier 17 is driven to move forward by the up-and-down swinging frame 37;

SB2, at the take-out station 12, first, the position of the robot support 40 is adjusted by the robot according to the position of the carrier 17; then, the rotating shaft part 41 rotates, so that the Y-shaped finger swing formed by the swinging plate 42 and the transverse lower pressing plate 43 swings and toggles the reset lower swinging plate 25 or the external force lower pressing plate 23;

SC, the main board is sent to the assembly line 15 by the gripping robot 14 to be placed into the case member 1.

The method is carried on the structures such as portable equipment, fixed equipment and the like, and is used for modifying equipment, and the basic structure comprises a shell part 1, a main board assembly 2 and an upper cover part 3; in order to improve the recognition of the main board and reduce the misloading, a process notch 4 is designed, thereby being an invention point. From another thought, the invention can protect a device for improving the inter-interpretation semantic recognition of the Chinese law, and also has the functions of automatic recognition and self recognition, thereby preventing misloading.

As extension protection, after assembly, each conventional module can be read, written and detected and updated regularly, program setting is carried out on a processor, a database is prestored in a memory, a loading and conveying group 5 realizes conveying of carriers, a mould adjusting station 6 realizes positioning and avoids deflection, so that a main board can be correctly placed, an opening station is put into 7 to realize active main board placement, a detection and adjusting station 8 realizes detection and adjustment, misplacement is avoided, and a circulating conveying assembly 9 realizes circulation of the carriers. The direction changing station 10 realizes the adjustment of the production line, avoids the overlong production line, opens the manipulator 11 and the transfer device 13 at the taking-out station 12,

And the clamping manipulator 14 realizes the taking of the main board. The assembly line 15 is used as a main line, and realizes assembly, detection and other parts of main parts. The horizontal accommodation and the vertical propelling movement have been realized through L type lower carriage 16, the transport of mainboard spare has also been realized to carrier 17, can also be the transport of casing or lid, well connecting portion 18 is the support connection, tip clamping part 19 is two symmetry settings, lateral adjustment cardboard 20 realizes the side direction joint, spring return clamping assembly 21 realizes the centre gripping to mainboard upper portion or lateral part (can realize), it has rubber pad or other flexible pads to press board portion 22, external force holding down plate 23 conveniently pushes down, the preceding pressure of waiting of front guide board 24 realization assembly line forward, swing board 25 resets, it opens to have realized the top realization, the commonality is strong, it is preferred, press and hold board portion through spring return. The isolating support frame 26, the guide support frame 27, the guide lower inclined surface 28, the push-down jacking column 29, the detection support frame 30, the rotary swinging frame 31, the upper jacking sensor 32, the lower contact sensor 33 realize dislocation detection and alarm, the knife-type toggle plate 34, the middle clearance channel 35 improves the process rationality, the support frame 36 is not limited by the attached drawing, the L-shaped baffle is improved, the upper swinging frame 37, the lower clearance part 38 and the inclined support plate 39 are improved, thereby realizing the pushing of the manufacturability and the compatibility, and the manipulator support 40, the rotating shaft part 41, the swinging plate 42 and the transverse lower pressing plate 43 are simultaneously realized, so that the automatic opening is realized, and the opening degree is large.

The above embodiments are not intended to limit the scope of the present application, so: all equivalent changes in structure, shape and principle of the application should be covered in the scope of protection of the application.

Claims

1. A Chinese and French voice semantic recognition method based on preposition sentences is characterized by comprising the following steps of: when detecting that the input native language contains prepositional sentences, executing the following steps of the method;

firstly, analyzing preposition sentences into target objects and background objects according to native language grammar, wherein the background objects comprise entities serving as reference objects or background objects, have spatial characteristics and can directly provide spatial references for the target objects;

when the background object is judged to be a real entity, modeling is directly carried out according to the corresponding object of the mother language of the background object, and the modeling is stored in a database or an existing model prestored in the database is called; when the background object is judged to be an abstract entity, metaphor modeling cognition is carried out, and modeling is stored in a database or an existing model prestored in the database is called; the target object and the background object construct a background construction formula; when the background object of the abstract entity is considered to be accommodating in the French, the abstract entity is correspondingly regarded as a container characteristic; when the background object of the abstract entity is Chinese and is not considered to have container characteristics, and French is considered to have accommodative property, introducing an expert database, in French, correspondingly taking the abstract entity as container characteristics, and establishing three-dimensional space dimension of the abstract entity so as to form a background object by entering sentences;

When the preposition phrase adopts a static space azimuth relation, setting a French constitution formula to adopt prepositions and background objects, and setting a Chinese constitution formula to adopt prepositions, background objects and azimuth words; when the background object and the target object are identified as the bearing relation of the accommodation space, the bearing object is set as prepositions dans+the background object in French, and the bearing object is set as prepositions+the background object in Chinese; in the process of analyzing the relationship, aiming at the modeled model, enhancing the feature part of the relationship and the object, regarding the feature part as the feature of the whole entity, and discarding the other features or dimensions;

when the Chinese is translated into French, if the model relation between the background object and the target object is an accommodation space azimuth relation, invoking preposition dans; otherwise, invoking preposition sur; when translating the background object into Chinese language, if the model relation between the background object and the target object is the accommodation space azimuth relation, calling the azimuth word or the azimuth word, and outputting the translated text after the target object and the background object are translated into translated text semantics.

2. The preposition-based medium french language semantic recognition method according to claim 1, wherein: the system comprises a voice input module, a semantic analysis modeling module and a text processing module, wherein the voice input module converts the native language voice into characters and then carries out semantic analysis modeling and/or directly carries out semantic analysis modeling on the native language voice;

3. The preposition-based medium french language semantic recognition method according to claim 1, wherein: wherein the background object is static and/or dynamic relative to the object;

the context includes a real entity and/or an abstract entity;

the background object has concave-convex degree and/or flatness;

4. The preposition-based medium french language semantic recognition method according to claim 1, wherein: when the relation of the object in the background object is translated into French in Chinese, when the object is positioned at the high position of the background object and is in direct or indirect contact with the background in the model relation, the relation is smaller than the background object, and the gravity of the object is opposite to the relative motion direction between the object and the background, the preposition sur is called; when the boundary part or the whole of the target object is contained by the background object, the preposition dans is called;

when the background object is modeled as a plane, invoking sur;

when translating into Chinese in French, the relation model of the background object and the target object comprises a protection type, a limiting type, a position limiting type, a sight limiting type and a related type;

when the object is of a protection type or a limiting type, the object is placed in a background object, is wrapped by the background object and is subjected to constraint force in an inward direction, and a corresponding Chinese translation sentence is intersected with a dynamic image pattern on the basis of recognizing and analyzing a container pattern, so that attention is focused on a relation between a position limiting feature and an application force, namely, a source direction of the constraint force applied to the content by the container or a specific position and a direction of the force generation on the application force surface can be determined on the azimuth word;

in Chinese, when the model is considered as one-dimensional and two-dimensional surface call, the three-dimensional model is considered as the body call; the method is matched with and updated with a Chinese database and a French database, and the Chinese database and the French database are linked through the relation in a model database; the French database establishes a category applicable to a real entity and an abstract entity in the received azimuth constitutive dans+ background scene; finding out real entities or abstract entities which are matched with the background object + in or on the background object + in the Chinese database containing azimuth structure in the categories;

5. The Chinese and French voice semantic recognition device based on preposition sentences is characterized in that: the medium french-language semantic recognition device for implementing the method according to any one of the preceding claims 1 to 4, the medium french-language semantic recognition device comprising a housing part (1); a main board assembly (2) is arranged on the shell piece (1); the shell piece (1) is buckled with an upper cover piece (3); a main board of the main board assembly (2) is provided with a process notch (4); the method is characterized in that: a main board of the main board assembly (2) is provided with a process notch (4); a processor is arranged on the main board; the processor is electrically connected with a voice input module, a wireless transceiver module, a power supply, a voice output module and a memory; the memory is used for storing data, the database comprises a Chinese database and a French database, and the Chinese database and the French database are linked through the relation in the model database; the French database establishes a category applicable to a real entity and an abstract entity in the received azimuth constitutive dans+ background scene; find out the background object + inner or the background object + upper of the category and the Chinese database receiving direction.

6. The preposition-based medium french language semantic recognition apparatus of claim 5, wherein: the French database is used for determining a space background category suitable for accommodating the azimuth structure based on an informationized French treasury database, a French assistant online dictionary and/or French modern grammar; in a Chinese database, based on a Xinhua dictionary, a modern Chinese grammar and/or a thesaurus construction, when a target object is in a closed space of a background object model container, calling a position word structure background object+ and when the background object model container has openness relative to the target object, calling the position word structure background object+ on the position word structure background object; the main board component (2) is transmitted through a production line of the Chinese and French semantic recognition device based on preposition sentences.

7. A production method of Chinese and French voice semantic recognition equipment based on preposition sentences is characterized by comprising the following steps of: the method for realizing the medium French semantic recognition device according to any one of the claims 1 to 4 comprises the following steps of SA, placing empty carriers (17) on a feeding conveying group (5) and conveying the carriers, and lifting the carriers (17) carrying the workpieces at each station through an L-shaped lower bracket (16) to block the carriers from advancing;

SA1, in a mould adjusting station (6), adjusting the positioning of a carrier (17) through a corresponding manipulator and a sensor;

SA2, opening a station and putting in (7); firstly, under the downward pressing of a guiding downward inclined surface (28), the guiding downward inclined surface is gradually contacted with an external force pressing plate (23) through a front guiding plate (24), the external force pressing plate (23) is pressed down through a pressing top column (29) against the spring force, and the pressing plate part (22) swings upwards to be opened; then, the main board is connected to the middle connecting part (18) through a mechanical arm;

SA3, opening a detection adjustment station (8); the main board is provided with a process notch (4), and the rotary swinging frame (31) is arranged in a rotary way;

when the process notch (4) is not corresponding to the rotary swinging frame (31), the fan plate of the rotary swinging frame (31) presses down or lifts up the main board, so that the rear part of the carrier (17) is tilted up or swung down, and is contacted with the corresponding upper top sensor (32) or lower contact sensor (33), and the corresponding manipulator clamps the carrier (17) to leave; when the process gap (4) corresponds to the rotary swinging frame (31), the carrier (17) is not contacted with the upper top sensor (32) or the lower contact sensor (33);

SB, the carrier (17) for conveying the output of the feeding conveying group (5) in a turning way through the circulating conveying component (9); the circulating conveying assembly (9) outputs the empty carrier (17) after passing through the taking-out station (12), and waits for being sent to the feeding conveying group (5) for the second time after being taken out;

SB1, at the turning station (10), firstly, the carrier (17) is sent to the inclined supporting plate (39) and slides downwards to the lower part of the upper and lower swinging frame (37), when the carrier (17) falls on the upper and lower swinging frame (37), the upper and lower swinging frame (37) swings downwards by self weight and falls to the lower gap part (38); then, the carrier (17) is driven to move forwards by the up-and-down swinging frame (37);

SB2, in the take-out station (12), firstly, according to the position of the carrier (17), the position of the manipulator support (40) is adjusted by the manipulator; then, the rotating shaft part (41) rotates, so that a Y-shaped finger formed by the swinging plate (42) and the transverse lower pressing plate (43) swings to toggle the reset lower swinging plate (25) or the external force lower pressing plate (23); SC, the main board is sent to an assembly line (15) by a clamping manipulator (14) to be placed into the housing part (1).