A method for identifying and extracting ideographic components where a bilingual sentence database is generated after sentence alignment is performed on corpora of multilingual or bilingual text versions of same content, which is characterized in that different language texts are further associated by using ideographs according to a fact that sentences of different language texts may express same semantic content, ideographs of different language texts are implemented by using four types of common ideographic components, and the four types of ideographic components are identified and extracted as follows;
- S1. reading a sentence and matching a frame (102);
- a sentence includes two parts:
- a sentence frame and a sentence cabin;
- in each round of identifying and extracting ideographic components, texts of languages A and B are selected;
- a machine reads a bilingual sentence pair of languages A and B from a bilingual sentence database, inquires a sentence frame database by using the current bilingual sentence pair to perform a sentence frame match operation, integrates the bilingual sentence pair into a matched sentence frame, and displays “Cabin Detection” and “Search, Match, and Label” buttons;
- after integration, if each sentence cabin is even and complete without stacking, and the “Cabin Detection” button is clicked, it is positive, and a matched sentence frame exists, and then S3 cabin detection is performed;
- after integration, if each sentence cabin is uneven or incomplete or stacking exists, and the “Search, Match, and Label” button is clicked, it is negative, which indicates that an integration result is negated and that a new sentence frame needs to be extracted;
- then the machine performs a “search, match, and label” operation, and uses a word string of the sentence of language A to inquire a conventional electronic dictionary and generate a semantic match table for the current sentence pair according to whether the sentence of language B includes a word string corresponding to semantic content;
- then the cabin count N=0, and a preparation is made for extracting a new sentence frame;
- S2. identifying and labeling sentence cabins (103);
- the machine simultaneously identifies bilingual quantity strings, proper noun strings, and article noun strings sequentially, and pre-labels them as sentence cabins according to a time sequence by using N=N+1;
- by pre-labeling sentence cabins, a frame example percentage is decreased from 100% gradually;
- after the three types of word strings are labeled as sentence cabins, if the frame example percentage is higher than 50%, the machine continues to label zero-article noun strings or other semantic match word strings one by one as sentence cabins according to the semantic match table under a prerequisite that spacing exists, accepts a user's approval or modification or addition of a semantic match word string by using a sense group alignment operation, and continues to identify and label sentence cabins, until the conditions for identifying and labeling sentence cabins are insufficient and the frame example percentage is between 15% and 50%;
- then the operation of identifying and labeling sentence cabins ends;
- the machine checks the sentence frame format and collates the foregoing pre-labeled sentence cabins as formally labeled sentence cabins;
- this is a step of identification and extraction in the first round;
- parts of speech do not need to be identified in identification and extraction starting from the second round;
- instead, labeled language sentences are used as a template to identify and label sentence cabins of unlabeled language sentences;
- when N=>1, residual pans after sentence cabins are removed from the current sentence pair are sentence frame components, and they are stored in corresponding language component fields in a sentence frame database (105);
- when N=0, the current sentence pair is an idiom component;
- idiom components are some special sentence pairs that cannot be divided into sentence frames and sentence cabins, and the idiom components are stored in corresponding language idiom fields in an idiom database (104);
- the current bilingual sentence example is integrated back into the current sentence frame strictly according to cabin numbers, so that it is in a frame example form;
- and then the machine goes on to the next step;
- S3. detecting cabins and extracting a cabin model (106);
- the machine detects sentence cabins one by one;
- in the first round of identification and extraction, the machine counts word strings included in sentence cabins of a sentence of a Pinyin text of language A;
- if the quantity of word strings included in a sentence cabin is greater than or equal to five original word strings, the machine needs to perform an operation of composing complex words;
- if the quantity of word strings included in a sentence cabin is greater than or equal to eight original word strings, the machine performs a “search, match, and label” operation by using the current sentence cabin, and generates a semantic match table for the current sentence cabin, so that a cabin model is further extracted as a sentence cabin with a model;
- the machine extracts the cabin model, and simultaneously identifies bilingual quantity strings, proper noun strings, and article noun strings sequentially, and pre-labels them as cabin eyes;
- a model example percentage is decreased from 100% gradually;
- after pre-labeling is completed, if the quantity of word strings of the three types is still greater than 70%, the machine labels zero-article noun strings or other semantic match word strings as cabin eyes under a prerequisite that spacing exists, accepts the's approval or modification or addition of a semantic match word string by using a sense group alignment operation, and continues to pre-label cabin eyes;
- when conditions for pre-labeling do not exist and the model example percentage is between 50% and 70%, the operation of pre-labeling cabin eyes ends;
- the pre-labeled cabin eyes are modified as formally labeled according to a left-to-right order of English sentence cabins;
- differently, in extraction starting from the second round, labeled language sentence cabins are used as a template to identify and label unlabeled language cabin eyes;
- when the cabin eye quantity N is greater than or equal to 1, residual parts after cabin eyes are removed from the current sentence cabin are cabin model components, and they are saved to corresponding language cabin model fields in a cabin model database (107);
- and then the machine continues to detect other sentence cabins, until detection of sentence cabins of the whole sentence pair is completed;
- S4. receiving and storing sense group strings (108);
- the machine integrates the current bilingual example back into the currently extracted sentence frame strictly according to cabin numbers, integrates sentence cabins with cabin models back into the corresponding cabin models, and then searches for the sentence cabins or cabin eyes one by one;
- with reference to the semantic match table, if word fields of language A and word fields of language B are not empty in records of the match table of the corresponding content and semantic content of texts of languages A and B is aligned, the content is determined to be sense group string components, and fetched from the records in succession, and respectively stored in corresponding language fields in a same record in a sense group database (109);
- and then the operation of identifying and extracting the current sentence pair ends.