CN115114929A - Quantitative attribute comparison sentence understanding method, equipment and storage medium - Google Patents
Quantitative attribute comparison sentence understanding method, equipment and storage medium Download PDFInfo
- Publication number
- CN115114929A CN115114929A CN202210515296.5A CN202210515296A CN115114929A CN 115114929 A CN115114929 A CN 115114929A CN 202210515296 A CN202210515296 A CN 202210515296A CN 115114929 A CN115114929 A CN 115114929A
- Authority
- CN
- China
- Prior art keywords
- sentence
- fine
- comparison
- quantitative
- sentences
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 230000014509 gene expression Effects 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 11
- 230000009191 jumping Effects 0.000 claims description 19
- 230000011218 segmentation Effects 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 6
- 230000000052 comparative effect Effects 0.000 description 7
- 230000002776 aggregation Effects 0.000 description 6
- 238000004220 aggregation Methods 0.000 description 6
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 239000003607 modifier Substances 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000003058 natural language processing Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
Abstract
A quantitative attribute comparison sentence understanding method, equipment and storage medium comprises an input preprocessing step, a sentence recognition step, an argument extraction step and an output generation step. Firstly, the input preprocessing step can delete the contents irrelevant to sentence understanding in the sentence, and the sentence is analyzed by lexical and syntactic methods; then, in the sentence identification step, a trigger word, a regular expression and a quantity type attribute word list are used for identifying the sentence; secondly, extracting arguments from the sentence according to the sentence mode and the syntactic structure; and finally, in the output generation step, the time and place information in the sentence is identified, and the identified argument is standardized to obtain the structured data. The invention provides a method for effectively understanding a quantitative attribute comparison sentence, which is used for understanding the quantitative attribute comparison sentence and generating computer-understandable structured data. The invention is easy to deploy and has strong practicability.
Description
Technical Field
The invention belongs to the technical field of computer natural language processing, relates to the field of understanding of sentences related to quantitative attribute comparison by using an expert system, in particular to a quantitative attribute comparison-oriented sentence understanding method, which can be applied to a question-answering system construction process related to quantitative attribute comparison.
Background
With the advent of the information age and the development of artificial intelligence technology, the life style of people tends to be more convenient and intelligent. Researchers are increasingly concerned about how to make machines quickly and accurately understand and answer questions described in natural language by users, greatly driving the development of automated question and answer technology. In the question-answering system, understanding of the question semantics is a precondition for making correct answers, and the question-answering system may return correct answers only if information contained in the question is sufficiently extracted.
In some fields, especially in question-answering systems in professional fields, questions related to quantitative comparison of quantitative attributes such as "air temperature", "birth value", "birth rate", etc., that is, automatic answering of questions related to quantitative attribute comparison still faces many challenges. The existing automatic question-answering technology based on retrieval and neural network is difficult to fully understand the quantitative type attribute comparison sentences, extract important semantic information from the sentences, achieve better solving effect and has limitation on processing the problems. On one hand, the types of the quantitative attributes are various, the natural language expression forms of the same quantitative attributes are flexible and various, and the aggregation operation on the quantitative attributes can be involved in the sentences. The existing question-answering system is difficult to effectively understand the quantitative attributes in the questions and the related aggregation operation, for example, the problems that the daily average temperature, the daily maximum temperature and the annual poor temperature are different quantitative attributes are difficult to understand; the numerical attributes of the daily temperature difference, the daily change of the air temperature and the like are difficult to understand, namely the daily temperature difference and the daily change of the air temperature are both poor; it is difficult to understand that the birth rate increase, the birth rate increase amplitude and the birth rate increase amplitude need to calculate the increase amplitude of the corresponding birth rate in the given time interval in the sentence. On the other hand, the quantitative attribute comparison problem has a complex semantic structure, the semantic information contained in different sentence structures is different, and a plurality of entities and attributes are often involved in a sentence and are provided with modifiers. The existing question-answering technology is difficult to fully understand semantic information contained in sentences, for example, text forms of 'the birth rate of the population A is lower than that of the population B', 'the birth rate of the population A is always lower than that of the population B' and 'the birth rate of the population A is increased to be lower than that of the population B' are similar, but the meanings of expressions are greatly different.
The automatic solving process of the quantitative attribute comparison problem is very dependent on the understanding of the quantitative attribute comparison sentences, and the quantitative attributes in the sentences need to be analyzed, the comparison semantics of the quantitative attributes are understood, and the quantitative attributes are organized into computer-understandable structured data.
Disclosure of Invention
The invention aims to solve the problems that: the conventional question-answering system is difficult to fully understand the quantitative attribute comparison sentences, and has limitation on solving the problems. On one hand, the existing question-answering system is difficult to effectively understand the quantitative attribute and the related aggregation operation in the quantitative attribute comparison sentence; on the other hand, the conventional question-answering system is difficult to fully understand the comparison semantics of the quantitative attribute comparison class sentences. Therefore, the invention introduces expert rules to understand the quantitative attribute comparison sentences, and analyzes the quantitative attribute comparison sentences in the natural language form into the comparative assertions which can be understood by the computer.
The technical scheme of the invention is as follows: a quantitative attribute comparison type sentence understanding method generates computer-understandable structured data for sentences containing contents quantitatively compared with quantitative attributes, and comprises the following steps:
input preprocessing, namely deleting contents irrelevant to sentence understanding in a sentence, and performing lexical and syntactic analysis on the sentence to obtain a word segmentation result, a part of speech tagging result and a syntactic tree of the sentence;
sentence recognition, namely recognizing the preprocessed sentences by using trigger words, regular expressions and a quantitative attribute word list to obtain fine-grained categories of quantitative attribute comparison sentences;
extracting argument, namely extracting argument from a sentence according to the sentence recognition result and the syntactic structure;
and (4) output generation, namely recognizing time and place information in the sentence, and standardizing the recognized argument to obtain structured data.
Further, the input preprocessing step specifically includes:
step 100, determining a processing object as a sentence;
step 101, deleting the receiving and prompting words in the sentence, and removing the contents irrelevant to sentence understanding;
102, segmenting words of the sentence;
103, performing part-of-speech tagging on a result obtained after sentence segmentation;
and 104, performing syntactic analysis on the sentence based on the word segmentation and part-of-speech tagging results.
Further, the sentence recognition step recognizes the preprocessed sentence, specifically:
200, constructing a quantitative attribute word list according to expert rules;
step 201, setting fine-grained categories recognized by quantitative attribute comparison sentences, including comparison-level sentences, highest-level sentences and variation-level sentences;
202, designing a trigger word list for each fine-grained category;
step 203, designing a trigger word rule for each fine-grained category;
step 204, traversing each fine-grained category of the preprocessed sentences;
step 205, judging whether the traversal is finished, if so, jumping to step 211, otherwise, jumping to step 206;
step 206, judging the sentence according to the trigger word list of the category and the trigger rule;
step 207, judging whether the sentence belongs to the category, if so, skipping to step 208, otherwise, skipping to step 204;
step 208, further judging according to the quantity type attribute word list;
step 209, judging whether the sentence is a quantitative attribute sentence, if so, jumping to step 210, otherwise, jumping to step 204;
step 210, returning the fine-grained type of the current execution process as an identification result;
step 211, return other as recognition results.
Furthermore, the argument extraction is specifically as follows according to the fine-grained category of the sentence, the position of the trigger word and the syntax tree extraction semantic template definition:
step 300, designing a semantic template for each fine-grained category, wherein the template comprises arguments to be extracted;
step 301, summarizing sentence patterns of each fine-grained category;
step 302, according to the sentence recognition result, acquiring a sentence and a fine-grained category thereof, and the position of a trigger word corresponding to the fine-grained category;
step 303, identifying and obtaining the position of an auxiliary word extracted from an auxiliary argument in a sentence;
step 304, extracting argument according to the positions of the trigger words and the auxiliary words;
step 305, judging whether all arguments to be extracted are identified, if so, jumping to step 307, otherwise, jumping to step 306;
step 306, extracting argument by using a syntactic structure rule;
and 307, returning extraction results of all arguments to be extracted.
Further, the output generating step specifically includes:
step 400, identifying time in a sentence;
step 401, identifying a place in a sentence;
step 402, standardizing the expression of word meaning of the argument and time and place extracted from the sentence;
step 403, organizing the normalized result into structured data as a sentence analysis result;
step 404, structured data is returned.
The invention also provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the computer program realizes the quantitative attribute comparison sentence understanding method when executed.
The present invention also provides a computer-readable storage medium storing a computer program for executing the above-described quantitative attribute comparison-based sentence comprehension method.
The method is suitable for the understanding process of the quantitative attribute comparison sentences, and the quantitative attribute comparison sentences are identified and semantically analyzed by using expert knowledge and rules to organize into computer-understandable structured data.
The invention has the beneficial effects that: the invention provides a method for effectively understanding a quantitative attribute comparison sentence. On one hand, the quantitative attribute and the related aggregation operation in the sentence are fully understood, the numerical attribute and the similar expression of the aggregation operation are mapped to the standard expression, and the problem that the quantitative attribute and the aggregation operation are various in expression and difficult to understand is effectively solved; on the other hand, the numerical attribute comparison sentences are classified in a fine-grained manner, semantic information of each fine-grained type sentence is fully extracted, and the semantics of the sentences are restored as far as possible. On the basis of fully understanding the quantitative attribute comparison sentences, the method can effectively improve the accuracy of the final solution of the quantitative attribute comparison problems in the question-answering system. In addition, the invention is easy to deploy and has strong interpretability.
Drawings
FIG. 1 is a diagram of a main body frame according to an embodiment of the present invention.
FIG. 2 is a flowchart of input preprocessing steps according to an embodiment of the present invention.
FIG. 3 is a flowchart of sentence recognition steps according to an embodiment of the invention.
FIG. 4 is a flowchart of argument extraction steps according to an embodiment of the present invention.
FIG. 5 is a flowchart of an output generation step according to an embodiment of the present invention.
FIG. 6 is a flow chart illustrating understanding of a certain type of sentence with a certain type of attribute comparison according to an embodiment of the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.
The following describes a specific embodiment of a quantitative attribute comparison sentence understanding method, and this embodiment describes a specific process of the invention with a high school geography choice as a scene. The high school geography choice questions often comprise a background, a question stem, four options and a chart, in the invention, the question stem and each option are spliced to obtain a sentence as a processing object without special processing on the question background and the chart. The embodiment organizes the final parsing result into a comparative assertion list, which is a kind of computer-understandable structured data. In the following explanation, the following titles are taken as examples: the term "correctly described below" is selected from "2015 to 2020, the birth rate of the population in the area A is lower than that in the area B".
As shown in FIG. 1, the specific process of the present invention includes an input preprocessing step, a sentence recognition step, an argument extraction step, and an output generation step. The input preprocessing step organizes the question stem and the options in the high school geography choice questions into a sentence form, and obtains the word segmentation result, the part of speech tagging result and the syntax tree of the sentence; in the sentence recognition step, the sentences are divided into three fine-grained categories of a quantitative attribute comparison level, a highest level and a variation level and one of other sentences according to the trigger words and the quantitative attribute word list; in the argument extraction step, arguments defined by the semantic template are extracted according to fine-grained categories of the sentence, positions of the trigger words and the syntax tree, and an argument extraction result is output; in the output generation step, time and place information in the sentence is identified, and the identified arguments are standardized to obtain a comparative assertion list.
As shown in fig. 2 and 6, the input preprocessing steps are sequentially:
step 100, determining a processing object as a sentence;
step 101, deleting the receiving and prompting words in the sentence, and removing the contents irrelevant to sentence understanding;
102, segmenting words of the sentence;
103, performing part-of-speech tagging on a result obtained after sentence segmentation;
and 104, performing syntactic analysis on the sentence based on the word segmentation and part-of-speech tagging results.
In this embodiment, specifically, it is determined that the processing target is a sentence formed by splicing the question stem and the option in the high school geography choice questions (step 100), and each choice question is converted into a correct and incorrect judgment of four sentences, for example, the question stem of the choice question "correctly describe the following sentence" with one option "between 2015 and 2020, the population birth rate of a place is always lower than that of B" splicing the following sentence "between 2015 and 2020, and the population birth rate of a place is always lower than that of B"; in order to simplify the sentence structure, the carrying and the cue in the sentence need to be deleted, and the content irrelevant to the sentence understanding is removed (step 101), for example, the "right is described in the following description" in the above example sentence is deleted, and the sentence "2015 to 2020 year, the population birth rate of the place A is always lower than that of the place B" is obtained; then, an open source tool is used, for example: the standard NLP performs word segmentation (step 102), part-of-speech tagging (step 103), and syntactic analysis (step 104) on a sentence.
As shown in fig. 3 and 6, the sentence recognition steps are sequentially:
200, constructing a quantitative attribute word list according to expert rules;
step 201, setting fine-grained categories recognized by quantitative attribute comparison sentences, including comparison-level sentences, highest-level sentences and variation-level sentences;
202, designing a trigger word list for each fine-grained category;
step 203, designing a trigger word rule for each fine-grained category;
step 204, traversing each fine-grained category of the preprocessed sentences;
step 205, judging whether the traversal is finished, if so, jumping to step 211, otherwise, jumping to step 206;
step 206, judging the sentence according to the trigger word list of the category and the trigger rule;
step 207, judging whether the sentence belongs to the category, if so, skipping to step 208, otherwise, skipping to step 204;
step 208, further judging according to the quantity type attribute word list;
step 209, judging whether the sentence is a quantitative attribute sentence, if so, jumping to step 210, otherwise, jumping to step 204;
step 210, returning the fine-grained type of the current execution process as an identification result;
step 211, returning the other as the recognition result.
In the embodiment, for the high school geography test question, a high school geography quantitative attribute vocabulary is constructed by expert evaluation and listing quantitative attributes therein, such as "temperature", "population number", "population birth rate" and the like (step 200); determining the fine-grained classification recognized by the quantitative attribute comparison sentence, including a comparison stage, a highest stage and a variation sentence (step 201), so as to conveniently perform fine-grained processing on the quantitative attribute sentence; designing a trigger word list and trigger rules for each fine-grained category according to the sentence mode (step 202, step 203), wherein common trigger words in comparison-level sentences comprise 'greater than', 'less than' and the like; traversing each fine-grained category (step 204), judging whether the traversal is finished (step 205), if the three fine-grained categories are not successfully judged after being traversed, identifying the fine-grained categories as other sentences (step 211), otherwise, continuing to process according to the following steps; for each fine-grained category, judging according to a trigger word list and a trigger rule (step 206), judging whether a sentence belongs to the category (step 207), if so, further judging according to a quantitative attribute word list (step 208), otherwise, continuously judging the next fine-grained category (jumping to step 204), for example, in the period from "2015 to 2020," the population birth rate of A is lower than that of B "contains a comparison-level trigger word" lower ", and judging as a comparison-level sentence; when the quantitative attribute vocabulary is further determined (step 208), whether the sentence is a quantitative attribute sentence is determined (step 209), i.e. whether the sentence has the quantitative attribute in the quantitative attribute vocabulary, if so, the fine-grained type of the current execution process is returned as the identification result (step 210), otherwise, the next fine-grained type is continuously determined (step 204), for example, the comparison-level sentence "2015-2020, the population birth rate of the place A is lower than that of the place B" the population birth rate of the quantitative attribute is included ", and the comparison-level sentence is identified as the quantitative attribute comparison-level sentence. If no fine-grained category is identified after the traversal, the others are returned, indicating that the sentence is not of an understandable type (step 211).
As shown in fig. 4 and 6, the argument extraction step extracts arguments defined by the semantic template, specifically:
step 300, designing a semantic template for each fine-grained category, wherein the template comprises arguments to be extracted;
step 301, summarizing sentence patterns of each fine-grained category;
step 302, obtaining a sentence and a fine-grained category thereof and a position of a trigger word corresponding to the fine-grained category according to a sentence recognition result;
step 303, identifying and obtaining the position of an auxiliary word extracted from an auxiliary argument in a sentence;
step 304, extracting argument according to the positions of the trigger words and the auxiliary words;
step 305, judging whether all arguments to be extracted are identified, if so, jumping to step 307, otherwise, jumping to step 306;
step 306, extracting argument by using a syntactic structure rule;
and 307, returning the extraction results of all arguments to be extracted.
In an embodiment, a semantic template is designed for each fine-grained category, including arguments to be extracted (step 300), for example, the semantic template of a comparison-level sentence is a "comparison-level template (comparison subject, comparison object, comparison aspect, comparison result, comparison modifier)", where the comparison subject, the comparison object, the comparison aspect, the comparison result, and the comparison modifier are arguments to be extracted, the comparison subject is an object to be compared, the comparison object is an object to be compared, the comparison aspect is an aspect to be compared, the comparison result represents a result of "different heights" of comparison, and the comparison modifier is a modifier of the comparison result; summarizing sentence patterns of each fine-grained category (step 301), for example, one sentence pattern of the comparison-level sentence is "in" sentence-like "[ comparison subject ] [ comparison aspect ] [ comparison modification ] < more than ]. in > [ comparison object ]"; inputting a result of the sentence recognition step, including a sentence, a fine-grained category, and a trigger word position (step 302); obtaining auxiliary word positions of auxiliary argument extraction (step 303), such as words like "and", "and" in a sentence; then extracting arguments according to the positions of the trigger word and the auxiliary word (step 304), for example, in the sentence "2015-2020, the population birth rate of the place A is always lower than that of the place B" the comparison object "the place B" is extracted according to the position of the trigger word "lower"; judging whether all arguments to be extracted are identified (step 305), if so, directly returning the extraction results of all arguments to be extracted (step 307), otherwise, extracting the arguments by using a syntactic structure rule (step 306), and finally returning the extraction results of all arguments to be extracted (step 307), for example, in the period from the sentence "2015 to 2020, the population birth rate of the place A is always lower than that of the place B," extracting the text "always" corresponding to the AD (adverb) node before the trigger word "lower than" according to the syntactic structure rule as a comparative modification, "extracting the text" place A "corresponding to the first child node of the left sibling node NP (noun phrase) of the trigger word" lower than "ancestor node VP (phrase) as a comparative subject, and extracting the text" population birth rate "corresponding to the rest NP child nodes as a comparative aspect.
As shown in fig. 5 and 6, the output generation step specifically includes:
step 400, identifying time in a sentence;
step 401, identifying a place in a sentence;
step 402, standardizing the expression of word meaning of the argument and time and place extracted from the sentence;
step 403, organizing the normalized result into structured data as a sentence analysis result;
step 404, structured data is returned.
With open source tools, for example: the hundred-degree LAC is used for identifying time and places in a sentence (step 400 and step 401), the time and the places are important for solving the sentence, for example, the time of the sentence is "2015-2020, the population birth rate of the place A is lower than that of the place B" for solving, the time of the sentence is "2015-2020", and the population birth rate of the place A and the population birth rate of the place B in the time range are compared; standardizing arguments and time and places extracted from sentences (step 402), which facilitates subsequent solution, for example, uniformly expressing "birth rate" and "population birth rate" as quantitative attributes in comparison aspects of comparison sentences, uniformly expressing "all the time", "all the time" and "continuously" as "all the time" in comparison and modification of comparison sentences, uniformly expressing "less than", "less than" and "less than" in comparison results of comparison sentences, and uniformly expressing "B place", "B place" and "B city" as "B place"; finally, the parsing result is organized into a comparative assertion list and returned (steps 403 and 404), resulting in structured data.
In the embodiment shown in fig. 6, in the high school geographical choice questions, the input of the input preprocessing step is a question stem, "the following description is correct and" and options "2015-2020, the birth rate of the population in place a is always lower than that in place B", and a sentence is spliced, and the segmentation result, the part-of-speech tagging result and the syntax tree of the sentence are output through the steps shown in fig. 2; in the sentence recognition step, the sentence is judged to be a quantitative attribute comparison-level sentence, and a trigger word 'lower' is output; in the argument extraction step, important arguments in the quantitative attribute comparison-level sentence are identified, wherein the important arguments comprise a comparison subject 'A place', a comparison object 'B place', a comparison aspect 'population birth rate', a comparison result 'lower' and a comparison modification 'constant'; the output generation step identifies time information "2015-2020", normalizes the identified argument and time location, and outputs a comparison type assertion list.
It will be appreciated by those skilled in the art that the steps of the above-described quantitative attribute comparison sentence-like understanding method of the present invention can be implemented by a general purpose computing device, centralized on a single computing device or distributed across a network of computing devices, and implemented by program code executable by the computing device, such that the steps shown and described can be executed by the computing device stored in a memory device and, in some cases, executed out of order, or fabricated separately as individual integrated circuit modules, or fabricated as a single integrated circuit module from multiple modules or steps. Thus, embodiments of the invention are not limited to any specific combination of hardware and software.
Claims (8)
1. A quantitative attribute comparison type sentence understanding method is characterized in that sentences containing contents which are quantitatively compared with quantitative attributes are generated into computer-understandable structured data, and the method comprises the following steps:
input preprocessing, namely deleting contents irrelevant to sentence understanding in a sentence, and performing lexical and syntactic analysis on the sentence to obtain a word segmentation result, a part of speech tagging result and a syntactic tree of the sentence;
sentence recognition, namely recognizing the preprocessed sentences by using trigger words, regular expressions and a quantitative attribute word list to obtain fine-grained categories of quantitative attribute comparison sentences;
extracting argument, namely extracting argument from a sentence according to the sentence recognition result and the syntactic structure;
and (4) output generation, namely recognizing time and place information in the sentence, and standardizing the recognized argument to obtain structured data.
2. The method according to claim 1, wherein the input preprocessing step specifically comprises:
step 100, determining a processing object as a sentence;
step 101, deleting the receiving and prompting words in the sentence, and removing the contents irrelevant to sentence understanding;
102, segmenting words of the sentence;
103, performing part-of-speech tagging on the result after the sentence is participated;
and 104, performing syntactic analysis on the sentence based on the word segmentation and part-of-speech tagging results.
3. The method according to claim 1, wherein the sentence recognition step recognizes the preprocessed sentence, specifically:
200, constructing a quantitative attribute word list according to expert rules;
step 201, setting fine-grained categories recognized by quantitative attribute comparison sentences, including comparison-level sentences, highest-level sentences and variation-level sentences;
202, designing a trigger word list for each fine-grained category;
step 203, designing a trigger word rule for each fine-grained category;
step 204, traversing each fine-grained category of the preprocessed sentences;
step 205, judging whether the traversal is finished, if so, jumping to step 211, otherwise, jumping to step 206;
step 206, judging the sentence according to the trigger word list of the category and the trigger rule;
step 207, judging whether the sentence belongs to the category, if so, skipping to step 208, otherwise, skipping to step 204;
step 208, further judging according to the quantity type attribute word list;
step 209, judging whether the sentence is a quantitative attribute sentence, if so, jumping to step 210, otherwise, jumping to step 204;
step 210, returning the fine-grained type of the current execution process as an identification result;
step 211, return other as recognition results.
4. The method for understanding a quantitatively attribute-based sentence according to claim 1, wherein the argument extraction is based on a fine-grained category of the sentence, a trigger word position, and a syntax tree extraction semantic template, and specifically comprises:
step 300, designing a semantic template for each fine-grained category, wherein the template comprises arguments to be extracted;
step 301, summarizing sentence patterns of each fine-grained category;
step 302, obtaining a sentence and a fine-grained category thereof and a position of a trigger word corresponding to the fine-grained category according to a sentence recognition result;
step 303, recognizing and obtaining the position of an auxiliary word extracted from an auxiliary argument in the sentence;
step 304, extracting argument according to the positions of the trigger words and the auxiliary words;
step 305, judging whether all arguments to be extracted are identified, if so, jumping to step 307, otherwise, jumping to step 306;
step 306, extracting argument by using a syntactic structure rule;
and 307, returning the extraction results of all arguments to be extracted.
5. The method for understanding a quantitative attribute comparison-type sentence according to claim 1, wherein the output generating step specifically comprises:
step 400, identifying time in a sentence;
step 401, identifying a place in a sentence;
step 402, standardizing the expression of word meaning of the argument and time and place extracted from the sentence;
step 403, organizing the normalized result into structured data as a sentence analysis result;
step 404, structured data is returned.
6. The method as claimed in claim 1, wherein the sentence containing the quantitative comparison content is in the form of question-answer choice questions, and the question stem is connected to each choice to obtain a sentence as a processing object for sentence understanding.
7. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the computer program when executed implementing a quantitative attribute comparison sentence comprehension class method of any one of claims 1-6.
8. A computer-readable storage medium storing a computer program for executing the quantitative attribute comparison sentence interpretation method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210515296.5A CN115114929A (en) | 2022-05-12 | 2022-05-12 | Quantitative attribute comparison sentence understanding method, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210515296.5A CN115114929A (en) | 2022-05-12 | 2022-05-12 | Quantitative attribute comparison sentence understanding method, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115114929A true CN115114929A (en) | 2022-09-27 |
Family
ID=83326942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210515296.5A Pending CN115114929A (en) | 2022-05-12 | 2022-05-12 | Quantitative attribute comparison sentence understanding method, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115114929A (en) |
-
2022
- 2022-05-12 CN CN202210515296.5A patent/CN115114929A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7685082B1 (en) | System and method for identifying, prioritizing and encapsulating errors in accounting data | |
| US20180239758A1 (en) | Method and system for machine comprehension | |
| Kaplan et al. | Knowledge-based acquisition of causal relationships in text | |
| CN110727779A (en) | Question-answering method and system based on multi-model fusion | |
| US20200097265A1 (en) | Method and system for machine comprehension | |
| US20160154631A1 (en) | Method and system for machine comprehension | |
| CN108021703B (en) | Conversation type intelligent teaching system | |
| CN107301163A (en) | Text semantic analysis method and device comprising formula | |
| CN111143531A (en) | Question-answer pair construction method, system, device and computer readable storage medium | |
| CN112541337A (en) | Document template automatic generation method and system based on recurrent neural network language model | |
| CN117744661B (en) | Text generation model training method and text generation method based on prompt word engineering | |
| CN112417161B (en) | Method and storage device for recognizing upper and lower relationships of knowledge graph based on mode expansion and BERT classification | |
| CN115713072A (en) | Relation category inference system and method based on prompt learning and context awareness | |
| CN117648093A (en) | RPA flow automatic generation method based on large model and self-customized demand template | |
| CN113157887B (en) | Knowledge question and answer intention recognition method and device and computer equipment | |
| CN111680493B (en) | English text analysis method and device, readable storage medium and computer equipment | |
| CN114462387B (en) | Sentence pattern automatic discrimination method under no-label corpus | |
| CN116450855A (en) | Knowledge graph-based reply generation strategy method and system for question-answering robot | |
| Sinpang et al. | Detecting ambiguity in requirements analysis using Mamdani fuzzy inference | |
| CN115906818A (en) | Grammar knowledge prediction method, grammar knowledge prediction device, electronic equipment and storage medium | |
| CN115114929A (en) | Quantitative attribute comparison sentence understanding method, equipment and storage medium | |
| CN112989001B (en) | Question and answer processing method and device, medium and electronic equipment | |
| CN111949781B (en) | Intelligent interaction method and device based on natural sentence syntactic analysis | |
| CN117056524A (en) | Aspect-level emotion analysis method and system based on domain knowledge graph | |
| CN118674041A (en) | Method and device for executing business based on large language model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |