CN103870565A

CN103870565A - Semantic logic guide searching method based on interaction encyclopedic knowledge

Info

Publication number: CN103870565A
Application number: CN201410086557.1A
Authority: CN
Inventors: 何格; 顾君忠; 兰小敏
Original assignee: East China Normal University
Current assignee: East China Normal University
Priority date: 2014-03-11
Filing date: 2014-03-11
Publication date: 2014-06-18
Anticipated expiration: 2034-03-11
Also published as: CN103870565B

Abstract

The invention discloses a semantic logic guide searching method based on interaction encyclopedic knowledge. The method comprises the following steps: classifying interaction encyclopedic online descriptive knowledge data into concept, relation and entity, forming a three-element relation group according to relation among the three types of knowledge, and extracting and storing the interaction encyclopedic knowledge in a form of the three-element relation group; setting four semantic logic symbols, constructing an inquiry statement by use of the four semantic logic symbols by a user, processing the inquiry statement by use of the three-element relation group, obtaining a new inquiry statement submission search engine, and recommending the other inquiry statement obtained by processing to the user. According to the method, similar and relevant logics which can not be processed by a common search engine are used, the inquiry statement is processed into the new inquiry statement which can be effectively processed by the common search engine, and the search precision of the search engine is improved; meanwhile, a more accurate inquiry statement can be recommended to the user, so that the efficiency of the search engine is improved, and the inquiry satisfaction degree of the user is improved.

Description

Semantic logic guided search method based on interactive encyclopedia knowledge

Technical Field

The invention relates to information extraction, query processing and information retrieval, in particular to a knowledge set based on interactive encyclopedia massive ternary relationship group, which is combined with a knowledge matching algorithm to guide the search of a user.

Background

The information retrieval technology comprises a search engine which is a common tool for searching information in daily life of people, and particularly becomes an indispensable network application after the popularization of the use of the internet.

With the mass growth of internet information and the high accuracy requirement of people on information retrieval, a search engine which solely depends on the inverted index and word matching technology has an unsatisfactory search effect under the condition that query sentences input by users are generally short and cannot describe query intentions in detail and accurately. Especially, the ambiguity problem of the word makes the search effect worse. At present, many researches improve the search effect of the search engine to a certain extent from the aspects of the search engine, such as semantic disambiguation, query expansion, query log analysis, concept-based search and the like, and from the use angle of users, formulated query, user real-time interaction, query processing and the like. From the perspective of the user, much investment is needed to research how to input query sentences which can be effectively processed by the search engine in a more direct and natural language.

Disclosure of Invention

The invention aims to help a user to more accurately describe a search intention from the perspective of the user, in particular to a semantic logic guide search method based on interactive encyclopedia knowledge, which is provided for helping the user to process a query sentence with semantic logic which cannot be processed by a search engine into general logic which can be processed by the search engine depending on inverted and word matching. While other query statements recommended to the user may be returned.

The specific technical scheme for realizing the purpose of the invention is as follows:

a semantic logic guided search method based on interactive encyclopedia knowledge is characterized by comprising the following specific steps:

a) dividing knowledge data of the interactive encyclopedia online description into three types of concepts, relations and entities, forming a ternary relation group according to the relation among the three types of knowledge, and extracting and storing the knowledge of the interactive encyclopedia in the form of the ternary relation group;

b) setting four semantic logic symbols, constructing a query statement by a user by using the four semantic logic symbols, processing the query statement by using the characteristics of a ternary relationship group, submitting the obtained new query statement to a search engine, and recommending other processed query statements to the user; wherein,

the concept has unique semantics and is a title of the interactive encyclopedia; relationships are the description of the attributes of a concept and all the relationships that are linked to the concept; the entity does not have unique semantics, and the entity or a certain concept corresponds to a certain relation of the concept;

the relationship between the three types of knowledge is: concepts, entities and relationships between the two and concepts, concepts and relationships between the two; the concepts, the relations and the entities form a ternary relation group or the concepts, the concepts and the entities form a ternary relation group;

the extraction of the interactive encyclopedia knowledge in the form of the ternary relationship group is as follows: the title of the interactive encyclopedic webpage is determined as the main concept of the webpage, and information pairs in the interactive encyclopedic webpage have the rules of colon two sides, subordinate titles and subordinate texts thereof, wherein the information pairs have attributes, character relations and subordinate relations; the information pairs respectively correspond to the relationship and the entity or the relationship and the concept;

the four semantic logical symbols are:

".' dependent, forming the correlation logic: acquiring a corresponding entity or concept set according to the concepts and the related relations;

": associated, defining logic: specifying a concept based on a description of the concept or a word related to the concept;

"[ Lambda ] correlation logic: deducing a third element from two elements of the concepts, the relationships and the entities or the characteristics of the ternary relationship group among the concepts;

"to" similar logic: screening data using similar logic;

the query statement is constructed by using four semantic logic symbols, and the processing of the query statement by using the ternary relationship group is as follows: the user replaces the logic in the natural language with symbols and then processes the logic using the characteristics of the set of tri-relationships.

The invention divides encyclopedia knowledge into three types of data, utilizes a ternary relation group knowledge set extracted from the encyclopedia knowledge of interactive encyclopedia mass through four defined semantic logic symbols, combines the knowledge matching of word similarity to help a user construct query sentences in more natural language logic, particularly uses similar and related logic which can not be processed by a general search engine, processes the query sentences into new query sentences which can be effectively processed by the general search engine, and improves the search precision of the search engine. Meanwhile, more accurate query sentences can be recommended to the user, and even the knowledge which the user wants is directly returned to. Therefore, the efficiency of the search engine is improved, and the query satisfaction of the user is improved.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is a detailed flow of interactive encyclopedia knowledge extraction in the present invention.

FIGS. 3-6 are flow charts of the processing of four semantic logical symbols in a search according to the present invention;

fig. 7, 8 and 9 are diagrams for implementing the embodiment of the invention.

Detailed Description

Semantic logic symbol directed searches require editing query statements using semantic logic symbols, while requiring an interactive encyclopedia of triplet knowledge sets. Referring to fig. 1, the extraction of interactive encyclopedia knowledge and the definition and processing of semantic logical symbols are important parts, as described below.

1. Extraction of interactive encyclopedia knowledge

Like Wikipedia, the interactive encyclopedia comprises a large amount of manual editing, judgment and other participation works, and is an open, self-growing and continuously-improved encyclopedia. In general, various ternary relations can be seen on any webpage document of an interactive encyclopedia, and the ternary relation refers to a ternary relation group formed by two things and the relation between the two things. For the interactive encyclopedia webpage document, one element is the title of the webpage, and the other two elements are distributed between two sides of a colon or bold face and narrative text below the colon.

Here, the title of an interactive encyclopedia page is defined as a concept, and the other two elements are a relationship and an entity (the entity may also be a concept), respectively. That is, a set of three-way relationships is made up of concepts and concepts or concepts and entities plus relationships between them. Then the knowledge of the interactive encyclopedia is divided into three broad categories: concepts, relationships, entities.

An entity consists of a set of phrases or terms, which may be specific names of people, organization, books, specific events, etc. Many times, however, it is not possible to determine who is by one person's name alone, since there are many people who have the same name but who are different. For example, in the interactive encyclopedia, the name "Wangfei" includes "Wangfei [ Chinese female singer ]" and "Wangfei [ university professor ]" in addition to "Wangfei". That is, it cannot be determined by an entity what it is.

A concept represents a certain thing. For example, there is only one "Wangfei [ Chinese female singer ]" in the interactive encyclopedia.

There are many kinds of relationships between concepts or between concepts and entities, such as "blood type", "album", "movie work", and the like of "royal fiver [ chinese female singer ]".

The three-element knowledge is extracted by analyzing the HTML webpage according to the DOM frame and the rule of the interactive encyclopedia webpage document, and the process can be carried out in real time.

The DOM rules are mainly as follows (taking the interactive encyclopedia page of "royal fei" as an example):

1) the title "Wangfei [ Chinese female singer ]", which is a concept, was obtained by "div.content-h 1h 1".

2) "p # polysemyl part a" derives several concepts about "Wangfei": "Wangfei [ Chinese female singer ]", "Wangfei [ Guangyuan City generation of Guangyuan city, Sichuan province ]", and "Wangfei [ university professor ]".

3) Summary p get the summary.

4) "div # configuration lia" yields the human relationship: lei asia peng fu; xia Gui Ying mother; royal youth father, etc.

5) Module, zoom td "gets the attribute of" royal phenanthrene ": the name of Chinese: royal jelly; the name of English: faye Wong, etc.

6) "div # content p" acquires the profile information: "birth hospital: beijing collaborates with the hospital, reads primary school: beijing Ditan primary school' and the like.

7) "div # content 3" and get other correlations. Such as: album

In 1985: where the wind came from (Carle of charming) ("Miss of charming")

In 1986: hand division (Deng Li Jun Guxiang Bin Qing)

In 1989: king Jing (r) & gt.

The specific extraction process is shown in fig. 2.

2. Definition and processing of semantic logical symbols

In order to fully utilize the massive ternary relation group knowledge of the interactive encyclopedia to help information retrieval, four semantic logic symbols are defined. The basic principle is to determine the third element from two elements in the set of ternary relationships.

1) Corresponding entities are obtained according to concepts and relational words, and the symbols are used. The format is as follows: entity relation word

For example, "royal fei. daughter," plum-flower, sinojingtong "is available from the knowledge base of the ternary relationship group of the interactive encyclopedia.

The entity captured by this symbol may be multiple, for example, an "Wangfei album" results in a series of Wangfei song albums.

The first "entity" is a multi-meaning, such as "Wangfei" with multiple concepts "Chinese female singer", "university professor of the university", etc. Which concept is determined by whether there are two simultaneous instances of the given two entities. For example, "Wangfei album" may be determined to be an album for acquiring "Chinese female singer Wangfei" because of the other two concepts "Wangfei" — "Wangfei [ Guangyuan city, Sichuan province ]" and "Wangfei [ professor of the university of great homology ]" there is no "album".

Note that: the result of the processing is given two "entities" plus the third element obtained. For example, the 'Wangfei daughter' is treated to become 'Wangfei daughter plum'.

2) The third element is obtained from any two of the elements, using the symbol ^.

Entity ^ word

The term "herein may be a relational word or an entity.

The processed result is the third element. For example: "Aubama ^ USA" can obtain "president" from the knowledge base of the ternary relation group. The processing may result in more than one result, possibly "nationality", etc.

3) Definite concepts are determined from concept words or related words, using the notation ":".

Entity word

Specific concepts may be identified by words or related words that describe the specific concepts. Here, the specific concept word is determined by the collinearity of two words, like the previous example "royal phenanthrene. album".

For example, "Wangfei singer" or "Wangfei album" can be processed to obtain "Wangfei singer of Chinese female singer".

4) And selecting the word most similar to a certain word from the word set, and using the symbols- ".

Word set-word

Similar logic is often used in conjunction with obtaining the symbol ". times" of an entity set. For example, "gutianle. movie-sweet" means that the movie name most similar to "sweet" is found from the movie list of gutianle. The final processing result is 'Gutianle movie sweet language honey language'.

The similarity logic here requires similarity calculation using words. Consider that when a user submits a search, the inputs typically have the same words or phrases for unknown or ambiguous portions. Here, a method of counting the same words and phrases is used to calculate the similarity of the words, wherein the words are more important than the words, so the two should have different weight calculations:

where p is the statement to be collated and p_iIs one of the sets of sentences to be screened, Sch (p, p)_i) Is two statements p and p_iThe number of words in between, Lch (p) is the number of words in the statement p, Ste (p, p)_i) Is the number of words that are the same between two sentences, lte (p) is the number of words of sentence p, α and β are weighting parameters, satisfying α + β = 1; similarly, σ and τ are weighting parameters, and σ + τ =1 is satisfied.

In addition, the determination uses the symbol ": when determining a concept by inputting a concept word, since the user does not necessarily input a complete description of the concept, for example," royal fei: singer ", it is also necessary to select a concept word" chinese female singer "from among a plurality of concepts of" royal fei "in this case using word similarity calculation. Similarly, the same problem exists in the application of the symbols ". DELTA.s" and "^ s", and knowledge matching is also required through word similarity calculation, wherein the adjustment parameters used for calculating the word similarity corresponding to each semantic logic are not necessarily the same.

In general, the use of four semantic logic symbols requires the use of word similarity to match words entered by a user with knowledge of the interactive encyclopedia. The processing logic for the four symbols is shown in fig. 3-6, respectively.

The query sentence edited by the user by using four semantic logic symbols complies with the priority processing principle from left to right in the processing process.

3. Semantic logic guided search

The interactive encyclopedia knowledge-based semantic logic-guided search process is shown in figure 1, and mainly comprises the following steps:

firstly, a user inputs a query statement by using semantic logic symbols; for example, the query sentence "gutianle, movie-sweet princess" is input.

And secondly, analyzing the logic symbols in the query statement according to the use rule of each logic symbol according to a priority principle from left to right. And according to the sequence of the original query statement, adding the analyzed partial query terms into a new query statement sequence in sequence. No parts using semantic logical notation are added directly in order. Considering that there may be more than one result of the semantic logic processing, the new query statement is changed into a sequence of term groups (which may be represented as a two-bit array). For example, "gu tianle, movie to sweet girl princess" query sentences are processed in sequence as follows:

1) after processing the first logical symbol ".", the new set of query statements becomes:

gutianle (a Chinese character)	Film	O records the record of the record
			——	——	Male child going into bottle
——	——	Play fire
			——	——	。。。

2) After processing the second logical symbol "-" becomes:

gutianle (a Chinese character)

Film

Sweet language honey

3) And (3) processing a new query statement group:

gutianle (a Chinese character)

Film

Sweet language honey

Main angle of woman

And thirdly, extracting the first line of the new query statement group obtained by processing as the processed query statement. The new query statement in the example is "Gutianle movie sweet language maid hero".

And fourthly, submitting the new query statement to a search engine.

And fifthly, returning the webpage result obtained by query to the user. Meanwhile, if the obtained query statement group has more than one row, data of non-empty columns in other rows except the first row is recommended to the user.

According to the above description, a search method based on semantic logic guidance of interactive encyclopedia knowledge is realized by the Google search engine, and the returned results after the method is implemented are shown in FIGS. 7, 8 and 9. In embodiment 1, when the user does not know the movie name and only memorizes the word "sweet", the user can directly obtain the desired movie name "sweet language" through similar logic processing, thereby searching and obtaining the desired web page result. Without this processing, the returned web page results cannot meet their requirements. In example 2, in addition to "drunk boxing", other movies with "boxing" are recommended to the user, and more information can be provided to the user. In embodiment 3, the concept "royal membrane [ chinese female singer ]" is determined by "singer", and although the result desired by the user can be obtained by only "royal membrane singer", the determination of the concept by semantic logic symbols can help disambiguation, especially in concept-based search.

In summary, the present invention can help the user to construct query sentences according to the natural language logic to a certain extent, and simultaneously process the query sentences into query sentences which can be effectively processed by a search engine based on inversion and word matching, so as to recommend accurate information to the user, reduce the number of searches required for obtaining results, and even directly return the accurate information required by the user, thereby improving the user satisfaction.

Claims

1. A semantic logic guided search method based on interactive encyclopedia knowledge is characterized by comprising the following specific steps:

b) setting four semantic logic symbols, constructing a query statement by a user by using the four semantic logic symbols, processing the query statement by using a ternary relationship group, submitting the obtained new query statement to a search engine, and recommending other processed query statements to the user; wherein,

the four semantic logical symbols are:

"-" similar logic: screening data using similar logic;