CN108776940A - A kind of intelligent food and drink proposed algorithm excavated based on text comments - Google Patents
A kind of intelligent food and drink proposed algorithm excavated based on text comments Download PDFInfo
- Publication number
- CN108776940A CN108776940A CN201810566021.8A CN201810566021A CN108776940A CN 108776940 A CN108776940 A CN 108776940A CN 201810566021 A CN201810566021 A CN 201810566021A CN 108776940 A CN108776940 A CN 108776940A
- Authority
- CN
- China
- Prior art keywords
- user
- comment
- dining room
- marking
- word
- 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
- 235000013305 food Nutrition 0.000 title claims abstract description 27
- 230000008451 emotion Effects 0.000 claims abstract description 41
- 238000011156 evaluation Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims abstract description 15
- 239000000284 extract Substances 0.000 claims abstract description 4
- 230000000694 effects Effects 0.000 claims description 16
- 235000013311 vegetables Nutrition 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 7
- 239000004615 ingredient Substances 0.000 claims description 5
- RTHCYVBBDHJXIQ-UHFFFAOYSA-N N-methyl-3-phenyl-3-[4-(trifluoromethyl)phenoxy]propan-1-amine Chemical compound C=1C=CC=CC=1C(CCNC)OC1=CC=C(C(F)(F)F)C=C1 RTHCYVBBDHJXIQ-UHFFFAOYSA-N 0.000 claims description 3
- 230000002996 emotional effect Effects 0.000 claims 1
- 238000000034 method Methods 0.000 abstract description 6
- 235000021404 traditional food Nutrition 0.000 abstract description 3
- 238000004364 calculation method Methods 0.000 description 5
- 230000001755 vocal effect Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 238000000205 computational method Methods 0.000 description 1
- 230000036461 convulsion Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000035622 drinking Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 210000003739 neck Anatomy 0.000 description 1
- 238000011002 quantification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/12—Hotels or restaurants
-
- 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)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Tourism & Hospitality (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Machine Translation (AREA)
Abstract
The present invention relates to the intelligent food and drink proposed algorithms excavated based on text comments, and the invention belongs to computerized algorithm fields.The present invention establishes user dining room scoring matrix, calculates user's marking similarity by collecting user's food and drink data;Extract feature emotion word pair;Merge homogenous characteristics, quantifies the score of each feature;User preference scoring matrix is established, the preference similarity of user is calculated;Calculate user's comprehensive similarity;Calculate users' trust value;Calculate the degree of belief between user;It is given a mark for dining room by weighting evaluation value based on degree of belief between user;Recommend top n dining room.The present invention utilizes the method that text comments excavate, a series of decimation rules are formulated, form comment abstract, emotion information of the user to characteristic attributes such as environment, the services in dining room is extracted in making a summary from comment using interdependent syntax, totally marking combines in the dining room of score and user after characteristic attribute is quantified, recommends dining room.Compared with traditional food and drink is recommended, the accuracy rate of recommendation is improved.
Description
Technical field
The present invention relates to the intelligent food and drink proposed algorithms excavated based on text comments, and the invention belongs to computerized algorithm necks
Domain.
Background technology
Commending system be the information for referring to that user may be liked or material object (such as:It is film, music, books, new
Hear, picture) recommend a kind of application of user.Proposed algorithm is the core of commending system, and performance determines commending system
Quality.Therefore, the research center of gravity of commending system is consistently placed in proposed algorithm.Currently, commonly being pushed away in commending system
It recommends algorithm mainly and has and is following several:Content-based recommendation, the recommendation based on correlation rule, is based on effectiveness at collaborative filtering
Recommendation, Knowledge based engineering recommend and combined recommendation.These algorithms are played very important due to the pros and cons of itself in certain fields
Effect, but food and drink recommend accuracy rate on it is to be improved.
It is a kind of suggested design popular at present based on the recommendation that comment text excavates.Text mining is data mining
Application in Web texts, text message excavation, has become the main stream approach of personalized recommendation emphatically.Intelligent food and drink is pushed away
It recommends for field, consumer contains the comment in dining room that consumption suggestions, impression of having dinner, service experience etc. is largely valuable disappears
Charge information can accurately portray the various features in dining room.And traditional food and drink commending system does not account for these features, only provides
Consumer gives a mark to the totality in dining room so that the accuracy of commending system decreases.Therefore the present invention is directed to the food and drink of user
Text mining has been done in comment, and the accuracy recommended is improved using the characteristic information excavated.
Invention content
The present invention provides a kind of intelligent food and drink proposed algorithm excavated based on text comments regarding to the issue above.
The present invention adopts the following technical scheme that:
The intelligent food and drink proposed algorithm of the present invention excavated based on text comments, algorithm steps are as follows:
1) user's food and drink data, are collected, according to the user-dining room scoring matrix for collecting data information foundation such as following formula, are divided
It is user U, dining room G, user marking M, user comment not define food and drink data;
In formulaIndicate user UiTo dining room GjMarking
And it is calculate by the following formula to obtain Euclidean distance using user-dining room scoring matrix:
D represents distance, and d () indicates an operator, for calculating two vectorial Euclidean distances;
Marking similarity is calculate by the following formula using the Euclidean distance that above formula obtains:
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to dining room a,
A is user UiWith user UjExcessive dining room number is beaten jointly;For user UiWith user UjMarking similarity;
2), the user comment being directed in step 1) carries out participle and part of speech label;It is short to pretreated comment using LTP
Sentence carries out interdependent syntactic analysis;Obtain the dependency relationship type between each ingredient of sentence;Abstract decimation rule is formulated from interdependent sentence
Feature emotion word is extracted in method tree;
The form of feature emotion word pair is W=(wc,we), w in formulacIt is characterized attribute word, weThe qualifier of feature thus;
3), needle is using service respectively, and environment is hygienic, and vegetable is characterized;Merge homogenous characteristics by such as following formula and quantifies each spy
Sign:
Gj={ wc1:[we11,we12,…we1n];wc2:[we21,we22,…we2n];…wc4:[we41,we42,…we4n] in formula
wc1,wc2,wc3,wc4It corresponds to service respectively, environment, health, vegetable;wem1,wem2…wemnFor all qualifiers under this feature;
And positive emotion dictionary, Negative Affect dictionary, negative sentiment dictionary and degree adverb dictionary are established, each feature is commented
Point;
4), scoring is obtained for step 3) establish following user-preference scoring matrix:
Wherein:Indicate user UjTo 4 feature cmThe marking of (service, environment, health, vegetable);
And user preference similarity is calculated by such as following formula
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to feature b, B are
User UiWith user UjExcessive number of features is beaten jointly;For user UiWith user UjMarking similarity.
5), by the user obtained in step 1 marking similarityThe user preference similarity obtained with step 4)
Weights combination is carried out by following formula:
Wherein:β is equal to 0.5;
6) it, calculates separately family liveness and user's evaluation is efficient;User activity is obtained by following formula
Wherein HUFor the activity of the user, e is the nature truth of a matter, and n is the number of marking operation, tiFor the duration, υ is to live
Jerk;
Judge whether comment is effectively to comment on by following formula,
NfavourIt agrees with counting for comment, NagainstTo comment on antilogarithm;
It is calculate by the following formula the effective percentage of effective evaluation again:
Wherein EUFor user comment effective percentage, NEffectively commentFor the quantity effectively commented on, NCommentFor the comment sum of the user;
7) user activity and user's evaluation, obtained for step 6 is efficient, is calculate by the following formula users to trust degree:
TU=HU+EU
User activity HUIt is higher, users' trust value TUIt is higher;User comment effective percentage EUIt is higher, users' trust value TUMore
It is high;
8), pass through degree of belief between users to trust degree calculating user:
TDi,j=Simi,j×Tj
Wherein TDi,jIndicate user UiTo UjDegree of belief, Simi,jFor user UiAnd UjComprehensive similarity, TjFor user Uj
Trust value.User UjTrust value is higher, user UiTo UjDegree of belief it is higher;
9) degree of belief is given a mark by following formula for dining room by weighting evaluation value between, being based on step 8 user:
Wherein:Scorei,gIndicate user UiTo dining room GgWeighting evaluation value, TDi,jIndicate user UiTo UjDegree of belief,
Mj,gFor user UjTo dining room GgMarking, U be user set;
10), after by giving a mark to all dining rooms of not giving a mark, by being ranked up from high to low, recommend top n dining room.
The intelligent food and drink proposed algorithm of the present invention excavated based on text comments, using to comment data in step 1
Participle and part-of-speech tagging are carried out, interdependent syntactic analysis is carried out to pretreated comment short sentence using LTP, obtains each ingredient of sentence
Between dependency relationship type, formulate decimation rule and extract feature emotion word pair.
The intelligent food and drink proposed algorithm of the present invention excavated based on text comments, is commented for the sentiment dictionary of foundation
Divider is then as follows:
(1), each positive emotion word assigns weight 1, and each Negative Affect word assigns weight -1, and assumes emotion
Value meets linear superposition theorem;
(2) if, the qualifier under feature include corresponding word in dictionary, in addition corresponding weights;It negate language appropriate to the occasion
Weights opposite sign, degree adverb enable weights double;
(3) if, total weight value be that just, emotion is that commendation if total weight value is negative, for derogatory sense, is otherwise neutrality;Feature is beaten
Divide and use the five-grade marking system, commendation is 5 points, and derogatory sense is 1 point, and neutrality is 3;
It is to feature:Service, environment, health, vegetable quantify later as a result, for establishing subsequent user-preference
Scoring matrix.
Advantageous effect
The present invention only gives a mark to the totality in dining room for existing food and drink commending system, have ignored the vegetable in dining room, environment,
The poor problem of recommendation effect caused by the attributive character such as service quality.Propose a kind of intelligence excavated based on text comments
Food and drink proposed algorithm.The present invention utilizes the method that text comments excavate, and has formulated a series of decimation rules, forms comment abstract,
User is extracted in making a summary from comment using interdependent syntax to the emotion information of the characteristic attributes such as environment, the service in dining room, it will be special
The dining room totally marking combination for levying the score and user after attribute quantification, recommends dining room.Recommend with traditional food and drink
It compares, improves the accuracy rate of recommendation.
Specific implementation mode
It is clearer for the purpose and technical solution that make the embodiment of the present invention, below to the technical solution of the embodiment of the present invention
It is clearly and completely described.Obviously, described embodiment is a part of the embodiment of the present invention, rather than whole realities
Apply example.Based on described the embodiment of the present invention, those of ordinary skill in the art are in the institute of the under the premise of without creative work
The every other embodiment obtained, shall fall within the protection scope of the present invention.
The present invention is based on the intelligent food and drink proposed algorithms that text comments excavate
Step 1:User's food and drink data are crawled, user-dining room scoring matrix is established, calculate user's marking similarity:It is described
User's food and drink data are user U, dining room G, and user gives a mark M, and user comment, these fields can crawl from food and drink website.
The user-dining room matrix is as follows:
In formulaIndicate user UiTo dining room GjMarking
And it is calculate by the following formula to obtain Euclidean distance using user-dining room scoring matrix:
Marking similarity is calculate by the following formula using the Euclidean distance that above formula obtains:
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to dining room a,
A is user UiWith user UjExcessive dining room number is beaten jointly;For user UiWith user UjMarking similarity;D is
The meaning of distance, d () indicate an operator, are used for calculating two vectorial Euclidean distances here.
Step 2:Participle and part of speech label are carried out to user comment, using LTP (LTP is language technology platform) to pretreatment
Comment short sentence afterwards carries out interdependent syntactic analysis, obtains the dependency relationship type between each ingredient of sentence, formulates abstract and extracts rule
Feature emotion word pair is then extracted from interdependent syntax tree:The form of the feature emotion word pair is W=(wc,we), wherein wcFor
Characteristic attribute word, weThe qualifier of feature thus.
Extraction feature emotion word pair method be:Participle and part-of-speech tagging are carried out to food and drink comment data, knot can be used
Bar participle etc. tools complete.Interdependent syntactic analysis is carried out to pretreated comment short sentence using LTP, obtain sentence respectively at point it
Between dependency relationship type, formulate abstract decimation rule feature emotion word pair is extracted from interdependent syntax tree;It is specific to extract rule
It is then as follows:
(1) part of speech of core words is adjective:
Step1:Core word is stored in emotion word list;
Step2:Traverse the word that all grammatical relations are sent out from core word;
Step3:If the grammatical relation of this word and core word is COO, COO indicates coordination, and part of speech is a
(adjective), then the word is emotion word, is deposited into emotion word list;
Step4:If the grammatical relation of this word and core word is ADV, ADV indicates verbal endocentric phrase, then this word is
Adverbial word is deposited into adverbial word list;
Step5:If the grammatical relation of this word and core word is SBV, SBV indicates subject-predicate relationship, and part of speech is n
(noun), then the word is attributive character word, is deposited into attributive character word list;
Step6:Traverse the word that all grammatical relations are sent out from Feature Words;
Step7:If the grammatical relation of the word and Feature Words is COO, COO indicates coordination, then this word is also to belong to
Property Feature Words, are deposited into attributive character word list.
(2) part of speech of core words is common saying:
Step1:The word is stored in emotion word list;
Step2:Traverse the word that all grammatical relations are sent out from the emotion word;
Step3:If the grammatical relation of the word and emotion word is SBV, SBV indicates subject-predicate relationship, and part of speech is n (names
Word), then the word is attributive character word, is deposited into attributive character word list.
If the part of speech of core word is verb, clause is relatively complicated changeable, is divided into 4 kinds of situations and carries out analysis digging
Pick:
(a) core word is " liking ", " love ":
Step1:Core word is stored in verb list;
Step2:Traverse the word that all grammatical relations are sent out from core word;
Step3:If the grammatical relation of the word and core word is ADV, ADV indicates verbal endocentric phrase, then the word is adverbial word, will
It is stored in adverbial word list L1;
Step4:If the grammatical relation of the word and core word is VOB, VOB indicates to move guest's relationship, and part of speech is n (names
Word), attributive character word is deposited into attributive character word list.
(b) core word is "Yes", " being exactly ", " feeling ", " feeling ":
Step1:Traverse the word that all grammatical relations are sent out from core word;
Step2:If the grammatical relation of the word and core word is VOB, VOB indicates to move guest's relationship, and part of speech is that a (is described
Word), then the word is emotion word, is stored in emotion word list;
Step3:Traverse the word that all grammatical relations are sent out from emotion word;
Step4:If the relationship of the word and emotion word is ADV, ADV indicates verbal endocentric phrase, then the word is adverbial word, is deposited
Enter adverbial word list L2;
Step5:If the relationship of the word and emotion word is SBV, SBV indicates subject-predicate relationship, and part of speech is n (noun), then
The word is attributive character word, is deposited into attributive character word list.
(c) core word is " service ", "Off", " eating ", " drinking ":
Step1:" distance " is saved in category by core word deposit attributive character word list if core word is "Off"
Property feature word list;
Step2:Traverse the word that all grammatical relations are sent out from core word;
Step3:If the relationship of the word and core word is CMP, CMP indicates structure of complementation, and part of speech is a (adjective),
Then the word is emotion word, is deposited into emotion word list;
Step4:Traverse the word that all grammatical relations are sent out from emotion word;
Step5:If the relationship of the word and emotion word is ADV, ADV indicates verbal endocentric phrase, then the word is adverbial word, is deposited
Enter adverbial word list L2.
(d) core word is " needing ":
Step1:Core word is stored in verb list;
Step2:Traverse the word that all grammatical relations are sent out from core word;
Step3:If the relationship of the word and core word is SBV, SBV indicates subject-predicate relationship, then the word is attributive character word,
It is deposited into attributive character word list;
Step4:If the relationship of the word and core word is ADV, ADV indicates verbal endocentric phrase, then the word is adverbial word, is deposited
Enter adverbial word list L1;
Step5:If the relationship of the word and core word is VOB, VOB indicates to move guest's relationship, and part of speech is v (verb), then
The word is emotion word, is deposited into emotion word list.
Step 3:Merge homogenous characteristics, quantify the score of each feature, for the feature extracted, a kind of spy will be belonged to
Sign is merged together, then user's character representation of interest is:
Gj={ wc1:[we11,we12,…we1n];wc2:[we21,we22,…we2n];…wc4:[we41,we42,…we4n]}
Wherein wc1,wc2,wc3,wc4For 4 features, specifically correspond to:Service, environment, health, vegetable, wem1,wem2…wemn
For all qualifiers under this feature.
It is right to establish sentiment dictionary (positive emotion dictionary, Negative Affect dictionary, negative sentiment dictionary and degree adverb dictionary)
Each feature scores;The code of points is as follows:
(1) each positive emotion words of assign weight 1, and each Negative Affect word assigns weight -1, and assumes emotion
Value meets linear superposition theorem;
(2) if the qualifier under features includes corresponding word in dictionary, corresponding weights are just added.In addition, negative
Language appropriate to the occasion weights opposite sign, degree adverb enable weights double;
(3) if total weight values are just, emotion, for derogatory sense, is otherwise neutrality if total weight value is negative for commendation.Feature is beaten
Divide and use the five-grade marking system, commendation is 5 points, and derogatory sense is 1 point, and neutrality is 3.
Step 4:User-preference scoring matrix is established, the preference similarity of user is calculated:User's feature of interest is beaten
It is user preference marking to divide.The user-preference scoring matrix is as follows:
Wherein:Indicate user UjTo 4 feature cmThe marking of (service, environment, health, vegetable).
Calculate the preference similarity of user:The calculating similarity based method is calculate by the following formula to obtain user inclined with step 1
Good similarity:And user preference similarity is calculated by such as following formula
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to feature b, B
For user UiWith user UjExcessive number of features is beaten jointly;For user UiWith user UjMarking similarity.
Step 5:Calculate user's comprehensive similarity:The method for calculating user's comprehensive similarity is similar for user gives a mark
DegreeWith user preference similarityIt is combined using weights appropriate, formula is as follows:
Wherein:β is equal to 0.5.
Step 6:It is efficient according to user activity and user's evaluation, calculate users' trust value, the user activity meter
Calculation method is as follows:Often operation is marked in a dining room in user, and liveness adds υ, and one is only calculated daily in same dining room
It is secondary, it can add up between different dining rooms.Marking operation needs to position.Liveness linearly increases in the case where account is always maintained at active
Long, after the markup operation of account is reduced, temporally length t does forthright linear reduction again.Its calculation formula is:
Wherein HUFor the activity of the user, e is the nature truth of a matter, and n is the number of marking operation, tiFor the duration.
The user's evaluation effective percentage computational methods are as follows:User's food and drink comment can be visible to other users.Other users
It can be evaluated and judge agreeing (favour) or oppose (against).If certain comment,(NfavOur is that comment is agreed with counting, NagainstTo comment on antilogarithm), then the comment is effectively comment.
User's evaluation effective percentage calculation formula is:
Wherein EUFor user comment effective percentage, NEffectively commentFor the quantity effectively commented on, NCommentFor the comment sum of the user.
It is efficient according to user activity and user's evaluation, users' trust value is calculated into the users' trust value calculates public
Formula is:
TU=HU+EU
User activity HUIt is higher, users' trust value TUIt is higher;User comment effective percentage EUIt is higher, users' trust value TUMore
It is high.
Step 7:The degree of belief between user is calculated, the users to trust degree calculation formula is as follows:
TDi,j=Simi,j×Tj
Wherein TDi,jIndicate user UiTo UjDegree of belief, Simi,jFor user UiAnd UjComprehensive similarity, TjFor user Uj
Trust value.User UjTrust value is higher, user UiTo UjDegree of belief it is higher.
Step 8:It is given a mark for dining room by weighting evaluation value based on degree of belief between user:The marking calculation formula is as follows:
Wherein:Scorei,gIndicate user UiTo dining room GgWeighting evaluation value, TDi,jIndicate user UiTo UjDegree of belief,
Mj,gFor user UjTo dining room GgMarking, U be user set.
Step 9:N family dining room before recommending, after giving a mark to all dining rooms of not giving a mark, by score according to carrying out from high to low
Sequence, N family dining room before recommending.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
Any one skilled in the art in the technical scope disclosed by the present invention, the change or replacement that can be readily occurred in,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of the claims
Subject to.
Claims (3)
1. a kind of intelligent food and drink proposed algorithm excavated based on text comments, it is characterised in that:Algorithm steps are as follows:
1) user's food and drink data, are collected, it is fixed respectively according to the user-dining room scoring matrix for collecting data information foundation such as following formula
Adopted food and drink data be user U, dining room G, user give a mark M, user comment,
In formulaIndicate user UiTo dining room GjMarking;
And it is calculate by the following formula to obtain Euclidean distance using user-dining room scoring matrix:
Marking similarity is calculate by the following formula using the Euclidean distance that above formula obtains:
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to dining room a, A are to use
Family UiWith user UjExcessive dining room number is beaten jointly;For user UiWith user UjMarking similarity;
2), the user comment being directed in step 1) carries out participle and part of speech label;Using LTP to pretreated comment short sentence into
The interdependent syntactic analysis of row;Obtain the dependency relationship type between each ingredient of sentence;Abstract decimation rule is formulated from interdependent syntax tree
In extract feature emotion word;
The form of feature emotion word pair is W=(wc,we), w in formulacIt is characterized attribute word, weThe qualifier of feature thus;
3), needle is using service respectively, and environment is hygienic, and vegetable is characterized;Merge homogenous characteristics by such as following formula and quantifies each feature:
Gj={ wc1:[we11,we12,…we1n];wc2:[we21,we22,…we2n];…wc4:[we41,we42,…we4n] w in formulac1,
wc2,wc3,wc4It corresponds to service respectively, environment, health, vegetable;wem1,wem2…wemnFor all qualifiers under this feature;And
Positive emotion dictionary, Negative Affect dictionary, negative sentiment dictionary and degree adverb dictionary are established, is scored each feature;
4), scoring is obtained for step 3) establish following user-preference scoring matrix:
Wherein:Indicate user UjTo 4 feature cmThe marking of (service, environment, health, vegetable);
And user preference similarity is calculated by such as following formula
Wherein D in formulai,jFor user UiAnd UjEuclidean distance,WithMarking for user to feature b, B are user
UiWith user UjExcessive number of features is beaten jointly;For user UiWith user UjMarking similarity.
5), by the user obtained in step 1 marking similarityThe user preference similarity obtained with step 4)Pass through
Following formula carries out weights combination:
Wherein:β is equal to 0.5;
6) it, calculates separately family liveness and user's evaluation is efficient;User activity is obtained by following formula
Wherein HUFor the activity of the user, e is the nature truth of a matter, and n is the number of marking operation, tiFor the duration, υ is liveness;
Judge whether comment is effectively to comment on by following formula,
NfavourIt agrees with counting for comment, NagainstTo comment on antilogarithm;
It is calculate by the following formula the effective percentage of effective evaluation again:
Wherein EUFor user comment effective percentage, NEffectively commentFor the quantity effectively commented on, NCommentFor the comment sum of the user;
7) user activity and user's evaluation, obtained for step 6 is efficient, is calculate by the following formula users to trust degree:
TU=HU+EU
User activity HUIt is higher, users' trust value TUIt is higher;User comment effective percentage EUIt is higher, users' trust value TUIt is higher;
8), pass through degree of belief between users to trust degree calculating user:
TDi,j=Simi,j×Tj
Wherein TDi,jIndicate user UiTo UjDegree of belief, Simi,jFor user UiAnd UjComprehensive similarity, TjFor user UjLetter
Appoint value.User UjTrust value is higher, user UiTo UjDegree of belief it is higher;
9) degree of belief is given a mark by following formula for dining room by weighting evaluation value between, being based on step 8 user:
Wherein:Scorei,gIndicate user UiTo dining room GgWeighting evaluation value, TDi,jIndicate user UiTo UjDegree of belief, Mj,gFor
User UjTo dining room GgMarking, U be user set;
10), after by giving a mark to all dining rooms of not giving a mark, by being ranked up from high to low, recommend top n dining room.
2. the intelligent food and drink proposed algorithm according to claim 1 excavated based on text comments, it is characterised in that:Step 1
It is middle to carry out participle and part-of-speech tagging using to comment data, interdependent syntax point is carried out to pretreated comment short sentence using LTP
Analysis obtains the dependency relationship type between each ingredient of sentence, formulates decimation rule to extract feature emotion word pair.
3. the intelligent food and drink proposed algorithm according to claim 1 excavated based on text comments, it is characterised in that:For building
The code of points of vertical sentiment dictionary is as follows:
(1), each positive emotion word assigns weight 1, and each Negative Affect word assigns weight -1, and assumes that emotional value is full
Sufficient linear superposition theorem;
(2) if, the qualifier under feature include corresponding word in dictionary, in addition corresponding weights;Negate language appropriate to the occasion weights
Opposite sign, degree adverb enable weights double;
(3) if, total weight value be that just, emotion is that commendation if total weight value is negative, for derogatory sense, is otherwise neutrality;Feature marking is adopted
With the five-grade marking system, commendation is 5 points, and derogatory sense is 1 point, and neutrality is 3;
It is to feature:Service, environment, health, vegetable quantify later as a result, for establishing subsequent user-preference marking
Matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810566021.8A CN108776940A (en) | 2018-06-04 | 2018-06-04 | A kind of intelligent food and drink proposed algorithm excavated based on text comments |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810566021.8A CN108776940A (en) | 2018-06-04 | 2018-06-04 | A kind of intelligent food and drink proposed algorithm excavated based on text comments |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108776940A true CN108776940A (en) | 2018-11-09 |
Family
ID=64024609
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810566021.8A Pending CN108776940A (en) | 2018-06-04 | 2018-06-04 | A kind of intelligent food and drink proposed algorithm excavated based on text comments |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108776940A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109697657A (en) * | 2018-12-27 | 2019-04-30 | 厦门快商通信息技术有限公司 | A kind of dining recommending method, server and storage medium |
CN109886773A (en) * | 2019-01-17 | 2019-06-14 | 平安城市建设科技(深圳)有限公司 | Recommended method, device, equipment and storage medium based on lessee's credit appraisal |
CN110457460A (en) * | 2019-06-20 | 2019-11-15 | 拉扎斯网络科技(上海)有限公司 | Text recommended method, device, server and storage medium |
CN111061962A (en) * | 2019-11-25 | 2020-04-24 | 上海海事大学 | Recommendation method based on user score analysis |
CN111798337A (en) * | 2020-05-22 | 2020-10-20 | 平安国际智慧城市科技股份有限公司 | Environmental sanitation supervision method, device, equipment and storage medium for catering enterprises |
CN112100517A (en) * | 2020-09-17 | 2020-12-18 | 哈尔滨理工大学 | Method for relieving cold start problem of recommendation system based on content feature extraction |
CN112949322A (en) * | 2021-04-27 | 2021-06-11 | 李蕊男 | E-commerce opinion mining recommendation system driven by online text comments |
CN112991017A (en) * | 2021-03-26 | 2021-06-18 | 刘秀萍 | Accurate recommendation method for label system based on user comment analysis |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130007015A1 (en) * | 2006-12-28 | 2013-01-03 | Ebay Inc. | Collaborative content evaluation |
CN106791964A (en) * | 2016-12-26 | 2017-05-31 | 中国传媒大学 | Broadcast TV program commending system and method |
-
2018
- 2018-06-04 CN CN201810566021.8A patent/CN108776940A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130007015A1 (en) * | 2006-12-28 | 2013-01-03 | Ebay Inc. | Collaborative content evaluation |
CN106791964A (en) * | 2016-12-26 | 2017-05-31 | 中国传媒大学 | Broadcast TV program commending system and method |
Non-Patent Citations (2)
Title |
---|
DONGCHEN MIAO等: "A Recommendation System Based on Text Mining", 《2017 INTERNATIONAL CONFERENCE ON CYBER-ENABLED DISTRIBUTED COMPUTING AND KNOWLEDGE DISCOVERY (CYBERC)》 * |
张进: "社交网络中基于聚类分析的可信推荐系统", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109697657A (en) * | 2018-12-27 | 2019-04-30 | 厦门快商通信息技术有限公司 | A kind of dining recommending method, server and storage medium |
CN109886773A (en) * | 2019-01-17 | 2019-06-14 | 平安城市建设科技(深圳)有限公司 | Recommended method, device, equipment and storage medium based on lessee's credit appraisal |
CN110457460A (en) * | 2019-06-20 | 2019-11-15 | 拉扎斯网络科技(上海)有限公司 | Text recommended method, device, server and storage medium |
CN111061962A (en) * | 2019-11-25 | 2020-04-24 | 上海海事大学 | Recommendation method based on user score analysis |
CN111061962B (en) * | 2019-11-25 | 2023-09-29 | 上海海事大学 | Recommendation method based on user scoring analysis |
CN111798337A (en) * | 2020-05-22 | 2020-10-20 | 平安国际智慧城市科技股份有限公司 | Environmental sanitation supervision method, device, equipment and storage medium for catering enterprises |
CN112100517A (en) * | 2020-09-17 | 2020-12-18 | 哈尔滨理工大学 | Method for relieving cold start problem of recommendation system based on content feature extraction |
CN112991017A (en) * | 2021-03-26 | 2021-06-18 | 刘秀萍 | Accurate recommendation method for label system based on user comment analysis |
CN112949322A (en) * | 2021-04-27 | 2021-06-11 | 李蕊男 | E-commerce opinion mining recommendation system driven by online text comments |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108776940A (en) | A kind of intelligent food and drink proposed algorithm excavated based on text comments | |
CN106528656B (en) | A kind of method and system for realizing that course is recommended based on student's history and real-time learning state parameter | |
CN102682120B (en) | Method and device for acquiring essential article commented on network | |
CN109960786A (en) | Chinese Measurement of word similarity based on convergence strategy | |
CN106484829B (en) | A kind of foundation and microblogging diversity search method of microblogging order models | |
CN105045857A (en) | Social network rumor recognition method and system | |
Shimada et al. | Analyzing tourism information on twitter for a local city | |
CN107133214A (en) | A kind of product demand preference profiles based on comment information are excavated and its method for evaluating quality | |
CN106407235B (en) | A kind of semantic dictionary construction method based on comment data | |
CN103917968A (en) | System and method for managing opinion networks with interactive opinion flows | |
CN102999507A (en) | Recommendation processing method and device for information of network microblog celebrities | |
CN106354845A (en) | Microblog rumor recognizing method and system based on propagation structures | |
CN104268230B (en) | A kind of Chinese micro-blog viewpoint detection method based on heterogeneous figure random walk | |
CN106294744A (en) | Interest recognition methods and system | |
CN108334493A (en) | A kind of topic knowledge point extraction method based on neural network | |
JP2009099088A (en) | Sns user profile extraction device, extraction method and extraction program, and device using user profile | |
CN107092605A (en) | A kind of entity link method and device | |
CN102200973A (en) | Equipment and method for generating viewpoint pair with emotional-guidance-based influence relationship | |
CN104598648B (en) | A kind of microblog users interactive mode gender identification method and device | |
CN106933969A (en) | Personalized recommendation system and recommendation method based on industry upstream-downstream relationship | |
CN109086355A (en) | Hot spot association relationship analysis method and system based on theme of news word | |
CN104317881B (en) | One kind is based on the authoritative microblogging method for reordering of user's topic | |
CN110070410A (en) | A kind of population social activity analysis method and system based on big data | |
CN105912644A (en) | Network review generation type abstract method | |
CN104281565A (en) | Semantic dictionary constructing method and device |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181109 |
|
RJ01 | Rejection of invention patent application after publication |