CN107885768B

CN107885768B - User comment mining method for APP software use quality

Info

Publication number: CN107885768B
Application number: CN201710890944.4A
Authority: CN
Inventors: 姜瑛; 胡甜媛; 丁家满; 李凌宇; 汪海涛
Original assignee: Kunming University of Science and Technology
Current assignee: Kunming University of Science and Technology
Priority date: 2017-09-27
Filing date: 2017-09-27
Publication date: 2021-08-20
Anticipated expiration: 2037-09-27
Also published as: CN107885768A

Abstract

The invention relates to a user comment mining method for the use quality of APP software, and belongs to the field of APP software use quality analysis. Firstly, excavating APP software user comments reflecting the use quality by using initial comment seeds; then, for the user comment which is unsuccessfully matched with the comment seed, judging whether the user comment reflects the use quality or not according to the APP software use quality feature word list and extracting a comment mode; and finally, extracting new comment seeds based on the candidate comment mode, and further mining user comments related to the use quality. The method is beneficial to mining comments related to the use quality from a large number of APP software user comments; the method is helpful for obtaining information related to ISO/IEC 25010 use quality attributes in APP software user comments; the method is helpful for analyzing the APP software use quality.

Description

User comment mining method for APP software use quality

Technical Field

The invention relates to a user comment mining method for the use quality of APP software, and belongs to the field of APP software use quality analysis.

Background

At present, APP software is more and more diversified in type, not only provides information and simple services for people, but also can influence life, economy, even health and safety of people from various aspects. Therefore, assessing the quality of APP software becomes critical. According to the definition of the international organization for standardization ISO: "software quality is the sum of all features and all characteristics of a software product that are relevant to meeting the specified and implied requirements capabilities". Software quality includes internal quality, external quality, usage quality, and process quality. Wherein quality of use is the degree of effectiveness, efficiency, safety, and satisfaction that a particular user uses a product or system to meet the needs of the user to achieve a particular goal in a particular use environment. The use quality is the non-functional quality characteristic experienced from the user perspective in the software actual application environment, and the measurement thereof depends on the environment for measurement, and changes with different evaluators. Some users can give corresponding comments after using the APP software, the user comments are important sources of user experience information, reflect the use condition of the APP software, pay attention to specific attributes of the APP software by the users and the like, and are important bases for analyzing the use quality. In recent years, use quality analysis based on user comments has received increasing attention. How to extract comments related to the use quality from a large number of APP software user comments becomes an urgent problem to be solved.

Disclosure of Invention

Aiming at the defects in the prior art, the invention provides a user comment mining method aiming at the APP software use quality.

The technical scheme of the invention is as follows: a user comment mining method aiming at the APP software use quality,

firstly, excavating APP software user comments reflecting the use quality by using initial comment seeds; then, for the user comment which is unsuccessfully matched with the comment seed, judging whether the user comment reflects the use quality or not according to the APP software use quality feature word list and extracting a comment mode; and finally, extracting new comment seeds based on the candidate comment mode, and further mining user comments related to the use quality.

The method comprises the following specific steps:

step1, initialization data: c user comment comments after word segmentation are arranged in the comment library, wherein m is 1, s comment seeds are arranged in the comment seed library, and f use quality feature words feature _ word are arranged in the use quality feature word list;

step2, judging whether the undetermined user comments exist in the comment library: judging whether m is less than or equal to c: if yes, taking out an undetermined comment of the user_mDrawing comment_mComment mode of_mExecuting Step 3; otherwise, Step5 is executed;

wherein comment mode_m＝<mode_word_m1+…+mode_word_mn+…+ mode_word_mq，mode_speech_m1+…+mode_speech_mn+…+mode_speech_mq， mode_quality_in_use_m>N is more than or equal to 1 and less than or equal to q; mode _ word represents a word in the user comment, mode _ speech represents a part of speech of the word in the user comment, q represents the number of words/parts of speech in the comment mode, and mode _ quality _ in _ use_mRepresentative comment_mReflectingThe quality of use is one of 5 quality of use attributes in the usage quality feature vocabulary, when comment is not given_mJudging the use quality, wherein the value is unknown;

step3, judgment comment_mWhether it matches a certain comment seed:

if comment is commented on_mAnd comment seed_gSuccessful match, mode _ quality _ in _ use_m＝ seed_quality_in_use_gGive comment_mAnd (4) judging the result:<comment_m，seed_quality_in_use_g>(ii) a m + +, execute Step 2; wherein g is more than or equal to 1 and less than or equal to s;

if comment is commented on_mIf the matching with all comment seeds fails, Step4 is executed;

step4, judgment comment_mWhether to match a certain usage quality feature word:

if comment is commented on_mAnd use the quality feature word_f'Successful match, mode _ quality _ in _ use_m＝quality_in_use_f'Give comment_mAnd (4) judging the result:<comment_m， quality_in_use_f'>(ii) a Comment mode_m＝<mode_word_m1+…+mode_word_mn +…+mode_word_mq，mode_speech_m1+…+mode_speech_mn+…+mode_speech_mq， mode_quality_in_use_f'>N is more than or equal to 1 and less than or equal to q enters a candidate comment mode library, m + +, and Step2 is executed;

if comment is commented on_mMode _ quality _ in _ use, failure to match with the usage quality feature vocabulary_mComment is given as null_mAnd (4) judging the result:<comment_m，null>(ii) a m + +, execute Step 2;

wherein null represents comment_mDoes not reflect the quality of use; f' is more than or equal to 1 and less than or equal to f;

step5, calculating the text similarity of the candidate comment modes to obtain 5 comment modes with maximum word text similarity, wherein the 5 comment modes respectively belong to 5 different use qualities: { mode_max1，…，mode_max5Execution Step 6;

step6, selecting comment seeds in the candidate comment mode library to update the comment seed library: extraction comment mode_max1，…，mode_max5Seed of comment of_newkTo seed_newkAnd updating the comment seeds which are not empty in the database to the comment seed database.

The specific steps of Step3 are as follows:

step3.1, initialization data: i is 1;

step3.2, judging whether i is less than or equal to s: if yes, extracting the ith comment seed_iExecuting Step3.3; otherwise, executing Step3.7;

wherein seed_i＝<seed_word_i1+…+seed_word_ij+…+seed_word_ip，seed_speech_i1 +…+seed_speech_ij+…+seed_speech_ip，seed_dis_i，seed_quality_in_use_i>J is more than or equal to 1 and less than or equal to p, seed _ word represents a word in the comment seed, seed _ speech represents a part of speech of the word in the comment seed, p represents the number of words/parts of speech in the comment seed, and seed _ dis_iThe distance of the comment seed is the sum of the maximum distance between the evaluation object and the evaluation viewpoint and the number of words/parts of speech in the comment seed, and the value of seed _ quality _ in _ use_iThe use quality reflected by the comment seed is represented, and the use quality is one of 5 use quality attributes in the use quality feature vocabulary;

wherein, the evaluation object is also called as an emotion object or a viewpoint object and refers to a subject discussed in a certain text; the evaluation viewpoint refers to words with emotion tendencies capable of expressing the own viewpoint of the user, and is the fundamental basis for judging the emotion of the user on the evaluation object;

step3.3, comment_mWord "mode word_m1+…+mode_word_mn+…+ mode_word_mq"and comment seed_iWord "seed word_i1+…+seed_word_ij+…+seed_word_ip"match, calculate word match value commented word _ sim_mi：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.2, judging whether j is less than or equal to p: if so, executing Step3.3.3; otherwise, executing Step3.3.6;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than q: if so, executing Step3.3.5; otherwise, j + +, execute Step3.3.2;

step3.3.5, judging seed _ word_ijWhether and mode _ word_mnThe same is that: if yes, add n to the set pos _ word, j + +, execute Step3.3.2; otherwise, n + +, execute Step3.3.4;

step3.3.6 calculation of comment using equation (1)_mAnd seed_iMaximum distance between successfully matched words dis word_mi：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

In the formula (1), MAX represents the maximum value of elements in the calculation set, and MIN represents the minimum value of elements in the calculation set;

then executing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not it is less than or equal to seed _ dis_i: if so, then calculate comment using equation (2)_mAnd seed_iWord matching value commentWord _ sim_mi：

commentWord_sim_mi＝NUM(pos_word)/p (2)

In the formula (2), NUM represents the number of elements in the calculation set;

otherwise, commenting word _ sim_mi＝0；

Performing Step3.4;

step3.4, comment_mPart of speech "mode _ speech_m1+…+mode_speech_mn+…+ mode_speech_mq"and comment seed_iPart of speech "seed _ speech_i1+…+seed_speech_ij+…+ seed_speech_ip"match, calculate the part-of-speech match value commendspeech _ sim_mi：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.2, judging whether j is less than or equal to p: if so, executing Step3.4.3; otherwise, executing Step3.4.7;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than or equal to q: if so, executing Step3.4.5; otherwise, j + +, execute Step3.4.2;

step3.4.5, judging seed _ speed_ijWhether to sum mode _ speed_mnThe same is that: if so, executing Step3.4.6; otherwise, n + +, execute Step3.4.4;

step3.4.6, determine if n is greater than all elements in the set pos _ speed: if yes, add n to the set pos _ speed, j + +, execute Step3.4.2; otherwise, n + +, execute Step3.4.4;

step3.4.7 uses equation (3) to calculate comment_mAnd seed_iMaximum distance dis _ speech between successfully matched parts of speech_mi：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

Then executing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not it is less than or equal to seed _ dis_i: if so, then the comment is calculated using equation (4)_mAnd seed_iWord-part matching value commendstech sim_mi：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

Otherwise, commenting Speech _ sim_mi＝0；

Performing Step3.5;

step3.5, calculate comment using equation (5)_mAnd comment seed_iIntegrated match value com _ sim of_mi：

commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

Performing Step3.6;

step3.6, i + +, execute Step3.2;

step3.7, useFormula (6) obtains comment_mWith comment seed of all comment seeds_gMaximum integrated match value of (2) commentBoth _ sim_mg：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1≤g≤s)； (6)

Executing Step3.8;

step3.8, judge commentBoth _ sim_mgWhether or not it is greater than 0.5: if so, comment_mAnd comment seed_gMatching is successful; otherwise, comment_mFailure to match with all review seeds;

if comment is commented on_mAnd comment seed_gSuccessful match, mode _ quality _ in _ use_m＝ seed_quality_in_use_gGive comment_mAnd (4) judging the result:<comment_m，seed_quality_in_use_g>(ii) a m + +, execute Step 2;

if comment is commented on_mFailing to match all the review seeds, Step4 is performed.

The specific steps of Step4 are as follows:

step4.1, initialization variables: f ═ 1;

step4.2, judging whether f' is less than or equal to f: if so, executing Step4.3; otherwise, executing Step4.6;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f': if so, executing Step4.5; otherwise, executing Step4.4;

wherein, the f' th use quality feature word_f'The reflected use quality is quality _ in _ use_f'， 1≤f'≤f；

Step4.4, comment_mAnd use the quality feature word_f'Matching fails, f' ++, Step4.2 is executed;

step4.5, comment_mAnd use the quality feature word_f'Successful match, mode _ quality _ in _ use_m＝quality_in_use_f'Give comment_mAnd (4) judging the result:<comment_m， quality_in_use_f'>(ii) a Comment mode_m＝<mode_word_m1+…+mode_word_mn+…+ mode_word_mq，mode_speech_m1+…+mode_speech_mn+…+mode_speech_mq， mode_quality_in_use_f'>Entering a candidate comment mode library, m + +, and executing Step 2;

step4.6, comment_mMatching with the use quality feature vocabulary fails: mode _ quality _ in _ use_mComment is given as null_mAnd (4) judging the result:<comment_m，null>(ii) a m + +, execute Step 2; wherein null represents comment_mDoes not reflect the quality of use.

The specific steps of Step5 are as follows:

step5.1, initialization variables: QinU is a quality attribute, QinU is 1; wherein QinU ═ 1 represents "satisfaction"; QinU ═ 2 stands for "effectiveness"; QinU ═ 3 stands for "efficiency"; QinU ═ 4 stands for "risk avoidance"; QinU ═ 5 stands for "context coverage";

step5.2, judging whether QinU is less than or equal to 5: if so, executing Step5.3; otherwise, obtaining the comment modes with the maximum text similarity of 5 words respectively belonging to 5 different use qualities: { mode_max1，…， mode_max5Execution Step 6;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: if yes, extracting all comment modes with the quality attribute of QinU in the candidate comment mode library, and executing Step5.4; otherwise, mode_maxQinUNull, perform step5.10;

wherein, mode_maxQinUNull represents that the comment mode representing the quality of use attribute QinU is null;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: if so, executing Step5.9; otherwise, extracting the quality of use attribute as QiComment mode with text similarity not calculated in nU_mExecuting Step5.5;

the text similarity calculation comprises word text similarity calculation, part of speech text similarity calculation and comprehensive text similarity calculation;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: if so, executing Step5.8; otherwise, extracting a using quality attribute of a non-neutralized mode in QinU_mComment mode for text similarity calculation_m'M'! Executing step5.6;

step5.6, calculate mode using equation (7)_mAnd mode_m＇Word text similarity modelword sim_mm＇：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

Wherein, modeWord _ sim_mm＇Represents a mode_mAnd mode_m＇Word-text similarity of (dis word)_mm＇Represents a mode_mAnd mode_m＇Distance between word combinations of (2), lenjword_mAnd len _ word_m＇And the lengths of the characters in the two word combinations respectively;

calculate mode using equation (8)_mAnd mode_m＇Part-of-speech text similarity modeSpeech _ sim_mm＇：

modeSpeech_sim_mm＇＝1-dis_speech_mm＇/max(len_speech_m，len_speech_m＇) (8)

Wherein modeSpeech _ sim_mm＇Represents a mode_mAnd mode_m＇Part of speech text similarity of (dis _ speech)_mm＇Represents a mode_mAnd mode_m＇Distance between parts of speech combinations of (2), len _ speech_mAnd len _ speed_m＇The lengths of the characters in the two part-of-speech combinations are respectively;

then, Step5.7 is executed;

step5.7, calculate mode using equation (9)_mAnd mode_m＇Integrated text similarity modeboyh _ sim_mm＇：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

Then, Step5.5 is executed;

step5.8 calculates the comment mode using equation (10)_mAnd all other comment modes in its usage quality attributes_m＇Sum of word text similarity sum _ model word _ sim_m：

sum_modeWord_sim_m＝SUM{modeWord_sim_mm＇}(m'！＝m) (10)

In the formula (10), SUM represents the SUM of elements in the calculation set;

performing Step5.4;

step5.9, obtaining a comment mode with maximum word text similarity sum by using formula (11)_maxQinU：

sum_modeWord_sim_maxQinU＝MAX{sum_modeWord_sim_m}(1≤max≤m) (11)

Performing Step5.10;

step5.10, QinU + +, Step5.2 was performed.

The specific steps of Step6 are as follows:

step6.1, choose comment mode { mode_max1，…，mode_max5The first 3 comment modes with the maximum sum of word text similarity: mode_newk(k is more than or equal to 1 and less than or equal to 3), and extracting comment seed thereof_newk：

Step6.1.1, initializing k to 1, and executing step 6.1.2;

step6.1.2, judging whether k is less than or equal to 3: if so, extracting the comment mode_newkExecuting Step6.1.3; otherwise, executing Step6.2;

step6.1.3, judgment mode_newkWhether it is empty: if so, seed_newkNull, k + +, execute step6.1.2; otherwise, extract its comment seed_newk＝{seed_word_newk1+…+seed_word_newkj+…+ seed_word_newkp，seed_speech_newk1+…+seed_speech_newkj+…+speech_newkp，seed_dis_newk， seed_quality_in_use_newk}, k + +, perform Step6.1.2;

wherein seed _ dis_newkIs a comment mode_newkNumber of words with middle part of speech n, v, d, and a, seed _ quality _ in _ use_newkTo comment_newkReflected quality of use attributes;

step6.2 determination of seed_newkWhether the distances of the comment seeds which are not empty in the database are all updated is judged: if so, executing Step6.7; otherwise, executing Step6.3;

step6.3, extract a distance not empty and not updating the comment seed_newkSelecting and commenting seed_newkComment mode of_newkAll comment mode modes with a combined text similarity greater than 0.35_newk＇Executing Step6.4;

step6.4, calculate all comment mode_newk＇Distance dis of keywords_newk＇：

The keywords are words with parts of speech n, v, d and a in the comment mode;

step6.4.1, judge all comment mode modes_newk＇Whether the extraction of the keyword distance is finished: if so, executing Step6.5; otherwise, executing Step6.4.2;

step6.4.2 extracting a comment mode with an unextracted keyword distance_newk＇Initializing the set pos _ mode _ speed to be empty, and then executing step 6.4.3;

step6.4.3, judgment mode_newk＇Whether all parts of speech are judged is finished: if so, executing Step6.4.5; otherwise, executing Step6.4.4;

step6.4.4, extracting an undetermined part-of-speech_newk＇jJudging whether the part of speech is n or v or a or d: if yes, add j to the set pos _ mode _ speed, execute Step6.4.3; otherwise, executing Step6.4.3;

step6.4.5, calculate comment mode Using equation (12)_maxk＇Distance dis of keywords_newk＇：

dis_newk＇＝MAX{pos_mode_speech}-MIN{pos_mode_speech} (12)

Performing Step6.4.1;

step6.5, retrieve and mode using equation (13)_newkAll comment mode with integrated text similarity greater than 0.35_newk＇Dis of_newk＇Maximum value new _ seed _ dis in (1)_max：

new_seed_dis_max＝MAX{dis_newk＇} (13)

Performing Step6.6;

step6.6, judge new _ seed _ dis_maxWhether greater than seed_newkThe distance of (c): if so, seed will be_newkIs updated to new _ seed _ dis_maxExecuting Step6.2; otherwise, executing Step6.2;

step6.7, seed_newkAnd updating the comment seeds which are not empty in the database to the comment seed database.

The invention has the beneficial effects that:

(1) the comment seed is defined and matched with the comment seed, so that the comment related to the use quality can be mined from a large number of APP software user comments;

(2) according to the method, the use quality characteristic word list is defined, and the comments which are unsuccessfully matched with the comment seeds are matched with the use quality characteristic word list, so that the method is beneficial to obtaining information related to the ISO/IEC 25010 use quality attribute in the APP software user comments;

(3) the method and the device provided by the invention are beneficial to further mining the user comments related to the use quality by extracting the new comment seeds based on the candidate comment mode.

In conclusion, the method is beneficial to mining comments related to the use quality from a large number of APP software user comments; the method is helpful for obtaining information related to ISO/IEC 25010 use quality attributes in APP software user comments; the method is helpful for analyzing the APP software use quality.

Drawings

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

FIG. 2 is a Step3 in FIG. 1comment_mA specific flow chart matching with the comment seed;

FIG. 3 is the calculated word match value commentWord _ sim of FIG. 2_miA specific flowchart of (1);

FIG. 4 is a block diagram of the calculated part-of-speech match value commendsSpeech sim of FIG. 2_miA specific flowchart of (1);

FIG. 5 shows the comment in Step4 in FIG. 1_mA specific flow chart matched with the use quality characteristic words;

FIG. 6 is a detailed flowchart of Step5 in FIG. 1;

fig. 7 is a detailed flowchart of Step6 in fig. 1.

Detailed Description

Example 1: as shown in fig. 1-7, a user comment mining method for APP software usage quality,

Further, the specific steps of the method may be set as follows:

wherein comment mode_m＝<mode_word_m1+…+mode_word_mn+…+ mode_word_mq，mode_speech_m1+…+mode_speech_mn+…+mode_speech_mq， mode_quality_in_use_m>N is more than or equal to 1 and less than or equal to q; mode _ word represents a word in the user comment, mode _ speech represents a part of speech of the word in the user comment, q represents the number of words/parts of speech in the comment mode, and mode _ quality _ in _ use_mRepresentative comment_mThe reflected use quality is one of 5 use quality attributes in the use quality feature vocabulary, and the comment is not given_mJudging the use quality, wherein the value is unknown;

step3, judgment comment_mWhether it matches a certain comment seed:

step4, judgment comment_mWhether to match a certain usage quality feature word:

wherein null represents comment_mDoes not reflect the quality of use; f' is more than or equal to 1 and less than or equal tof；

Example 2: as shown in fig. 1 to 7, a user comment mining method for APP software use quality includes the following specific steps:

step1, initialization data:

the comment library has 5 (c is 5) segmented user comments, m is 1, the comment seed library has 1 (s is 1) comment seeds seed, and the quality feature word list has 15 (f is 15) feature words feature _ word;

the user comment is shown in table 2, and the usage quality feature word table is shown in table 1;

TABLE 1 word list using quality characteristics

Wherein vi represents a short verb, a represents an adjective, n represents a noun, and d represents an adverb;

TABLE 2 APP software user comment form

Numbering	APP software	User comment information	User comment information after word segmentation
				1	Youke	Can not download	Download/v is not/d is/y
2	Youke	Large internal memory	Memory/n is/a
				3	Youke	ok	ok/n
4	Youke	The memory is smaller than before	Memory/n ratio/p pre/f is small/a
				5	Youke	The memory space is really too large	Memory/n space/n true/ad too/d big/a

Initializing a comment seed in a comment seed repository₁：

seed₁＝<Installation + not + v + d + y, 3, effectiveness>

Step2, judging whether the undetermined user comments exist in the comment library:

judging whether m is less than or equal to c: 1<5, taking out an undetermined comment₁"download/v not/d not/y", extract comment₁Comment mode of₁＝<mode_word₁₁+mode_word₁₂+ mode_word₁₃，mode_speech₁₁+mode_speech₁₂+mode_speech₁₃， mode_quality_in_use₁>＝<Download + not + v + d + y, un-download>Executing Step 3;

step3, judgment comment_mWhether it matches a certain comment seed:

step3.1, initialization data: i is 1;

step3.2, judging whether i is less than or equal to s: 1-1, extract the 1 st comment seed₁＝<Installation + not + v + d + y, 3, effectiveness>Executing Step3.3;

step3.3, comment₁The word "download + not + seed" and comment seed₁The word "install + not + match" is matched, and the word matching value commentWord _ sim is calculated₁₁：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.2, judging whether j is less than or equal to p: 1<3, perform step3.3.3;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than q: 1<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₁₁: the 'downloading' is different, n is 2, and Step3.3.4 is executed;

step3.3.4, judging whether n is less than q: 2<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₁₂: "not" is not the same, n is 3, step3.3.4 is performed;

step3.3.4, judging whether n is less than q: 3, executing Step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnWhether or not the same as: word₁₁: "install" and word₁₃: "different," n ═ 4, step3.3.4 is performed;

step3.3.4, judging whether n is less than q: 4>3, j ═ 2, step3.3.2 is performed;

step3.3.2, judging whether j is less than or equal to p: 2<3, perform step3.3.3;

step3.3.3, initialization n is 1;

step3.3.4, judging whether n is less than q: 1<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₁₁: the 'downloading' is different, n is 2, and Step3.3.4 is executed;

step3.3.4, judging whether n is less than q: 2<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₁₂: not being the same, add 2 to the set pos _ word, j is 3, perform step3.3.2;

step3.3.2, judging whether j is less than or equal to p: 3, executing Step3.3.3;

step3.3.3, initialization n is 1;

step3.3.4, judging whether n is less than q: 1<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: ' s and word₁₁: the 'downloading' is different, n is 2, and Step3.3.4 is executed;

step3.3.4, judging whether n is less than q: 2<3, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: ' s and word₁₂: "not" is not the same, n is 3, step3.3.4 is performed;

step3.3.4, judging whether n is less than q: 3, executing Step3.3.5;

Step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: ' s and word₁₃: "same," add 3 to the set pos _ word, j is 4, perform step3.3.2;

step3.3.2, judging whether j is less than or equal to p: 4>3, execute step 3.3.6;

step3.3.6 calculation of comment using equation (1)₁And seed₁Maximum distance between successfully matched words dis word₁₁：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

dis_word₁₁＝3-2＝1

Performing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not it is less than or equal to seed _ dis_i：dis_word₁₁＝1<seed_dis₁Calculate comment using equation (2) as 3₁And seed₁Word matching value commentWord _ sim₁₁：

commentWord_sim_mi＝NUM(pos_word)/p (2)

commentWord_sim₁₁＝2/3＝0.67

Performing Step3.4;

step3.4, comment₁Part of speech "v + d + y" and comment seed₁The parts of speech "v + d + y" are matched, and a part of speech matching value commensuch _ sim is calculated₁₁：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.2, judging whether j is less than or equal to p: 1<3, perform step3.4.3;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<3, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₁₁: "v" is the same, Step3.4.6 is performed;

step3.4.6, determine if n is greater than all elements in the set pos _ speed: pos _ speed { }, 1 is greater than all elements in pos _ speed, 1 is added to the set pos _ speed, j ═ 2, step3.4.2 is performed;

step3.4.2, judging whether j is less than or equal to p: 2<3, perform step3.4.3;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<3, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₁₁: "v" is not the same, n ═ 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<3, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₁₂: "d" is the same, Step3.4.6 is performed;

step3.4.6, determine if n is greater than all elements in the set pos _ speed: pos _ speed ═ {1}, 2 is greater than all elements in pos _ speed, 2 is added to the set pos _ speed, j ═ 3, step3.4.2 is performed;

step3.4.2, judging whether j is less than or equal to p: 3, executing Step3.4.3;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<3, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₁₁: "v" is not the same, n ═ 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<3, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₁₂: "d" is not the same, n is 3, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 3, executing Step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₁₃: "y" is the same, Step3.4.6 is performed;

step3.4.6, determine if n is greater than all elements in the set pos _ speed: pos _ speed ═ {1, 2}, 3 is greater than all elements in pos _ speed, 3 is added to the set pos _ speed, j ═ 4, step3.4.2 is performed;

step3.4.2, judging whether j is less than or equal to p: 4>3, perform Step3.4.7;

step3.4.7, calculation of comment using equation (3)₁And seed₁Maximum distance dis _ speech between successfully matched parts of speech₁₁：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

dis_speech₁₁＝3-1＝2

Performing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not less than seed _ dis_i：dis_speech₁₁＝2<seed_dis₁Calculate comment using equation (4) as 3₁And seed₁Word-part matching value commendstech sim₁₁：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

commentSpeech_sim₁₁＝3/3＝1

Performing Step3.5;

step3.5, calculate comment using equation (5)₁And comment seed₁Integrated match value com _ sim of₁₁： commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

commentBoth_sim₁₁＝0.5*0.67+0.5*1＝0.835

Performing Step3.6;

step3.6, i ═ 2, step3.2 is performed;

step3.2, judging whether i is less than or equal to s: 2>1, performing Step3.7;

step3.7, obtaining comment using equation (6)₁With comment seed of all comment seeds_g(1< ＝g<S) maximum integrated match value commentBoth _ sim_1g：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1<＝g<＝s)； (6)

commentBoth_sim₁₁＝MAX{0.835}＝0.835 (g＝1)

Executing Step3.8;

step3.8, judge commentBoth _ sim_mgWhether or not to>0.5：0.835>0.5, comment₁And comment seed₁Matching is successful;

comment₁And comment seed₁Successful match, mode _ quality _ in _ use₁Given a comment as valid₁And (4) judging the result:<download/v No/d/y, availability>(ii) a And finishing the judgment of the 1 st comment.

After the 1 st comment is judged, the execution is started from the 2 nd comment, and Step2 is executed when m is 2;

judging whether m is less than or equal to c: 2<5, taking out an undetermined comment₂"memory/n is large/a", and comment is extracted₂Comment mode of₂＝<mode_word₂₁+mode_word₂₂，mode_speech₂₁+ mode_speech₂₂，mode_quality_in_use₂>＝<Memory + large, n + a, unknown>Executing Step 3;

step3, judgment comment_mWhether it matches a certain comment seed:

step3.1, initialization data: i is 1;

step3.3, comment₂The word "memory + big" and comment seed₁The word "install + not + match" is matched, and the word matching value commentWord _ sim is calculated₂₁：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than or equal to q: 1<2, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₂₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 2 ═ 2, step3.3.5 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₂₂: "big" is not the same, n is 3, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 3>2, j ═ 2, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₂₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₂₂: "big" is not the same, n is 3, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 3>2, j ═ 3, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₂₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₂₂: "big" is not the same, n is 3, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 3>2, j ═ 4, step3.3.2 is performed;

step3.3.6 calculation of comment using equation (1)₂And seed₁Maximum distance between successfully matched words dis word₂₁：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

dis_word₂₁＝0-0＝0

Performing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not less than seed _ dis_i：dis_word₂₁＝0<seed_dis₁Calculate comment using equation (2) as 3₂And seed₁Word matching value commentWord _ sim₂₁：

commentWord_sim_mi＝NUM(pos_word)/p (2)

commentWord_sim₂₁＝0/3＝0

Performing Step3.4;

step3.4, review comment₂Part of speech "n + a" and comment seed₁The parts of speech "v + d + y" are matched, and a part of speech matching value commensuch _ sim is calculated₂₁：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<2, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₂₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2 ═ 2, step3.4.5 is performed;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₂₂: "a" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3>2, j ═ 2, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<2, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₂₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2 ═ 2, step3.4.5 is performed;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₂₂: "a" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3>2, j ═ 3, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<2, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₂₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2 ═ 2, step3.4.5 is performed;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₂₂: "a" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3>2, j ═ 4, step3.4.2 is performed;

step3.4.7, calculation of comment using equation (3)₂And seed₁Maximum distance dis _ speech between successfully matched parts of speech₂₁：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

dis_speech₂₁＝0-0＝0

Performing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not less than seed _ dis_i：dis_speech₂₁＝0<seed_dis₁Calculate comment using equation (4) as 3₂And seed₁Word-part matching value commendstech sim₂₁：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

commentSpeech_sim₂₁＝0/3＝0

Performing Step3.5;

step3.5, calculate comment using equation (5)₂The composite match value commentBoth_sim₂₁： commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

commentBoth_sim₂₁＝0+0＝0

Performing Step3.6;

step3.6, i ═ 2, step3.2 is performed;

step3.2, judging whether i is less than or equal to s: 2>1, performing Step3.7;

step3.7, obtaining comment using equation (6)₂With comment seed of all comment seeds_g(1< ＝g<S) maximum integrated match value commentBoth _ sim_2g：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1<＝g<＝s)； (6)

commentBoth_sim₂₁＝Max{0}＝0 (g＝1)

Executing Step3.8;

step3.8, judge commentBoth _ sim_mgWhether or not to>0.5：0<0.5, comment₂Failure to match with all review seeds;

comment₂If the matching with all comment seeds fails, Step4 is executed;

step4, judgment comment_mWhether to match a certain usage quality feature word:

step4.1, initialization variables: f ═ 1;

step4.2, judging whether f' is less than or equal to f: 1<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₂Does not contain the use quality feature word₁: "fail/vi", execute step 4.4;

step4.4, comment₂And use the quality feature word₁Matching fails, f' is 2, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 2<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₂Does not contain the use quality feature word₂: "error/vi", perform Step4.4;

step4.4, comment₂And use the quality feature word₂Matching fails, f' is 3, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 3<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₂Does not contain the use quality feature word₃: "Utility/a", perform Step4.4;

step4.4, comment₂And use the quality feature word₃Matching fails, f' is 4, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 4<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₂Does not contain the use quality feature word₄: "function/n", perform Step4.4;

step4.4, comment₂And use the quality feature word₄Matching fails, f' is 5, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 5<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₂Including the use quality feature word₅: "memory/n", executing Step4.5;

step4.5, comment₂And use the quality feature word₅Matching is successful;

comment₂And use the quality feature word₅Successful match, mode _ quality _ in _ use₂Given a comment as efficiency₂And (4) judging the result:<memory/n is large/a, efficiency>(ii) a Comment mode₂＝<Memory + large, n + a, efficiency>Entering a candidate comment mode library; and finishing the judgment of the 2 nd comment.

After the 2 nd comment is judged, the execution is started from the 3 rd comment, and Step2 is executed when m is 3;

judging whether m is less than or equal to c: 3<5, taking out an undetermined comment₃Extract comment as ok/n₃Comment mode of₃＝<mode_word₃₁，mode_speech₃₁，mode_quality_in_use₃> ＝<ok，n，unknown>Executing Step 3;

step3, judgment comment_mWhether it matches a certain comment seed:

step3.1, initialization data: i is 1;

step3.3, comment₃The word "ok" with comment seed₁The word "install + not + match" is matched, and the word matching value commentWord _ sim is calculated₃₁：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than or equal to q: 1 to 1, performing step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₃₁: "ok" is not the same, n ═ 2, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 2>1, j ═ 2, Step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₃₁: "ok" is not the same, n ═ 2, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 2>1, j ═ 3, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₃₁: "ok" is not the same, n ═ 2, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 2>1, j ═ 4, step3.3.2 is performed;

step3.3.6 calculation of comment using equation (1)₃And seed₁Maximum distance between successfully matched words dis word₃₁：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

dis_word₃₁＝0-0＝0

Performing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not less than seed _ dis_i：dis_word₃₁＝0<seed_dis₁Calculate comment using equation (2) as 3₃And seed₁Word matching value commentWord _ sim₃₁：

commentWord_sim_mi＝NUM(pos_word)/p (2)

commentWord_sim₃₁＝0/3＝0

Performing Step3.4;

step3.4, comment₃Part of speech "n" and comment seed₁The parts of speech "v + d + y" are matched, and a part of speech matching value commensuch _ sim is calculated₃₁：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1 to 1, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₃₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2>1, j ═ 2, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1 to 1, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₃₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2>1, j ═ 3, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1 to 1, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₃₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2>1, j ═ 4, step3.4.2 is performed;

step3.4.7, calculation of comment using equation (3)₃And seed₁Maximum distance dis _ speech between successfully matched parts of speech₃₁：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

dis_speech₃₁＝0-0＝0

Performing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not less than seed _ dis_i：dis_speech₃₁＝0<seed_dis₁Calculate comment using equation (4) as 3₃And seed₁Word-part matching value commendstech sim₃₁：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

commentSpeech_sim₃₁＝0/3＝0

Performing Step3.5;

step3.5, calculate comment using equation (5)₃Composite match value commentBoth _ sim₃₁： commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

commentBoth_sim₃₁＝0+0＝0

Performing Step3.6;

step3.6, i ═ 2, step3.2 is performed;

step3.2, judging whether i is less than or equal to s: 2>1, performing Step3.7;

step3.7, obtaining comment using equation (6)₃With comment seed of all comment seeds_g(1< ＝g<S) maximum integrated match value commentBoth _ sim_3g：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1<＝g<＝s)； (6)

commentBoth_sim₃₁＝Max{0}＝0 (g＝1)

Executing Step3.8;

step3.8, judge commentBoth _ sim_miWhether or not to>0.5：0<0.5, comment₃Failure to match with all review seeds;

comment₃If the matching with all comment seeds fails, Step4 is executed;

step4, judgment comment_mWhether to match a certain usage quality feature word:

step4.1, initialization variables: f ═ 1;

step4.2, judging whether f' is less than or equal to f: 1<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₁: "fail/vi", execute step 4.4;

step4.4, comment₃And use the quality feature word₁Matching fails, f' is 2, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 2<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₂: "error/vi", perform Step4.4;

step4.4, comment₃And use the quality feature word₂Matching fails, f' is 3, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 3<15, perform step4.3;

step4.3, judgmentcomment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₃: "Utility/a", perform Step4.4;

step4.4, comment₃And use the quality feature word₃Matching fails, f' is 4, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 4<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₄: "function/n", perform Step4.4;

step4.4, comment₃And use the quality feature word₄Matching fails, f' is 5, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 5<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₅: "memory/n", execute step 4.4;

step4.4, comment₃And use the quality feature word₅Matching fails, f' is 6, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 6<15, perform Step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₆: "efficiency/n", perform Step4.4;

step4.4, comment₃And use the quality feature word₆Matching fails, f' is 7, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 7<15, perform Step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₇: "speed/n", execute Step4.4;

step4.4, comment₃And use the quality feature word₇Matching fails, f' is 8, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 8<15, perform Step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₈: "safe/a", perform step 4.4;

step4.4, comment₃And use the quality feature word₈Matching fails, f' is 9, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 9<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₉: "charge/vi", perform step 4.4;

step4.4, comment₃And use the quality feature word₉Matching fails, f' is 10, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 10<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₁₀: "Ad/n", execute Step4.4;

step4.4, comment₃And use the quality feature word₁₀Matching fails, f' is 11, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 11<15, perform step4.3;

step4.3, judgment comment_mIn isWhether to include the use quality feature word_f'：comment₃Does not contain the use quality feature word₁₁: "version/n", perform Step4.4;

step4.4, comment₃And use the quality feature word₁₁Matching fails, f' is 12, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 12<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₁₂: "product/n", perform Step4.4;

step4.4, comment₃And use the quality feature word₁₂Matching fails, f' is 13, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 13<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₁₃: "system/n", perform Step4.4;

step4.4, comment₃And use the quality feature word₁₃Matching fails, f' is 14, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 14<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₃Does not contain the use quality feature word₁₄: "good/a", perform step 4.4;

step4.4, comment₃And use the quality feature word₁₄Matching fails, f' is 15, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 15, executing Step4.3;

step4.3, judgment comment_mWhether or notInvolving the use of quality feature words feature word_f'：comment₃Does not contain the use quality feature word₁₅: "difference/a", perform step4.4;

step4.4, comment₃And use the quality feature word₁₅Matching fails, f ═ 16, step4.2 is performed;

step4.2, judging whether f' is less than or equal to f: 16>15, perform Step4.6;

step4.6, comment₃Matching with the use quality feature vocabulary fails;

comment₃Mode _ quality _ in _ use, failure to match with the usage quality feature vocabulary₃Comment is given as null₃And (4) judging the result:<comment₃，null>(ii) a And the 3 rd comment is judged to be finished.

After the 3 rd comment is judged, the execution is started from the 4 th comment, and Step2 is executed when m is 4;

judging whether m is less than or equal to c: 4<5, taking out an undetermined comment₄Extracting comment as "memory/n ratio/p pre/f small/a₄Comment mode of₄＝<mode_word₄₁+mode_word₄₂+ mode_word₄₃+mode_word₄₄，mode_speech₄₁+mode_speech₄₂+mode_speech₄₃+ mode_speech₄₄，mode_quality_in_use₄>＝<Memory + is smaller than + before, n + p + f + a, undnown>Executing Step 3;

step3, judgment comment_mWhether it matches a certain comment seed:

step3.1, initialization data: i is 1;

step3.3, comment₄The word "memory + less than + before + and comment seed₁The word "install + not + match" is matched, and the word matching value commentWord _ sim is calculated₄₁：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than or equal to q: 1<4, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₄₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 2<4, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₄₂: the ratio is different, n is 3, and Step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 3<4, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₄₃: "before" is not the same, n is 4, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 4-4, performing step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₄₄: "small" is not the same, n is 5, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 5>4, j ═ 2, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₄₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₄₂: the ratio is different, n is 3, and Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₄₃: "before" is not the same, n is 4, step3.3.4 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₄₄: "small" is not the same, n is 5, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 5>4, j ═ 3, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₄₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₄₂: the ratio is different, n is 3, and Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₄₃: "before" is not the same, n is 4, step3.3.4 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₄₄: "small" is not the same, n is 5, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 5>4, j ═ 4, step3.3.2 is performed;

step3.3.6 calculation of comment using equation (1)₄And seed₁Maximum distance between successfully matched words dis word₄₁：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

dis_word₄₁＝0-0＝0

Performing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not less than seed _ dis_i：dis_word₄₁＝0<seed_dis₁Calculate comment using equation (2) as 3₄And seed₁Word matching value commentWord _ sim₄₁：

commentWord_sim_mi＝NUM(pos_word)/p (2)

commentWord_sim₄₁＝0/3＝0

Performing Step3.4;

step3.4, comment₄Part of speech "n + p + f + a" and comment seed₁The parts of speech "v + d + y" are matched, and a part of speech matching value commensuch _ sim is calculated₄₁：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₄₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₄₂: "p" is not the same, n ═ 3, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 3<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₄₃: "f" is not the same, n is 4, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 4-4, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₄₄: "a" is not the same, n is 5, and step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 5>4, j ═ 2, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₄₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₄₂: "p" is not the same, n ═ 3, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 3<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₄₃: "f" is not the same, n is 4, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 4-4, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₄₄: "a" is not the same, n is 5, and step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 5>4, j ═ 3, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₄₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₄₂: "p" is not the same, n ═ 3, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 3<4, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₄₃：“f”Different, n is 4, execute step3.4.4;

step3.4.4, judging whether n is less than q: 4-4, performing step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₄₄: "a" is not the same, n is 5, and step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 5>4, j ═ 4, step3.4.2 is performed;

step3.4.7, calculation of comment using equation (3)₄And seed₁Maximum distance dis _ speech between successfully matched parts of speech₄₁：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

dis_speech₄₁＝0-0＝0

Performing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not less than seed _ dis_i：dis_speech₄₁＝0<seed_dis₁Calculate comment using equation (4) as 3₄And seed₁Word-part matching value commendstech sim₄₁：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

commentSpeech_sim₄₁＝0/3＝0

Performing Step3.5;

step3.5, calculate comment using equation (5)₄Composite match value commentBoth _ sim₄₁： commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

commentBoth_sim₄₁＝0+0＝0

Performing Step3.6;

step3.6, i ═ 2, step3.2 is performed;

step3.2, judging whether i is less than or equal to s: 2>1, performing Step3.7;

step3.7, Utility maleEquation (6) get comment₄With comment seed of all comment seeds_g(1< ＝g<S) maximum integrated match value commentBoth _ sim_4g：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1<＝g<＝s)； (6)

commentBoth_sim₄₁＝Max{0}＝0 (g＝1)

Executing Step3.8;

step3.8, judge commentBoth _ sim_miWhether or not to>0.5：0<0.5, comment₄Failure to match with all review seeds;

comment₄If the matching with all comment seeds fails, Step4 is executed;

step4, judgment comment_mWhether to match a certain usage quality feature word:

step4.1, initialization variables: f ═ 1;

step4.2, judging whether f' is less than or equal to f: 1<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₄Does not contain the use quality feature word₁: "fail/vi", execute step 4.4;

step4.4, comment₄And use the quality feature word₁Matching fails, f' is 2, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 2<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₄Does not contain the use quality feature word₂: "error/vi", perform Step4.4;

step4.4, comment₄And use the quality feature word₂Fail to match, f' is 3, Step is executed4.2；

Step4.2, judging whether f' is less than or equal to f: 3<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₄Does not contain the use quality feature word₃: "Utility/a", perform Step4.4;

step4.4, comment₄And use the quality feature word₃Matching fails, f' is 4, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 4<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₄Does not contain the use quality feature word₄: "function/n", perform Step4.4;

step4.4, comment₄And use the quality feature word₄Matching fails, f' is 5, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 5<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₄Including the use quality feature word₅: "memory/n", executing Step4.5;

step4.5, comment₄And use the quality feature word₅Matching is successful;

comment₄And use the quality feature word₅Successful match, mode _ quality _ in _ use₄Given a comment as efficiency₄And (4) judging the result:<memory/n ratio/p front/f is small/a, efficiency>(ii) a Comment mode₄＝<Memory + is smaller than + before, n + p + f + a, efficiency>Entering a candidate comment mode library; the judgment of item 4 was completed.

After the 4 th comment is judged, the execution is started from the 5 th comment, and Step2 is executed when m is 5;

judging whether m is less than or equal to c: taking out one piece of comment which is not judged if 5 is 5₅Extracting comment as "memory/n space/n true/ad too/d big/a₅Comment mode of₅＝<mode_word₅₁+ mode_word₅₂+mode_word₅₃+mode_word₅₄+mode_word₅₅，mode_speech₅₁+ mode_speech₅₂+mode_speech₅₃+mode_speech₅₄+mode_speech₅₅， mode_quality_in_use₅>＝<Memory + space + true + too + big, n + n + ad + d + a, unbnown>Executing Step 3;

step3, judgment comment_mWhether it matches a certain comment seed:

step3.1, initialization data: i is 1;

step3.3, comment₅The word "memory + space + really + too + big" and comment seed₁The word "install + not + match" is matched, and the word matching value commentWord _ sim is calculated₅₁：

Step3.3.1, initializing j to 1, and setting pos _ word to be null;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.4, judging whether n is less than or equal to q: 1<5, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₅₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 2<5, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₅₂: "space" is not the same, n is 3, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 3<5, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₅₃: "do" is not the same, n is 4, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 4<5, perform step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₅₄: "too" is not the same, n ═ 5, step3.3.4 is performed;

step3.3.4, judging whether n is less than or equal to q: 5, executing Step3.3.5;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₁: "install" and word₅₅: "big" is not the same, n is 6, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 6>5, j ═ 2, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₅₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₅₂: "space" is not the same, n is 3, step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₅₃: "do" is not the same, n is 4, step3.3.4 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₅₄: "too" is not the same, n ═ 5, step3.3.4 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₂: not and word₅₅: "big" is not the same, n is 6, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 6>5, j ═ 3, step3.3.2 is performed;

step3.3.3, initializing n to 1, and executing step 3.3.4;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₅₁: the memories are different, and when n is 2, Step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₅₂: "space" is not the same, n is 3, step3.3.4 is executed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₅₃：“Indeed "not the same, n ═ 4, step3.3.4 was performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₅₄: "too" is not the same, n ═ 5, step3.3.4 is performed;

step3.3.5, judgment word_ijWhether to sum word_mnThe same is that: word₁₃: 's' and word₅₅: "big" is not the same, n is 6, step3.3.4 is executed;

step3.3.4, judging whether n is less than or equal to q: 6>5, j ═ 4, step3.3.2 is performed;

step3.3.6 calculation of comment using equation (1)₅And seed₁Maximum distance between successfully matched words dis word₅₁：

dis_word_mi＝MAX{pos_word}-MIN{pos_word} (1)

dis_word₅₁＝0-0＝0

Performing Step3.3.7;

step3.3.7, determine dis _ word_miWhether or not less than seed _ dis_i：dis_word₅₁＝0<seed_dis₁Calculate comment using equation (2) as 3₅And seed₁Word matching value commentWord _ sim₅₁：

commentWord_sim_mi＝NUM(pos_word)/p (2)

commentWord_sim₅₁＝0/3＝0

Performing Step3.4;

step3.4, comment₅Part of speech "n + n + ad + d + a" and comment seed₁The parts of speech "v + d + y" are matched, and a part of speech matching value commensuch _ sim is calculated₅₁：

Step3.4.1, initializing j to 1, and setting pos _ speed to be empty;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₅₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₅₂: "n" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₅₃: "ad" is not the same, n ═ 4, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 4<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₅₄: "d" is not the same, n is 5, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 5, executing Step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₁: "v" and speed₅₅: "a" is not the same, n is 6, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 6>5, j ═ 2, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₅₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₅₂: "n" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₅₃: "ad" is not the same, n ═ 4, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 4<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₂: "d" and speed₅₄: "d" is the same, Step3.4.6 is performed;

step3.4.6, determine if n is greater than all elements in the set pos _ speed: pos _ speed { }, 4 is greater than all elements in pos _ speed, 4 is added to the set pos _ speed, j ═ 3, step3.4.2 is performed;

step3.4.3, initializing n to 1, and executing step 3.4.4;

step3.4.4, judging whether n is less than q: 1<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₅₁: "n" is not the same, n is 2, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 2<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₅₂: "n" is not the same, n is 3, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 3<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₅₃: "ad" is not the same, n ═ 4, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 4<5, perform step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₅₄: "d" is not the same, n is 5, step3.4.4 is performed;

step3.4.4, judging whether n is less than q: 5, executing Step3.4.5;

step3.4.5, judgment of speed_ijWhether or not to harmonize with speed_mnThe same is that: speed₁₃: "y" and speed₅₅: "a" is not the same, n is 6, step3.4.4 is executed;

step3.4.4, judging whether n is less than q: 6>5, j ═ 4, step3.4.2 is performed;

step3.4.7, calculation of comment using equation (3)₅And seed₁Maximum distance dis _ speech between successfully matched parts of speech₅₁：

dis_speech_mi＝MAX{pos_speech}-MIN{pos_speech} (3)

dis_speech₅₁＝4-4＝0

Performing Step3.4.8;

step3.4.8, determine dis _ speed_miWhether or not less than seed _ dis_i：dis_speech₅₁＝0<seed_dis₁Calculate comment using equation (4) as 3₅And seed₁Word-part matching value commendstech sim₅₁：

commentSpeech_sim_mi＝NUM(pos_speech)/p (4)

commentSpeech_sim₅₁＝1/3＝0.33

Performing Step3.5;

step3.5, calculate comment using equation (5)₅Composite match value commentBoth _ sim₅₁： commentBoth_sim_mi＝0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi (5)

commentBoth_sim₅₁＝0+0.165＝0.165

Performing Step3.6;

step3.6, i ═ 2, step3.2 is performed;

step3.2, judging whether i is less than or equal to s: 2>1, performing Step3.7;

step3.7, obtaining comment using equation (6)₅With comment seed of all comment seeds_g(1< ＝g<S) maximum integrated match value commentBoth _ sim_5g：

commentBoth_sim_mg

＝MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1<＝g<＝s)； (6)

commentBoth_sim₅₁＝Max{0.165}＝0.165 (g＝1)

Executing Step3.8;

step3.8, judge commentBoth _ sim_miWhether or not to>0.5：0.165<0.5, comment₅Failure to match with all review seeds;

comment₅If the matching with all comment seeds fails, Step4 is executed;

step4, judgment comment_mWhether to match a certain usage quality feature word:

step4.1, initialization variables: f ═ 1;

step4.2, judging whether f' is less than or equal to f: 1<15, perform step4.3;

step4.3, judgment comment_mWhether or not to include quality of useFeature word_f'：comment₅Does not contain the use quality feature word₁: "fail/vi", execute step 4.4;

step4.4, comment₅And use the quality feature word₁Matching fails, f' is 2, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 2<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₅Does not contain the use quality feature word₂: "error/vi", perform Step4.4;

step4.4, comment₅And use the quality feature word₂Matching fails, f' is 3, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 3<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₅Does not contain the use quality feature word₃: "Utility/a", perform Step4.4;

step4.4, comment₅And use the quality feature word₃Matching fails, f' is 4, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 4<15, perform step4.3;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_f'：comment₅Does not contain the use quality feature word₄: "function/n", perform Step4.4;

step4.4, comment₅And use the quality feature word₄Matching fails, f' is 5, and step4.2 is executed;

step4.2, judging whether f' is less than or equal to f: 5<15, perform step4.3;

step4.3, judgment comment_mWhether or not to include the use quality feature word feature_word_f'：comment₅Including the use quality feature word₅: "memory/n", executing Step4.5;

step4.5, comment₅And use the quality feature word₅Matching is successful;

comment₅And use the quality feature word₅Successful match, mode _ quality _ in _ use₅Given a comment as efficiency₅And (4) judging the result:<memory/n space/n true/ad too/d greater/a, efficiency>(ii) a Comment mode₅＝<Memory + space + true + too + big, n + n + ad + d + a, efficiency>Entering a candidate comment mode library; and 5, finishing the judgment of the comment.

After the 5 th comment is judged, if no comment which is not judged exists in the comment library, Step5 is executed;

at this time, there are three candidate comment modes:

mode₂＝<memory + large, n + a, efficiency>；

mode₄Memory + is smaller than + before, n + p + f + a, efficiency >;

mode₅memory + space + really + too + big, n + n + ad + d + a, efficiency >;

step5, calculating the text similarity of the candidate comment modes:

step5.1, initialization variables: QinU is a quality of use attribute, QinU 1,

wherein QinU ═ 1 represents "satisfaction"; QinU ═ 2 stands for "effectiveness"; QinU ═ 3 stands for "efficiency"; QinU ═ 4 stands for "risk avoidance"; QinU ═ 5 stands for "context coverage";

step5.2, judging whether QinU is less than or equal to 5: 1<5, perform step5.3;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: there is no comment pattern using a quality attribute of 1 in the candidate comment pattern library, mode_max1Null, perform step5.10;

step5.10, QinU ═ 2, step5.2 was performed;

step5.2, judging whether QinU is less than or equal to 5: 2<5, perform step5.3;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: there is no comment pattern using a quality attribute of 2 in the candidate comment pattern library, mode_max2Null, perform step5.10;

step5.10, QinU ═ 3, step5.2 was performed;

step5.2, judging whether QinU is less than or equal to 5: 3<5, execute step5.3;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: the candidate comment mode library has a comment mode with a use quality attribute of 3, all comment modes with the use quality attribute of 3 in the candidate comment mode library are extracted, and Step5.4 is executed;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: extracting a comment mode with the quality attribute of 3 and the text similarity of all comment modes in the category 3 not calculated₂Executing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₂And extracting a mode with the use quality attribute of 3 and the mode with the use quality attribute of 3 without calculating the text similarity of other comment modes with the use quality attribute of 3₂Comment mode for text similarity calculation₄Executing Step5.6;

step5.6, calculate mode using equation (7)₂And mode₄Word text similarity modelword sim₂₄：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₂₄＝0.33

Calculate mode using equation (8)₂And mode₄Part-of-speech text similarity modeSpeech _ sim₂₄：

modeSpeech_sim₂₄＝0.43

Performing Step5.7;

step5.7, calculate mode using equation (9)₂And mode₄Integrated text similarity modeboyh _ sim₂₄：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₂₄＝0.38

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in the category QinU is calculated is finished: mode₂Extracting a category 3 un-neutralized mode when the similarity of the text with other comment modes in the category 3 is not calculated₂Comment mode for text similarity calculation₅Executing Step5.6;

step5.6, calculate mode using equation (7)₂And mode₅Word text similarity modelword sim₂₅：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₂₅＝0.38

Calculate mode using equation (8)₂And mode₅Part-of-speech text similarity modeSpeech _ sim₂₅：

modeSpeech_sim₂₅＝0.3

Performing Step5.7;

step5.7, calculate mode using equation (9)₂And mode₅Integrated text similarity modeboyh _ sim₂₅：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₂₅＝0.34

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in the category QinU is calculated is finished: mode₂After the similarity calculation with other comment mode texts in the category 3 is finished, Step5.8 is executed;

step5.8, calculate comment mode using equation (10)₂Sum of word text similarity for all other comment modes in its usage quality attribute sum _ mode word _ sim₂：

sum_modeWord_sim_m＝SUM{modeWord_sim_mm＇}(m'！＝m) (10)

sum_modeWord_sim₂＝modeWord_sim₂₄+modeWord_sim₂₅＝0.33+0.38＝0.71

Performing Step5.4;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: extracting a comment mode with the quality attribute of 3 and the text similarity of all comment modes in the category 3 not calculated₄Executing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₄And extracting a mode with the use quality attribute of 3 and the mode with the use quality attribute of 3 without calculating the text similarity of other comment modes with the use quality attribute of 3₄Comment mode for text similarity calculation₂Executing Step5.6;

step5.6, calculate mode using equation (7)₄And mode₂Word text similarity modelword sim₄₂：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₄₂＝0.33

Calculate mode using equation (8)₄And mode₂Part-of-speech text similarity modeSpeech _ sim₄₂：

modeSpeech_sim₄₂＝0.43

Performing Step5.7;

step5.7, calculate mode using equation (9)₄And mode₂Integrated text similarity modeboyh _ sim₄₂：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₄₂＝0.38

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₄And extracting a mode with the use quality attribute of 3 and the mode with the use quality attribute of 3 without calculating the text similarity of other comment modes with the use quality attribute of 3₄Comment mode for text similarity calculation₅Executing Step5.6;

step5.6, calculate mode using equation (7)₄And mode₅Word text similarity modelword sim₄₅：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₄₅＝0.25

Calculate mode using equation (8)₄And mode₅Part-of-speech text similarity modeSpeech _ sim₄₅：

modeSpeech_sim₄₅＝0.5

Performing Step5.7;

step5.7, calculate mode using equation (9)₄And mode₅HeddleCombined text similarity modeBoth _ sim₄₅：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₄₅＝0.375

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₄And after the similarity of the text of other comment modes with the quality attribute of 3 is calculated, executing Step5.8;

step5.8, calculate comment mode using equation (10)₄Sum of word text similarity for all other comment modes in its usage quality attribute sum _ mode word _ sim₄：

sum_modeWord_sim_m＝SUM{modeWord_sim_mm＇}(m'！＝m) (10)

sum_modeWord_sim₄＝modeWord_sim₄₂+modeWord_sim₄₅＝0.33+0.25＝0.58

Performing Step5.4;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: extracting the comment mode with the quality attribute of 3 and the text similarity of all the comment modes not calculated₅Executing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₅And extracting a mode with the use quality attribute of 3 and the mode with the use quality attribute of 3 without calculating the text similarity of other comment modes with the use quality attribute of 3₅Comment mode for text similarity calculation₂Executing Step5.6;

step5.6, calculate mode using equation (7)₅And mode₂Word text similarity modelword sim₅₂：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₅₂＝0.38

Calculate mode using equation (8)₅And mode₂Part-of-speech text similarity modeSpeech _ sim₅₂：

modeSpeech_sim₅₂＝0.3

Performing Step5.7;

step5.7, calculate mode using equation (9)₅And mode₂Integrated text similarity modeboyh _ sim₅₂：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₅₂＝0.34

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₅And extracting a mode with the use quality attribute of 3 and the mode with the use quality attribute of 3 without calculating the text similarity of other comment modes with the use quality attribute of 3₅Comment mode for text similarity calculation₄Executing Step5.6;

step5.6, calculate mode using equation (7)₅And mode₄Word text similarity modelword sim₅₄：

modeWord_sim_mm＇＝1-dis_word_mm＇/max(len_word_m，len_word_m＇) (7)

modeWord_sim₅₄＝0.25

Calculate mode using equation (8)₅And mode₄Part-of-speech text similarity modeSpeech _ sim₅₄：

modeSpeech_sim₅₄＝0.5

Performing Step5.7;

step5.7, calculate mode using equation (9)₅And mode₄Integrated text similarity modeboyh _ sim₅₄：

modeBoth_sim_mm＇＝0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇ (9)

modeBoth_sim₅₄＝0.375

Performing Step5.5;

step5.5, judge mode_mAnd whether the similarity of the text of other comment modes in QinU is calculated by using the quality attribute: mode₅And after the similarity of the text of other comment modes with the quality attribute of 3 is calculated, executing Step5.8;

step5.8, calculate comment mode using equation (10)₅Sum of word text similarity for all other comment modes in its usage quality attribute sum _ mode word _ sim₅：

sum_modeWord_sim_m＝SUM{modeWord_sim_mm＇}(m'！＝m) (10)

sum_modeWord_sim₅＝modeWord_sim ₅₂+modeWord_sim ₅₄＝0.38+0.25＝0.63

Performing Step5.4;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: after the similarity of all comment mode texts with the quality attribute of 3 is calculated, executing Step5.9;

step5.9, obtaining a comment mode with the maximum sum of word text similarity in the quality attribute of 3 by using a formula (11)_maxQinU：

sum_modeWord_sim_maxQinU＝MAX{sum_modeWord_sim_m}(1<＝max<＝m) (11)

mode_max3＝MAX{sum_modeWord_sim₂，sum_modeWord_sim₄， sum_modeWord_sim₅}＝MAX{0.71，0.58，0.63}＝mode₂

Performing Step5.10;

step5.10, QinU ═ 4, step5.2 was performed;

step5.2, judging whether QinU is less than or equal to 5: 4<5, perform step5.3;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: there is no comment pattern using a quality attribute of 4 in the candidate comment pattern library, mode_max4Null, perform step5.10;

step5.10, QinU ═ 5, step5.2 was performed;

step5.2, judging whether QinU is less than or equal to 5: 5, executing Step5.3;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: there is no comment pattern using a quality attribute of 5 in the candidate comment pattern library, mode_max5Null, perform step5.10;

step5.10, QinU ═ 6, step5.2 was performed;

step5.2, judging whether QinU is less than or equal to 5: 6>5, obtaining 5 word text similarity maximum comment modes which respectively belong to 5 different use qualities: { mode_max1，…，mode_max5}＝{null，null，mode₂Null, null }, Step6 is executed;

step6, selecting comment seeds in the candidate comment mode library to update the comment seed library:

step6.1, choose comment mode null, null, mode₂The first 3 comment modes with the largest sum of word text similarities in null, null } are as follows: mode_new1＝null、mode_new2＝null、mode_new3＝mode₂Extract its comment seed_new1、seed_new2、seed_new3：

Step6.1.1, initializing k to 1, and executing step 6.1.2;

step6.1.2, judging whether k is less than or equal to 3: 1<Extracting comment mode_new1Executing Step6.1.3;

step6.1.3, judgment mode_new1Whether it is empty: mode_new1Is empty, seed_new1Null, k 2, step6.1.2;

step6.1.2, judging whether k is less than or equal to 3: 2<Extracting comment mode_new2Executing Step6.1.3;

step6.1.3, judgment mode_new2Whether it is empty: mode_new2Is empty, seed_new2Null, k 3, step6.1.2;

step6.1.2, judging whether k is less than or equal to 3: extracting comment mode with 3 ═ 3_new3Executing Step6.1.3;

step6.1.3, judgment mode_new3Whether it is empty: mode_new3Not null, comment mode_new3The number of words with the part of speech n, v, d and a is 2, and the comment seed is extracted_new3Step6.1.2 is performed, where { memory + is large, n + a, 2, efficiency }, and k is 4;

step6.1.2, judging whether k is less than or equal to 3: 4>3, performing Step6.2;

step6.2 determination of seed_newkWhether the distances of the comment seeds which are not empty in the database are all updated is judged: seed_newkIf the distances of the comment seeds which are not empty in the comment database are not completely updated, executing Step6.3;

step6.3, extract a distance not empty and not updating the comment seed_new3Selecting and commenting seed_new3Comment mode of_new3All comment modes with a combined text similarity greater than 0.35: mode₄Where memory + is smaller than + before, n + p + f + a, efficiency > is performed step6.4;

step6.4, calculate all comment mode_newk＇Distance dis of keywords_newk＇：

Step6.4.1, judge all comment mode modes_newk＇Whether the extraction of the keyword distance is finished: comment mode₄Executing Step6.4.2 if the distance of the key words is not extracted;

step6.4.2, extracting a comment mode with the distance of the unextracted keyword₄Initializing the set pos _ mode _ speed to be null;

step6.4.3, judgment mode_newk＇Whether all parts of speech are judged is finished: mode₄After all parts of speech are not judged, executing Step6.4.4;

step6.4.4, extracting an undetermined part of speech mode _ speech₄₁And judging whether the part of speech is: n or v or a or d: speed₄₁For n, add 1 to the set pos _ mode _ speed, perform step 6.4.3;

step6.4.4, extracting an undetermined part of speech mode _ speech₄₂And judging whether the part of speech is: n or v or a or d: speed₄₂Is p, speed₄₂Instead of n or v or a or d, Step6.4.3 is performed;

step6.4.4, extracting an undetermined part of speech mode _ speech₄₃And judging whether the part of speech is: n or v or a or d: speed₄₃Is f, speed₄₃Instead of n or v or a or d, Step6.4.3 is performed;

step6.4.4, extracting an undetermined part of speech mode _ speech₄₄And judging whether the part of speech is: n or v or a or d: speed₄₄To a, add 4 to the set pos _ mode _ speed, perform step 6.4.3;

step6.4.3, judgment mode_newk＇Whether all parts of speech are judged is finished: mode₄After all parts of speech are judged, executing Step6.4.5;

step6.4.5, calculate comment mode Using equation (12)₄Distance dis of keywords₄：

dis_newk＇＝MAX{pos_mode_speech}-MIN{pos_mode_speech} (12)

dis₄＝4-1＝3

Performing Step6.4.1;

step6.4.1, judge all comment mode modes_newk＇Whether the extraction of the keyword distance is finished: all comment mode_newk＇After the keyword distance is extracted, executing Step6.5;

step6.5, retrieve and mode using equation (13)_new3All comment mode with integrated text similarity greater than 0.35_newk＇Dis of_newk＇Maximum value new _ seed _ dis in (1)_max：

new_seed_dis_max＝MAX{dis_newk＇} (13)

new_seed_dis_max＝MAX{3}＝3

Performing Step6.6;

step6.6, judge new _ seed _ dis_maxWhether greater than seed_newkThe distance of (c): 3>seed_dis _new32, seed_new3Update the distance to 3, and execute step 6.2;

step6.2 determination of seed_newkWhether the distances of the comment seeds which are not empty in the database are all updated is judged: seed_newkThe distances of the comment seeds which are not empty in the comment database are all updated, and Step6.7 is executed;

step6.7, seed_newkSeed of comments not empty: seed_new3The { memory + large, n + a, 3, efficiency } is updated to the review seed repository.

While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.

Claims

1. A user comment mining method aiming at APP software use quality is characterized in that:

firstly, excavating APP software user comments reflecting the use quality by using initial comment seeds; then, for the user comment which is unsuccessfully matched with the comment seed, judging whether the user comment reflects the use quality or not according to the APP software use quality feature word list and extracting a comment mode; finally, extracting new comment seeds based on the candidate comment mode, and further mining user comments related to the use quality;

the method comprises the following specific steps:

step1, initialization data: the comment library has c pieces of user comment after word segmentation_mM =1, there are s comment seeds seed in the comment seed library, and there are f use quality feature words feature _ word in the use quality feature word list;

wherein comment mode_m= <mode_word_m1 + … +mode_word_mn +… + mode_word_mq，mode_speech_m1 +… + mode_speech_mn +…+ mode_speech_mq ，mode_quality_in_use_m >N is more than or equal to 1 and less than or equal to q; mode _ word represents the word in the user comment, mode _ speech represents the part of speech of the word in the user comment, q represents the number of words/parts of speech in the comment mode, mode _ quality _ in _ use_mRepresentative comment_mThe reflected use quality is one of 5 use quality attributes in the use quality feature vocabulary, and the comment is not given_mJudging the use quality, wherein the value is unknown;

step3, judgment comment_mWhether it matches a certain comment seed:

if comment is commented on_mAnd comment seed_gSuccessful match, mode _ quality _ in _ use_m = seed_quality_in_use_gGive comment_mAnd (4) judging the result:<comment_m，seed_quality_in_use_g>(ii) a m + +, execute Step 2; wherein g is more than or equal to 1 and less than or equal to s;

step4, judgmentBreak comment_mWhether to match a certain usage quality feature word:

if comment is commented on_mAnd use the quality feature word_{f '}Successful match, mode _ quality _ in _ use_m= quality_in_use_{f '}Give comment_mAnd (4) judging the result:<comment_m，quality_in_use_{f '}>(ii) a Comment mode_m = < mode_word_m1 + … +mode_word_mn +…+mode_word_mq，mode_speech_m1 +… + mode_speech_mn +…+ mode_speech_mq，mode_quality_in_use_{f '}>N is more than or equal to 1 and less than or equal to q enters a candidate comment mode library, m + +, and Step2 is executed;

if comment is commented on_mMode _ quality _ in _ use, failure to match with the usage quality feature vocabulary_m= null, give comment_mAnd (4) judging the result:<comment_m，null>(ii) a m + +, execute Step 2;

2. The user comment mining method for APP software usage quality of claim 1, characterized in that: the specific steps of Step3 are as follows:

step3.1, initialization data: i = 1;

wherein seed_i = <seed_word_i1 + … + seed_word_ij + … + seed_word_ip，seed_speech_i1 +… + seed_speech_ij +… + seed_speech_ip，seed_dis_i，seed_quality_in_use_i>J is more than or equal to 1 and less than or equal to p, seed _ word represents a word in the comment seed, seed _ speech represents a part of speech of the word in the comment seed, p represents the number of words/parts of speech in the comment seed, and seed _ dis_iThe distance of the comment seed is the sum of the maximum distance between the evaluation object and the evaluation viewpoint and the number of words/parts of speech in the comment seed, and the value of seed _ quality _ in _ use_iThe use quality reflected by the comment seed is represented, and the use quality is one of 5 use quality attributes in the use quality feature vocabulary;

step3.3, comment_mWord "mode word_m1 + … + mode_word_mn + … + mode_word_mq"and comment seed_iWord "seed word_i1 + … + seed_word_ij + …+seed_word_ip"match, calculate word match value commented word _ sim_mi：

Step3.3.1, initialization j =1, and set pos _ word is null;

step3.3.3, initialize n =1, execute step3.3.4;

dis_word_mi = MAX{pos_word} -MIN{pos_word} （1）

then executing Step3.3.7;

commentWord_sim_mi=NUM(pos_word)/p （2）

otherwise, commenting word _ sim_mi=0；

Performing Step3.4;

step3.4, comment_mPart of speech "mode _ speech_m1 + … + mode_speech_mn + … + mode_speech_mq"and comment seed_iPart of speech "seed _ speech_i1 + … + seed_speech_ij + … + seed_speech_ip"match, calculate the part-of-speech match value commendspeech _ sim_mi：

Step3.4.1, initialization j =1, and set pos _ speed is empty;

step3.4.3, initialize n =1, execute step3.4.4;

dis_speech_mi = MAX{pos_speech} -MIN{pos_speech} （3）

Then executing Step3.4.8;

commentSpeech_sim_mi=NUM(pos_speech)/p （4）

Otherwise, commenting Speech _ sim_mi=0；

Performing Step3.5;

commentBoth_sim_mi=0.5*commentWord_sim_mi+0.5*commentSpeech_sim_mi （5）

Performing Step3.6;

step3.6, i + +, execute Step3.2;

step3.7, obtaining comment using equation (6)_mWith comment seed of all comment seeds_gMaximum integrated match value of (2) commentBoth _ sim_mg：

commentBoth_sim_mg

=MAX{commentBoth_sim_m1,…,commentBoth_sim_mg,…,commentBoth_sim_ms}

(1≤g≤s) ；（6）

Executing Step3.8;

if comment is commented on_mAnd comment seed_gSuccessful match, mode _ quality _ in _ use_m = seed_quality_in_use_gGive comment_mAnd (4) judging the result:<comment_m，seed_quality_in_use_g>(ii) a m + +, execute Step 2;

3. The user comment mining method for APP software usage quality of claim 1, characterized in that: the specific steps of Step4 are as follows:

step4.1, initialization variables: f' = 1;

step4.3, judgment comment_mWhether or not the use quality feature word feature _ word is included in_{f '}: if so, executing Step4.5; otherwise, executing Step4.4;

wherein, the f' th use quality feature word_{f '}The reflected use quality is quality _ in _ use_{f '}，1≤ f '≤ f；

Step4.4, comment_mAnd use the quality feature word_{f '}Matching fails, f' ++, Step4.2 is executed;

step4.5, comment_mAnd use the quality feature word_{f '}Successful match, mode _ quality _ in _ use_m = quality_in_use_{f '}Give comment_mAnd (4) judging the result:<comment_m，quality_in_use_{f '}>(ii) a Comment mode_m = < mode_word_m1 + … +mode_word_mn +… + mode_word_mq，mode_speech_m1+… + mode_speech_mn +…+ mode_speech_mq， mode_quality_in_use_{f '}>Entering a candidate comment mode library, m + +, and executing Step 2;

step4.6, comment_mMatching with the use quality feature vocabulary fails: mode _ quality _ in _ use_m= null, give comment_mAnd (4) judging the result:<comment_m，null>(ii) a m + +, execute Step 2; wherein null represents comment_mDoes not reflect the quality of use.

4. The user comment mining method for APP software usage quality of claim 1, characterized in that: the specific steps of Step5 are as follows:

step5.1, initialization variables: QinU is the quality of use attribute, QinU = 1; wherein QinU =1 represents "satisfaction"; QinU =2 for "effectiveness"; QinU =3 for "efficiency"; QinU =4 represents "risk avoidance"; QinU =5 for "context coverage";

step5.2, judging whether QinU is less than or equal to 5: if so, executing Step5.3; otherwise, obtaining the comment modes with the maximum text similarity of 5 words respectively belonging to 5 different use qualities: { mode_max1，…，mode_max5Execution Step 6;

step5.3, judging whether a comment mode with a quality attribute of QinU exists in the candidate comment mode library: if yes, extracting all comment modes with the quality attribute of QinU in the candidate comment mode library, and executing Step5.4; otherwise, mode_maxQinU= null, perform step5.10;

wherein, mode_maxQinU= null to indicate that the comment mode representing a quality of use attribute QinU is null;

step5.4, judging whether the text similarity of all comment modes in QinU is calculated according to the quality attribute: if so, executing Step5.9; otherwise, extracting a comment mode using an uncalculated text similarity in QinU as a quality attribute_mExecuting Step5.5;

step5.5, judge mode_mAnd using the quality attribute in QinUWhether the similarity of the texts in other comment modes is calculated is finished: if so, executing Step5.8; otherwise, extracting a using quality attribute of a non-neutralized mode in QinU_mComment mode for text similarity calculation_m'M' | = m, execute step5.6;

modeWord_sim_mm＇=1 - dis_word_mm＇ /max(len_word_m，len_word_m＇) （7）

modeSpeech_sim_mm＇ =1 - dis_speech_mm＇ /max(len_speech_m，len_speech_m＇) （8）

then, Step5.7 is executed;

modeBoth_sim_mm＇= 0.5*modeWord_sim_mm＇+0.5*modeSpeech_sim_mm＇（9）

Then, Step5.5 is executed;

sum_modeWord_sim_m= SUM{modeWord_sim_mm＇}(m'! = m) （10）

In the formula (10), SUM represents the SUM of elements in the calculation set;

performing Step5.4;

sum_modeWord_sim_maxQinU= MAX{sum_modeWord_sim_m}(1≤ max≤ m) （11）

Performing Step5.10;

step5.10, QinU + +, Step5.2 was performed.

5. The user comment mining method for APP software usage quality of claim 1, characterized in that: the specific steps of Step6 are as follows:

Step6.1.1, initialize k =1, execute step6.1.2;

step6.1.3, judgment mode_newkWhether it is empty: if so, seed_newk= null, k + +, execute step6.1.2; otherwise, extract its comment seed_newk={ seed_word_newk1 + … + seed_word_newkj + … + seed_word_newkp，seed_speech_newk1 +… + seed_speech_newkj +… + speech_newkp，seed_dis_newk，seed_quality_in_use_newk}, k + +, perform Step6.1.2;

step6.4, calculate all comment mode_newk＇Distance dis of keywords_newk＇：

The keywords are words with parts of speech n, v, d and a in the comment mode;

dis_newk＇=MAX{ pos_mode_speech}-MIN{pos_mode_speech} （12）

Performing Step6.4.1;

step6.5, using equation (13)) Acquisition and mode_newkAll comment mode with integrated text similarity greater than 0.35_newk＇Dis of_newk＇Maximum value new _ seed _ dis in (1)_max：

new_seed_dis_max = MAX{dis_newk＇} （13）

Performing Step6.6;