CN104239383A - MicroBlog emotion visualization method - Google Patents

Abstract

The invention discloses a microBlog emotion visualization method. By the microBlog emotion visualization method, a microBlog hot issue national attention tendency chart, a microBlog hot issue emotion national distribution diagram and a microBlog hot issue area distribution dagram are made according to relative strategies based on keyword frequency data acquired statistically and eight-dimensional emotion results obtained by emotion computing.

Description

A kind of microblog emotional method for visualizing

Technical field

The present invention relates to microblog emotional analytical approach field, specifically a kind of microblog emotional method for visualizing.

Background technology

Affection computation becomes one of current popular research field, and text emotion calculates particularly burning hot.Along with the rise of this short-text message pattern of microblogging, a large amount of texts being rich in affective characteristics can obtain easily, for text emotion research is provided convenience.Due to the difficult point in text emotion tolerance, such that text emotion is visual faces many difficult problems, microblog emotional is visual so same.

Summary of the invention

The object of this invention is to provide a kind of microblog emotional method for visualizing, to realize the displaying microblog text affective of visual pattern.

In order to achieve the above object, the technical solution adopted in the present invention is:

A kind of microblog emotional method for visualizing, is characterized in that: comprise the following steps:

(1) set of appointment topic keyword, is expanded:

Due to the colloquial style in content of microblog, in the appointment topic microblog data got, specify the original keyword seed of topic to be not the statement that standardizes, now need keyword seed spoken language words, the slang of will topic be specified original, expansion step is as follows:

(1.1), topic microblogging text participle will be specified, statistics word frequency, and determine to specify the original keyword seed of topic;

(1.2), by word frequency sequence, get front 20 words and alternatively specify topic keyword seed;

(1.3), calculate 20 candidates according to formula (1) and specify topic keyword seed and the similarity of specifying the original keyword seed of topic:

$d=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\log \frac{p(\mathrm{word}\_{\mathrm{seed}}_j,{\mathrm{word}}_i)}{p\left(\mathrm{word}\_{\mathrm{seed}}_j\right)p\left({\mathrm{word}}_i\right)}---\left(1\right)$

Wherein, word_seed _jrepresent and specify the original keyword seed of topic, word _irepresent that candidate specifies topic keyword seed, p (word_seed _j, word _i) represent the probability of specifying the original keyword seed of topic and candidate to specify topic keyword seed simultaneously to occur in microblogging text, p (word_seed _j) represent the probability of specifying the original keyword seed of topic to occur in microblogging text, p (word _i) represent that the probability that candidate specifies topic keyword seed to occur in microblogging text, d represent that candidate specifies topic keyword seed and the similarity of specifying the original keyword seed of topic;

(1.4), according to the result of calculation of step (1.3), the candidate getting first 10 of similarity rank specifies topic keyword seed as the keyword seed expanded, the keyword seed expanded as topic keyword set, is designated as K together with the original keyword seed of appointment topic;

(2) appointment topic microblog data, is separated: appointment topic microblog data is split as regional microblog data according to city belonging to microblogging, is designated as D _city; According to microblogging issuing time, in units of sky, appointment topic microblog data is split as time microblog data, is designated as D _time;

(3) the regional microblog data, by step (2) obtained is split as area time-division microblog data according to the time in units of sky, is designated as D _city ^time;

(4) the time microblog data D after, statistics specifies topic microblog data to be separated _timein the frequency of keyword seed, the frequency and being daily calculating all keyword seed to specify in topic microblog data the attention rate on this topic same day, according to statistics, adopt broken line graph, different topic selects different colors to distinguish, with keyword frequency for the longitudinal axis, take time as transverse axis, national attention rate trend map in the appointment topic fixed time section in units of sky can be obtained; Statistics area time-division microblog data D _city ^timein the frequency of keyword seed, according to the method described above, with keyword frequency for the longitudinal axis, with time and city for transverse axis, can obtain specifying topic area attention rate trend comparison diagram, in the attention rate trend comparison diagram of actualite area, adopt tufted histogram graph representation comparative information;

(5), do to specify topic whole nation emotion distribution plan and Area distribution figure, process is as follows:

(5.1) the time microblog data D specifying topic microblog data, is calculated _timeand area time-division microblog data D _city ^time; Obtain the 8 dimension microblog emotional results of specifying topic every day, as shown in formula (2):

E＝(e _hate,e _anger,e _sorrow,e _anxiety,e _surprise,e _love,e _joy,e _expect) (2)

Wherein, the vector element in formula (2) represent successively specify topic microblogging hatred, anger, sadness, anxiety, surprised, like, emotion intensity level under happiness, expectation 8 kinds of emotions;

(5.2), the three-dimensional emotion intensity level piling up histogram graph representation appointment topic microblogging every day is adopted, use respectively RGB look #EE9572, #9AC0CD, #CD8162, #5CACEE, #5D478B, #6E8B3D, #8B2500, #3A5FCD represent hatred, anger, sadness, anxiety, surprised, like, happiness, expectation 8 kinds of emotions, with emotion intensity for transverse axis, with timeline and area for the longitudinal axis, make and specify topic microblogging area emotion distribution plan, and with emotion intensity for transverse axis, be the longitudinal axis with timeline, make and specify topic microblogging whole nation emotion distribution plan.

The present invention is based on the keyword word frequency data of statistics acquisition and 8 dimension emotion results of affection computation acquisition, make the microblog hot event whole nation according to corresponding strategies and pay close attention to trend map, microblog hot event emotion whole nation distribution plan and microblog hot event Area distribution figure, can the displaying microblog text affective of visual pattern.

Accompanying drawing explanation

Fig. 1 specifies national attention rate trend map in topic fixed time section in the present invention.

Fig. 2 specifies topic area attention rate trend comparison diagram in the present invention.

Fig. 3 specifies topic microblogging area emotion distribution plan in the present invention.

Fig. 4 specifies topic microblogging whole nation emotion distribution plan in the present invention.

Embodiment

A kind of microblog emotional method for visualizing, comprises the following steps:

(1) set of appointment topic keyword, is expanded:

Due to the colloquial style in content of microblog, in the appointment topic microblog data got, specify the original keyword seed of topic to be not the statement that standardizes, now need keyword seed spoken language words, the slang of will topic be specified original, expansion step is as follows:

(1.1), topic microblogging text participle will be specified, statistics word frequency, and determine to specify the original keyword seed of topic;

(1.2), by word frequency sequence, get front 20 words and alternatively specify topic keyword seed;

(1.3), calculate 20 candidates according to formula (1) and specify topic keyword seed and the similarity of specifying the original keyword seed of topic:

$d=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\log \frac{p(\mathrm{word}\_{\mathrm{seed}}_j,{\mathrm{word}}_i)}{p\left(\mathrm{word}\_{\mathrm{seed}}_j\right)p\left({\mathrm{word}}_i\right)}---\left(1\right)$

Wherein, word_seed _jrepresent and specify the original keyword seed of topic, word _irepresent that candidate specifies topic keyword seed, p (word_seed _j, word _i) represent the probability of specifying the original keyword seed of topic and candidate to specify topic keyword seed simultaneously to occur in microblogging text, p (word_seed _j) represent the probability of specifying the original keyword seed of topic to occur in microblogging text, p (word _i) represent that the probability that candidate specifies topic keyword seed to occur in microblogging text, d represent that candidate specifies topic keyword seed and the similarity of specifying the original keyword seed of topic;

(1.4), according to the result of calculation of step (1.3), the candidate getting first 10 of similarity rank specifies topic keyword seed as the keyword seed expanded, the keyword seed expanded as topic keyword set, is designated as K together with the original keyword seed of appointment topic;

(2) appointment topic microblog data, is separated: appointment topic microblog data is split as regional microblog data according to city belonging to microblogging, is designated as D _city; According to microblogging issuing time, in units of sky, appointment topic microblog data is split as time microblog data, is designated as D _time;

(3) the regional microblog data, by step (2) obtained is split as area time-division microblog data according to the time in units of sky, is designated as D _city ^time;

(4) the time microblog data D after, statistics specifies topic microblog data to be separated _timein the frequency of keyword seed, the frequency and being daily calculating all keyword seed to specify in topic microblog data the attention rate on this topic same day, according to statistics, adopt broken line graph, different topic selects different colors to distinguish, and with keyword frequency for the longitudinal axis, take time as transverse axis, national attention rate trend map in the appointment topic fixed time section in units of sky can be obtained, as shown in Figure 1; Statistics area time-division microblog data D _city ^timein the frequency of keyword seed, according to the method described above, with keyword frequency for the longitudinal axis, with time and city for transverse axis, can obtain specifying topic area attention rate trend comparison diagram, as shown in Figure 2, in the attention rate trend comparison diagram of actualite area, adopt tufted histogram graph representation comparative information;

(5), do to specify topic whole nation emotion distribution plan and Area distribution figure, process is as follows:

(5.1) the time microblog data D specifying topic microblog data, is calculated _timeand area time-division microblog data D _city ^time; Obtain the 8 dimension microblog emotional results of specifying topic every day, as shown in formula (2):

E＝(e _hate,e _anger,e _sorrow,e _anxiety,e _surprise,e _love,e _joy,e _expect) (2)

Wherein, the vector element in formula (2) represent successively specify topic microblogging hatred, anger, sadness, anxiety, surprised, like, emotion intensity level under happiness, expectation 8 kinds of emotions;

(5.2), adopt the three-dimensional emotion intensity level piling up histogram graph representation appointment topic microblogging every day, use RGB look #EE9572 respectively, #9AC0CD, #CD8162, #5CACEE, #5D478B, #6E8B3D, #8B2500, #3A5FCD represents hatred, angry, sad, anxiety, surprised, like, glad, expect 8 kinds of emotions, with emotion intensity for transverse axis, with timeline and area for the longitudinal axis, make and specify topic microblogging area emotion distribution plan, as shown in Figure 3, and with emotion intensity for transverse axis, take timeline as the longitudinal axis, make and specify topic microblogging whole nation emotion distribution plan, as shown in Figure 4.

Claims (1)

1. a microblog emotional method for visualizing, is characterized in that: comprise the following steps:

(1) set of appointment topic keyword, is expanded:

Due to the colloquial style in content of microblog, in the appointment topic microblog data got, specify the original keyword seed of topic to be not the statement that standardizes, now need keyword seed spoken language words, the slang of will topic be specified original, expansion step is as follows:

(1.1), topic microblogging text participle will be specified, statistics word frequency, and determine to specify the original keyword seed of topic;

(1.2), by word frequency sequence, get front 20 words and alternatively specify topic keyword seed;

(1.3), calculate 20 candidates according to formula (1) and specify topic keyword seed and the similarity of specifying the original keyword seed of topic:

$d=\underset{j=1}{\overset{n}{\mathrm{\Σ}}}\log \frac{p(\mathrm{word}\_{\mathrm{seed}}_j,{\mathrm{word}}_i)}{p\left(\mathrm{word}\_{\mathrm{seed}}_j\right)p\left({\mathrm{word}}_i\right)}---\left(1\right)$

Wherein, word_seed _jrepresent and specify the original keyword seed of topic, word _irepresent that candidate specifies topic keyword seed, p (word_seed _j, word _i) represent the probability of specifying the original keyword seed of topic and candidate to specify topic keyword seed simultaneously to occur in microblogging text, p (word_seed _j) represent the probability of specifying the original keyword seed of topic to occur in microblogging text, p (word _i) represent that the probability that candidate specifies topic keyword seed to occur in microblogging text, d represent that candidate specifies topic keyword seed and the similarity of specifying the original keyword seed of topic;

(1.4), according to the result of calculation of step (1.3), the candidate getting first 10 of similarity rank specifies topic keyword seed as the keyword seed expanded, the keyword seed expanded as topic keyword set, is designated as K together with the original keyword seed of appointment topic;

(2) appointment topic microblog data, is separated: appointment topic microblog data is split as regional microblog data according to city belonging to microblogging, is designated as D _city; According to microblogging issuing time, in units of sky, appointment topic microblog data is split as time microblog data, is designated as D _time;

(3) the regional microblog data, by step (2) obtained is split as area time-division microblog data according to the time in units of sky, is designated as D _city ^time;

(4) the time microblog data D after, statistics specifies topic microblog data to be separated _timein the frequency of keyword seed, the frequency and being daily calculating all keyword seed to specify in topic microblog data the attention rate on this topic same day, according to statistics, adopt broken line graph, different topic selects different colors to distinguish, with keyword frequency for the longitudinal axis, take time as transverse axis, national attention rate trend map in the appointment topic fixed time section in units of sky can be obtained; Statistics area time-division microblog data D _city ^timein the frequency of keyword seed, according to the method described above, with keyword frequency for the longitudinal axis, with time and city for transverse axis, can obtain specifying topic area attention rate trend comparison diagram, in the attention rate trend comparison diagram of actualite area, adopt tufted histogram graph representation comparative information;

(5), do to specify topic whole nation emotion distribution plan and Area distribution figure, process is as follows:

(5.1) the time microblog data D specifying topic microblog data, is calculated _timeand area time-division microblog data D _city ^time; Obtain the 8 dimension microblog emotional results of specifying topic every day, as shown in formula (2):

E＝(e _hate,e _anger,e _sorrow,e _anxiety,e _surprise,e _love,e _joy,e _expect) (2)

Wherein, the vector element in formula (2) represent successively specify topic microblogging hatred, anger, sadness, anxiety, surprised, like, emotion intensity level under happiness, expectation 8 kinds of emotions;

(5.2), the three-dimensional emotion intensity level piling up histogram graph representation appointment topic microblogging every day is adopted, use respectively RGB look #EE9572, #9AC0CD, #CD8162, #5CACEE, #5D478B, #6E8B3D, #8B2500, #3A5FCD represent hatred, anger, sadness, anxiety, surprised, like, happiness, expectation 8 kinds of emotions, with emotion intensity for transverse axis, with timeline and area for the longitudinal axis, make and specify topic microblogging area emotion distribution plan, and with emotion intensity for transverse axis, be the longitudinal axis with timeline, make and specify topic microblogging whole nation emotion distribution plan.