CN110309320B - NBA basketball news automatic generation method combining NBA event knowledge map - Google Patents
NBA basketball news automatic generation method combining NBA event knowledge map Download PDFInfo
- Publication number
- CN110309320B CN110309320B CN201910574961.6A CN201910574961A CN110309320B CN 110309320 B CN110309320 B CN 110309320B CN 201910574961 A CN201910574961 A CN 201910574961A CN 110309320 B CN110309320 B CN 110309320B
- Authority
- CN
- China
- Prior art keywords
- time
- key
- dif
- event
- nba
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
Abstract
The invention discloses an NBA basketball news automatic generation method combining an NBA event knowledge map, which comprises the following steps: pre-processing NBA character live text data crawled by a network, removing a crawler webpage label, removing stop words in a character text, and expressing the stop words by quintuple; according to the provided segmentation algorithm, carrying out data segmentation on the preprocessed live text data to obtain a competition development trend; extracting special events according to the definition of the provided special events of the basketball game; defining a basketball news description template; combining the data segmentation result, the special event extraction result and the corresponding news description template to generate a news primary draft; combining the knowledge graph to generate match background information to obtain a news final draft; therefore, the automatic generation of the NBA event news is realized, the quality of the generated NBA event news is improved, and the generated news content can be better controlled.
Description
Technical Field
The invention relates to the field of automatic news generation methods, in particular to an NBA basketball news automatic generation method combined with an NBA event knowledge map.
Background
In the data age, news reporting in combination with news data is increasingly receiving attention of media people, and the traditional news reporting mode is gradually transformed to 'data news'. Among the numerous categories of news, sports news is one of the categories with a large audience, and the combination of event data mining analysis and data visualization technology to drive the reporting of event data news has become a common sports event news reporting means for large media in the world. Basketball is a popular item in sports, and research and analysis of event data by combining data visualization is the direction of media emphasis and prominence.
At present, a plurality of sports portals and data analysis websites aim at data analysis after basketball match, but the expression forms of the sports portals and the data analysis websites are simple tabulated visual presentation of statistical data and word summary after basketball match, and deep mining and knowledge expansion of match data are lacked in the analysis process. In particular, the practice of basketball background information expansion is lacked, and the use experience of users for reading basketball games and knowing basketball knowledge is influenced.
Disclosure of Invention
The invention provides an NBA basketball news automatic generation method combined with an NBA event knowledge map, aiming at solving the defects of the existing basketball event visualization field.
An NBA basketball news automatic generation method combining an NBA event knowledge map comprises the following steps:
1) the method comprises the steps of crawling NBA character live broadcast text data through a network, removing crawler webpage labels, removing stop words in character texts, representing the obtained text data in quintuple mode to obtain a plurality of text data (quintuple sets for short) represented by the quintuple, wherein each match corresponds to a plurality of text data represented by the quintuple;
2) segmenting a plurality of text data expressed by quintuple obtained in the step 1) to obtain a competition development trend;
3) summarizing the basketball event special events from the event events in the text data expressed by the quintuple obtained in the step 1), and extracting the special events by combining the segments obtained in the step 2) to obtain the quintuple of the special events;
4) summarizing the competition trend of each section obtained in the step 2) and filling a news description template with the quintuple of the special event obtained in the step 3) to generate a news manuscript for the basketball game;
5) and inquiring and acquiring match background information according to the NBA match knowledge map, and automatically generating NBA basketball news by combining with the basketball match news initial draft.
In the step 1), the quintuple comprises a quater, time, team, event and score, wherein the quater refers to the current number of nodes, the time is the rest time of the node, the team is the team with the event, the score is the score after the event occurs, and the event is the event;
the types of events that occur are fixed and generally include seven cases, pause, miss, cricket, score, iron strike, formation adjustment, foul, and are independent, with no necessary connection between events.
In step 2), the segmentation adopts a segmentation algorithm to obtain a competition development trend, and specifically comprises the following steps:
A) calculating the score after the occurrence of the event in the text data expressed by the quintuple by adopting the formula (1);
dif(t)＝Scorea,t-Scoreb,t (1)
dif (t) represents the difference of t time, when the difference reaches the maximum and minimum, the corresponding t time is determined as the key time of the competition, Scorea,tRepresents the total Score, of the home team at time tb,tRepresenting the total score at time t of the team. Then, the match is segmented according to the positive and negative of the difference of the key time and the appearance sequence relation.
B) Obtaining two key time points key _ time by using the formula (2) and the formula (3)1And key _ time2；
key_time1＝argmax(dif(t)) (2)
key_time2＝argmin(dif(t)) (3)
argmax (dif (t)) represents a time t at which the difference dif (t) is maximized, as key _ time1；
argmin (dif (t)) represents a time t at which the difference dif (t) is minimized, and is taken as key _ time2；
For key time point key _ time1、key_time2Correcting according to the formula (4);
in the formula (4), n is the number of each remaining time in the text data represented by the quintuple corresponding to each game, and when i is 1, the key _ timeiIs key _ time1When i is 2, key _ timeiIs key _ time2，；
If the key time point key _ timeiThe time of one sixth of the first competition is shown, the starting time t of the competition is set0As a key point in time, a start time t0As key _ timei；
If the key time point key _ timeiThe time which appears one sixth after the game is the end time t of the gamenAs a key point in time, an end time tnAs key _ timei；
If the critical timePoint key _ timeiThe time that does not appear in the first sixth of the game and does not appear in the last sixth of the game, the key time point key _ timeiNo modification is made;
dif(key_time1) To the corrected key time point key _ time1A difference of (d); dif (key _ time)2) To the corrected key time point key _ time2A difference of (d);
C) hereinafter, it will be classified as dif (key _ time)1)-dif(key_time2)>8 and dif (key _ time)1)-dif(key_time2) Judging the overall trend condition of the competition according to the difference condition of the two key time points less than or equal to 8;
when the difference of scores>8 (i.e. dif (key _ time)1)-dif(key_time2)>8) And in time, the difference fluctuation range of the competition is large, and the difference between the front and the back of the key time point can be regarded as large. At this moment, the match is suitable to be cut into three segments according to time points, and the situation on the match field is summarized into the following 6 conditions:
a)key_time1>key_time2,dif(key_time1)>0,dif(key_time2)<0, firstly, expanding superiority, then, reversely surpassing and then reducing inferiority;
b)key_time1<key_time2,dif(key_time1)>0,dif(key_time2)<0, firstly expanding the inferior situation, then reversely exceeding and then reducing the superiority;
c)key_time1>key_time2,dif(key_time1)<0, firstly reducing the disadvantages, then expanding the disadvantages and then reducing the disadvantages;
d)key_time1<key_time2,dif(key_time1)<0, expanding the disadvantages after expanding the disadvantages;
e)key_time1>key_time2,dif(key_time2)>0, firstly, expanding the advantages, then reducing the advantages and then expanding the advantages;
f)key_time1>key_time2,dif(key_time2)>0 first "minification advantage"And then the advantages are expanded.
When the score difference is less than or equal to 8 (i.e. dif (key _ time)1)-dif(key_time2) Not more than 8), then this section match discrepancy fluctuation range is less, and the competition situation is inclined to the dragsaw state, can handle this section match as whole section this moment, specifically divide into following three kinds of condition:
g)dif(key_time2)>10, the minimum difference is more than 10, the variation amplitude of the difference is not large, and the situation of the competition field is summarized as the stable leading of the leading team;
h)dif(key_time1)<10, the minimum difference is less than 10 below zero, the variation amplitude of the difference is not large, and the situation of the competition field is summarized as the stable state of the leading team;
i)dif(key_time 1)>0，dif(key_time 2)<the 0: the difference changes around 0 min, and the situation of the competition field is summarized as 'gummy' in two teams.
In the step 3), quintuple of the special event comprises player, team, action, result and time, wherein the player refers to a player, and the team refers to the team; action is a specific action, mostly a basic event; result is the score after the action has taken place; time is the time at which the action occurred.
The extracted special event may be null.
Compared with the prior art, the invention has the following advantages:
the NBA basketball news automatic generation method combined with the NBA event knowledge map can extract special events and game trends according to NBA character live broadcast data, meanwhile, the NBA event knowledge map is used for obtaining background information of a game, and then the NBA event news is generated by combining with a defined news template. The method not only can efficiently generate news, but also can ensure that the generated results are clear in logic and rich in content, can effectively reduce the workload of NBA event news journalists, and can greatly improve the reading experience of NBA fans.
The invention realizes the automatic generation of the NBA event news, improves the quality of the generated NBA event news and can better control the generated news content.
Drawings
FIG. 1 is a flow chart of the NBA basketball news automatic generation method combining the NBA event knowledge map of the present invention.
Detailed Description
In order to better explain the relevant contents of the present invention, the technical solutions involved in the present invention will be fully and clearly described below with reference to the attached flowcharts. It is to be understood that the described embodiments are merely exemplary of some, not all, and that the invention should not be limited to the disclosed embodiments, but rather should be understood to provide those skilled in the art with the benefit of the description herein.
As shown in fig. 1, a NBA basketball news automatic generation method combined with NBA event knowledge map includes the following steps:
1) the method comprises the steps of crawling NBA character live broadcast text data through a network, removing crawler webpage labels, removing stop words in character texts, representing the obtained text data in quintuple mode to obtain a plurality of text data (quintuple sets for short) represented by the quintuple, wherein each match corresponds to a plurality of text data represented by the quintuple;
2) segmenting a plurality of text data expressed by quintuple obtained in the step 1) to obtain a competition development trend;
3) summarizing the basketball event special events from the event events in the text data expressed by the quintuple obtained in the step 1), and extracting the special events by combining the segments obtained in the step 2) to obtain the quintuple of the special events;
4) summarizing the competition trend of each section obtained in the step 2) and filling a news description template with the quintuple of the special event obtained in the step 3) to generate a news manuscript for the basketball game;
5) and inquiring and acquiring match background information according to the NBA match knowledge map, and automatically generating NBA basketball news by combining with the basketball match news initial draft.
In the step 1), the quintuple comprises a quater, time, team, event and score, wherein the quater refers to the current section number, the time is the rest time of the section, the team with the event occurs, the score is the score after the event occurs, and the event is the event occurs;
the types of events that occur are fixed and generally include seven cases, pause, miss, cricket, score, iron strike, formation adjustment, foul, and are independent, with no necessary connection between events.
In step 2), a segmentation algorithm is adopted in a segmentation manner to obtain a competition development trend, and the method specifically comprises the following steps:
A) calculating the score after the occurrence of the event in the text data expressed by the quintuple by adopting the formula (1);
dif(t)＝Scorea,t-Scoreb,t (1)
dif (t) represents the difference of t time, when the difference reaches the maximum and minimum, the corresponding t time is determined as the key time of the competition, Scorea,tRepresents the total Score, of the home team at time tb,tRepresenting the total score at time t of the team. Then, the match is segmented according to the positive and negative of the difference of the key time and the appearance sequence relation.
B) Obtaining two key time points key _ time by using the formula (2) and the formula (3)1And key _ time2；
key_time1＝argmax(dif(t)) (2)
key_time2＝argmin(dif(t)) (3)
argmax (dif (t)) represents a time t at which the difference dif (t) is maximized, as key _ time1；
argmin (dif (t)) represents a time t at which the difference dif (t) is minimized, and is taken as key _ time2；
For key time point key _ time1、key_time2Correcting according to the formula (4);
in the formula (4), n is the rest of each of the text data expressed by quintuple corresponding to each gameWhen i is 1, key _ timeiIs key _ time1When i is 2, key _ timeiIs key _ time2，；
If the key time point key _ timeiThe time of one sixth of the first competition is shown, the starting time t of the competition is set0As a key point in time, a start time t0As key _ timei；
If the key time point key _ timeiThe time which appears one sixth after the game is the end time t of the gamenAs a key point in time, an end time tnAs key _ timei；
If the key time point key _ timeiThe time that does not appear in the first sixth of the game and does not appear in the last sixth of the game, the key time point key _ timeiNo modification is made;
dif(key_time1) To the corrected key time point key _ time1A difference of (d); dif (key _ time)2) To the corrected key time point key _ time2A difference of (d);
C) hereinafter, it will be classified as dif (key _ time)1)-dif(key_time2)>8 and dif (key _ time)1)-dif(key_time2) Judging the overall trend condition of the competition according to the difference condition of the two key time points less than or equal to 8;
when the difference of scores>8 (i.e. dif (key _ time)1)-dif(key_time2)>8) And in time, the difference fluctuation range of the competition is large, and the difference between the front and the back of the key time point can be regarded as large. At this moment, the match is suitable to be cut into three segments according to time points, and the situation on the match field is summarized into the following 6 conditions:
a)key_time1>key_time2,dif(key_time1)>0,dif(key_time2)<0, firstly, expanding superiority, then, reversely surpassing and then reducing inferiority;
b)key_time1<key_time2,dif(key_time1)>0,dif(key_time2)<0 first "enlarge the inferior position" and then "reverse super"Reduced dominance ";
c)key_time1>key_time2,dif(key_time1)<0, firstly reducing the disadvantages, then expanding the disadvantages and then reducing the disadvantages;
d)key_time1<key_time2,dif(key_time1)<0, expanding the disadvantages after expanding the disadvantages;
e)key_time1>key_time2,dif(key_time2)>0, firstly, expanding the advantages, then reducing the advantages and then expanding the advantages;
f)key_time1>key_time2,dif(key_time2)>0, firstly, narrowing the advantage, then narrowing the advantage and then expanding the advantage.
When the score difference is less than or equal to 8 (i.e. dif (key _ time)1)-dif(key_time2) Not more than 8), then this section match discrepancy fluctuation range is less, and the competition situation is inclined to the dragsaw state, can handle this section match as whole section this moment, specifically divide into following three kinds of condition:
g)dif(key_time2)>10, the minimum difference is more than 10, the variation amplitude of the difference is not large, and the situation of the competition field is summarized as the stable leading of the leading team;
h)dif(key_time1)<10, the minimum difference is less than 10 below zero, the variation amplitude of the difference is not large, and the situation of the competition field is summarized as the stable state of the leading team;
i)dif(key_time 1)>0，dif(key_time 2)<the 0: the difference changes around 0 min, and the situation of the competition field is summarized as 'gummy' in two teams.
In the step 3), quintuple of the special event comprises player, team, action, result and time, wherein the player refers to a player, and the team refers to the team; action is a specific action, mostly a basic event; result is the score after the action has taken place; time is the time at which the action occurred.
The extracted special event may be null.
The invention provides an NBA basketball news automatic generation method combining an NBA event knowledge graph, which comprises the following steps:
step 1: data crawling and preprocessing;
each NBA match has corresponding textual live broadcast data available on a particular website, such as "https:// NBA. hupu. com/games/playbyply/156253". The live text data of each game is divided into four sections, and each section generally has about 100 text data describing the game. In the text live broadcast data, there are five key fields, which are section number, time, team, event, score, and can be expressed as quintuple data (quater, time, team, event, score), where quater refers to the current section number, time is the time left in the section, team is the team who has an event, score is the score after the event has occurred, and event is the event that has occurred. In the event, it can be subdivided into the following seven types, respectively: pause, error, cricket, score, iron, formation adjustment, foul.
In order to better cover the information of the game, the following basic information, namely the type of the event, the time of the event, the executor of the event and the score after the event, needs to be extracted from the live text. Each piece of data in the live text broadcast is classified according to event types through keyword matching, and score _ event, reb _ event, miss _ event, stop _ event, adjust _ event, fol _ event and error _ event are obtained and respectively represent a scoring event set, a backboard event set, a casting event set, a pause event set, a formation adjustment event set, a foul event set and a fault event set. And then integrating the main information in the text live broadcast into the set. Thus, a score _ event { sc1, sc 2.. and scn }, a reb _ event { r1, r 2.. and rn }, a miss _ event { m1, m 2.. and mn }, a stop _ event { st1, st 2.. and stn }, an adjust _ event { a1, a 2.. and an }, a foul _ event { f1, f 2.. and fn }, an error _ event { e1, e 2.. and en }, are obtained. The elements in the set are all quadruplets formed by the type of an event, the time of the event, the executor of the event and the score after the event occurs.
Step 2: data segmentation processing;
each piece of data in the live text records an event occurring on the court, but the individual events can not all appear in the generated news, and important events are selected to generate news content after being summarized. There are many different situations where events are important, such as a player scoring continuously during that time, a team scoring continuously, etc. Therefore, to obtain important events, the live text must be segmented and important events occurring in different time periods must be summarized, so as to generate news content.
The invention provides a segmentation method based on key time points. The method based on the key time point firstly obtains the key time point in the live text broadcast, wherein the key time point is the key time point of the section when the difference reaches the maximum and the difference reaches the minimum, the reason is that the key time point is a turning point of the section of match when the difference reaches the maximum and the difference reaches the minimum, and the key time point is a process of expanding the advantages of the leading team or reducing the disadvantages of the leading team before the difference reaches the maximum; before the difference reaches the minimum, the method is a process of reducing the advantages of the main team or expanding the disadvantages of the main team. Events important to a game often occur in these processes. The method comprises the following steps:
key_time1＝argmaxdif(t)；key_time2＝argmindif(t)；
wherein key _ time1Represents the time point of maximum dispersion, key _ time2Representing the point in time when the difference is minimal. And segmenting the match according to the sequence relation of the positive and negative differences and the occurrence of the differences at the key moment.
In case 1, if the maximum difference is greater than 0 and the minimum difference is less than 0, the maximum difference appears before the minimum difference, which generally indicates that the home team expands the difference to reach the maximum difference in the game, then is leveled up to exceed the score and laggard, and finally narrows the lag difference. The process can be summarized as expanding superiority, being inverted superordinate and narrowing inferiority.
Case 2, if the maximum differential is greater than 0 and the minimum differential is less than 0, the maximum differential appears after the minimum differential. This situation illustrates that the home team lags behind the game by the maximum score, then exceeds the score by the maximum, and finally the advantage of the lead is reduced. The process can be summarized as expanding the disadvantages, reversing the advantages and reducing the advantages.
Case 3, if the maximum difference is less than 0. The situation shows that the home team is always behind in the competition, and the process can be summarized as disadvantages reduction, disadvantages expansion and disadvantages reduction.
Case 4, if the maximum difference is less than 0. This situation indicates that the home team is behind in the competition, and the process can be summarized as expanding the disadvantages, contracting the disadvantages and expanding the disadvantages.
Case 5, if the minimum differential is greater than 0. This situation indicates that the lead team is leading in the game, and the process can be summarized as expanding the advantage, contracting the advantage and expanding the advantage.
Case 6, if the minimum differential is greater than 0. This situation indicates that the lead team is leading in the game, and the process can be summarized as a reduction advantage, an expansion advantage, and a reduction advantage.
In case 7, if the minimum differential is larger than 10, the variation of the differential is not large. This situation shows that the lead team is steady and leading, and the advantage is not further expanded, nor is there any sign of shrinkage. This process can be summarized as steady lead.
In case 8, if the maximum differential is less than-10, and the variation of the differential is not large. This situation shows that the disadvantages are not further expanded and there is no sign of shrinkage after the lead is stabilized. This process can be summarized as falling behind steady state.
Case 9, if the variance varies around 0. This situation illustrates that the home team and the guest team are very sticky and you are competing for their arrival and their arrival. The process can be summarized as gluing. This advantageously divides the data into segments.
And step 3: extracting special events;
for the event set acquired in step 1, the following specified special events are extracted:
special event 1, the highest score, represents that the player helped the team to contribute the most scores during a certain period of time.
Special event 2, important score 1, at some time the player scores help the team to be out of score.
Special event 3, important score 2, at some time the player scores to help the team lead to a maximum.
Special event 4, important score 3, in the case of petty scorching, the player scores at the moment of truth.
Special event 5, important score 4, scoring in case the team does not score for a long time.
Special event 6, the key backboard, the backboard that was preempted before the key score.
The special event 7 has the most faults, and the faults are the most in a time period with expanded disadvantages.
Special event 8, iron hit at most, and throw the most balls in the time period of extended disadvantages.
Special event 9, critical pause, pause near critical point in time.
The special event 10, attack climax, reaches the maximum lead time, team scores the number.
And 4, step 4: generating a primary draft according to the extracted special events and the obtained segmentation result;
for the selected special event, selecting a corresponding description template, and filling the extracted key information into the template, taking the case 1 in the step 3 as an example:
pelicans took slightly initiative at the beginning of the race, leading 76 people by about 2 points. Then, the Joer-Enbide tie 76 person attacks reversely, and the person cuts down 13 points, so that the difference is smoothed and 16 points are exceeded. The first race was completed, and 76 pelicans led 15 points ahead.
And 5: combining the NBA event knowledge graph to generate a news final draft;
a basketball game involving two teams may have some relationship between them that is a good complement to the creation of news items. The relation of the teams in the battle array can be inquired through the knowledge graph, and one historical hand-engaging score of the two teams is obtained. For example, if the two parties of the battle of the match are Saint Antonio Ma Ci and Jinzhou warrior, the historical victory or defeat relationship between them is obtained by direct relationship inquiry. Background information can be obtained according to the win-lose relationship: "how much the saint antonio wins the knight's win in the state. Or "saint antoni horse thorn on maw warrior 62 win 38 minus" directly. Besides the direct historical match performance relationship among the teams, the indirect relationship among the teams can be obtained, such as the match of the team with the latest match, which is also a good information supplement for the match news. In the basketball game knowledge map, the entity of the game is defined, in order to obtain the recent hand-delivering performance of two teams, the nodes of the game entity related to the two teams are obtained from the knowledge map,
the nodes store information such as recent hand-to-hand competition time, final score, single-section score and the like of two teams, and the information can be used for reasoning to obtain background description between the teams. For example, if team a and team B have both lost their nearest hands of hand, a "team B has a significant advantage over team a" may be obtained. In addition to team context information, player context information may also be obtained from the knowledge-graph, such as hand-of-hand records between players, the player's trump, the player's recent status, and the like. And after the supplementary information is obtained, combining the supplementary information with the first draft to generate a final draft.
The data of the invention are basically from the tiger NBA website, and the complete examples generated are as follows:
beijing time 2018, 11 months, 22 days brief, and NBA pelican arena challenges 76 people. This is the comparison between a western team and an eastern team, and the game is very vigorous. 76 people have the current performance of 12 win 7 minus, pelican has the current performance of 10 win 7 minus and 76 people have the dominant performance. In the last surcharge, 76 pelicans were delivered.
Pelicans took slightly initiative at the beginning of the race, leading 76 people by about 2 points. Then, the Joer-Enbide tie 76 person attacks reversely, and the person cuts down 13 points, so that the difference is smoothed and 16 points are exceeded. The first race was completed, and 76 pelicans led 15 points ahead.
Pelican lagged 76 for 15 points at the beginning of the second race, with 76 holding a tremendous advantage. Pelicans in subsequent races have attempted to close the score gap, slightly closing the score at the end of the festival. The next race was completed, and 76 pelicans were led by 10 points.
The warfare is easy, 76 people always keep leading, and the highest leading reaches 14 minutes. Pelican did not have the scores expanded further, and finally moved back some scores. Three quarters were completed and 76 leading pelicans were assigned 9 points.
At the end of the day, 76 people remain leading, with the highest leading reaching 16 points. Pelicans subsequently produced a wave of 32-17 small climax, closing the score. Finally, 76 people defeat pelicans.
Pelican data: antoni-davis 12 points 16 backboard; nigula-milofitqi 13 in 13 backboard; 30 minutes and 10 minutes of Zhu-Huoludi for assisting attack; zhu Li Yes-Lande 22 fen 10 backboard.
Data of 76 persons: Joer-Enbide 31 points 19 backboard; 13 minutes and 4 minutes of JJ-Leideck for assisting tapping; gemi-butler 13 points 6 backboard; this-simons 22 points to 8 backboard.
[ list of injuries and diseases ]
Pelican: an Elfred-Pelton finger injury was absent.
Claims (3)
1. An NBA basketball news automatic generation method combining an NBA event knowledge map is characterized by comprising the following steps:
1) the method comprises the steps of crawling NBA character live text data through a network, removing a crawler webpage label and stop words in a character text, representing the obtained text data in quintuple, obtaining a plurality of text data represented by quintuple, wherein each match corresponds to a plurality of text data represented by quintuple;
the quintuple comprises a quater, a time, a team, an event and a score, wherein the quater refers to the current number of nodes, the time is the rest time of the node, the team with the event occurs, the score is the score after the event occurs, and the event is the event occurs;
2) segmenting a plurality of text data expressed by quintuple obtained in the step 1) to obtain a competition development trend;
the segmentation adopts a segmentation algorithm to obtain the development trend of the competition, and the competition is segmented according to the positive and negative of the difference of the key moments and the occurrence sequence relation, and the segmentation method specifically comprises the following steps:
A) calculating the score after the occurrence of the event in the text data expressed by the quintuple by adopting the formula (1);
dif(t)＝Scorea,t-Scoreb,t (1)
dif (t) represents the difference of t time, when the difference reaches the maximum and minimum, the corresponding t time is determined as the key time of the competition, Scorea,tRepresents the total Score, of the home team at time tb,tThe total score of the passenger team at the time t is represented;
B) obtaining two key time points key _ time by using the formula (2) and the formula (3)1And key _ time2；
key_time1＝argmax(dif(t)) (2)
key_time2＝argmin(dif(t)) (3)
argmax (dif (t)) represents a time t at which the difference dif (t) is maximized, as key _ time1；
argmin (dif (t)) represents a time t at which the difference dif (t) is minimized, and is taken as key _ time2；
For key time point key _ time1、key_time2Correcting according to the formula (4);
in the formula (4), when i is 1, key _ timeiIs key _ time1When i is 2, key _ timeiIs key _ time2；
If the key time point key _ timeiThe time of one sixth of the first competition is shown, the starting time t of the competition is set0As a key point in time, a start time t0As key _ timei；
If the key time point key _ timeiThe time which appears one sixth after the game is the end time t of the gamenAs a key point in time, an end time tnAs key _ timei；
If the key time point key _ timeiThe time that does not appear one sixth of the game before and one sixth of the game after, then offKey time _ timeiNo modification is made;
dif(key_time1) To the corrected key time point key _ time1A difference of (d); dif (key _ time)2) To the corrected key time point key _ time2A difference of (d);
C) hereinafter, it will be classified as dif (key _ time)1)-dif(key_time2) > 8 and dif (key _ time)1)-dif(key_time2) Judging the overall trend condition of the competition according to the difference condition of the two key time points less than or equal to 8;
in step C), when dif (key _ time)1)-dif(key_time2) When the competition is more than 8, the competition is suitable to be cut into three sections according to time points, and the following 6 conditions are divided according to the situation on the competition field:
a)key_time1＞key_time2,dif(key_time1)＞0,dif(key_time2) Less than 0, firstly, expanding superiority, then, reducing inferiority;
b)key_time1＜key_time2,dif(key_time1)＞0,dif(key_time2) Less than 0, firstly expanding the inferior position, then reversely exceeding and then reducing the superiority;
c)key_time1＞key_time2,dif(key_time1) Less than 0, firstly reducing the disadvantages, then expanding the disadvantages and then reducing the disadvantages;
d)key_time1＜key_time2,dif(key_time1) Less than 0, firstly expanding the disadvantages, then reducing the disadvantages and then expanding the disadvantages;
e)key_time1＞key_time2,dif(key_time2) If more than 0, firstly expanding the advantages, then reducing the advantages and then expanding the advantages;
f)key_time1＜key_time2,dif(key_time2) More than 0, firstly reducing the advantage, then expanding the advantage and then reducing the advantage;
when dif (key _ time)1)-dif(key_time2) When the match is less than or equal to 8, the match is treated as a whole, and the method is divided into the following three conditions:
g)dif(key_time2) If the difference is more than 10, the minimum difference is more than 10, the variation amplitude of the difference is not large, and the situation of the competition field is summarized as the stable leading of the leading team;
h)dif(key_time1) -10, the minimum difference is less than-10, the variation range of the difference is not large, and the situation of the competition field is summarized as the stable state of the leading team;
i)dif(key_time1)＞0，dif(key_time2) The difference is changed around 0 min < 0, and the situation of the competition field is summarized as two teams 'gumming';
3) summarizing the basketball event special events from the event events in the text data expressed by the quintuple obtained in the step 1), and extracting the special events by combining the segments obtained in the step 2) to obtain the quintuple of the special events;
4) summarizing the competition trend of each section obtained in the step 2) and filling a news description template with the quintuple of the special event obtained in the step 3) to generate a news manuscript for the basketball game;
5) and inquiring and acquiring match background information according to the NBA match knowledge map, and automatically generating NBA basketball news by combining with the basketball match news initial draft.
2. The method of claim 1 wherein the events include pauses, mistakes, cricket, scoring, iron, formation adjustments, infractions.
3. The method for automatically generating NBA basketball news combined with NBA event knowledge-graph in claim 1, wherein in step 3), quintuple of special event comprises player, team, action, result, time, wherein player refers to player and team refers to team; action is a specific action; result is the score after the action has taken place; time is the time at which the action occurred.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910574961.6A CN110309320B (en) | 2019-06-28 | 2019-06-28 | NBA basketball news automatic generation method combining NBA event knowledge map |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910574961.6A CN110309320B (en) | 2019-06-28 | 2019-06-28 | NBA basketball news automatic generation method combining NBA event knowledge map |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110309320A CN110309320A (en) | 2019-10-08 |
| CN110309320B true CN110309320B (en) | 2021-04-06 |
Family
ID=68077833
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910574961.6A Active CN110309320B (en) | 2019-06-28 | 2019-06-28 | NBA basketball news automatic generation method combining NBA event knowledge map |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110309320B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111177396B (en) * | 2019-11-13 | 2023-06-06 | 浙江广播电视集团 | Automatic analysis and visualization method for basketball event shooting event combined with knowledge graph |
| CN110765753B (en) * | 2019-12-27 | 2023-07-14 | 广东博智林机器人有限公司 | Document generation method, system, computer device and storage medium |
| CN111324748B (en) * | 2020-02-28 | 2023-08-04 | 北京百度网讯科技有限公司 | Sports war report generation method and device, electronic equipment and storage medium |
| CN113411623A (en) * | 2021-06-15 | 2021-09-17 | 首都师范大学 | Automatic news generation method and system based on difference-time function algorithm and computer readable storage medium |
| CN113497949A (en) * | 2021-06-15 | 2021-10-12 | 首都师范大学 | Live broadcast method based on difference-time function algorithm, event live broadcast terminal, electronic equipment and computer readable storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105912526A (en) * | 2016-04-15 | 2016-08-31 | 北京大学 | Sports game live broadcasting text based sports news automatic constructing method and device |
| CN106021389A (en) * | 2016-05-12 | 2016-10-12 | 新华通讯社 | System and method for automatically generating news based on template |
| CN106407343A (en) * | 2016-09-06 | 2017-02-15 | 首都师范大学 | Automatic generation method for NBA competition news |
| CN106776523A (en) * | 2017-01-22 | 2017-05-31 | 百度在线网络技术（北京）有限公司 | News speed report generation method and device based on artificial intelligence |
| CN107644085A (en) * | 2017-09-22 | 2018-01-30 | 百度在线网络技术（北京）有限公司 | The generation method and device of competitive sports news |
-
2019
- 2019-06-28 CN CN201910574961.6A patent/CN110309320B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105912526A (en) * | 2016-04-15 | 2016-08-31 | 北京大学 | Sports game live broadcasting text based sports news automatic constructing method and device |
| CN106021389A (en) * | 2016-05-12 | 2016-10-12 | 新华通讯社 | System and method for automatically generating news based on template |
| CN106407343A (en) * | 2016-09-06 | 2017-02-15 | 首都师范大学 | Automatic generation method for NBA competition news |
| CN106776523A (en) * | 2017-01-22 | 2017-05-31 | 百度在线网络技术（北京）有限公司 | News speed report generation method and device based on artificial intelligence |
| CN107644085A (en) * | 2017-09-22 | 2018-01-30 | 百度在线网络技术（北京）有限公司 | The generation method and device of competitive sports news |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110309320A (en) | 2019-10-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110309320B (en) | NBA basketball news automatic generation method combining NBA event knowledge map | |
| US7668791B2 (en) | Distinguishing facts from opinions using a multi-stage approach | |
| CN103886034B (en) | A kind of method and apparatus of inquiry input information that establishing index and matching user | |
| US11514242B2 (en) | Method for automatically summarizing internet web page and text information | |
| US10198480B2 (en) | Method and apparatus for determining hot user generated contents | |
| CN108845982B (en) | Chinese word segmentation method based on word association characteristics | |
| CN102682120B (en) | Method and device for acquiring essential article commented on network | |
| DE112011106041T5 (en) | search device | |
| Vikram et al. | An effective pre-processing algorithm for information retrieval systems | |
| Barot et al. | Analysis and prediction for the Indian Premier League | |
| WO2017028706A1 (en) | Method and device for live text broadcasting of match | |
| JP2011070291A (en) | Device, system and method for extraction of topic word, and program | |
| George et al. | The effect of altitude and travel on rugby union performance: analysis of the 2012 super rugby competition | |
| Dyreson | World Harmony or an athletic ‘Clash of Civilizations’? The Beijing Olympic spectacle, BMX bicycles and the American contours of globalisation | |
| CN114997161A (en) | Keyword extraction method and device, electronic equipment and storage medium | |
| Agarwal et al. | A datamining approach for emotions extraction and discovering cricketers performance from stadium to sensex | |
| Burdsey | Football fans must take the lead in the fight against racism | |
| Schauberger et al. | Modeling football results in the German Bundesliga using match-specific covariates | |
| Pellegrin et al. | INAOE-UNAL at ImageCLEF 2015: Scalable Concept Image Annotation. | |
| Toprak et al. | The positive effect of PMI on the selection of meaningful words | |
| Pramudita et al. | Clustering Sports News in Indonesian Using Modified K-Medoid Method | |
| Xi et al. | ZZISTI at TREC2013 Temporal Summarization Track. | |
| CN104572621B (en) | A kind of term decision method based on decision tree | |
| Rojo-Labaien | Sport as a pillar of representation of current basque identity | |
| Georgiev et al. | The Modern English |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |