KR20090061520A - Method for providing program and person name information management in electronic program guide service - Google Patents

Abstract

A method for managing broadcast program names and person names in an EPG search service is provided to remove the necessity for a user to memorize and search a correct name included in an EPG DB. The original name is indexed through the space elimination of a program name included in an EPG DB(S200). A broadcast program name and a person name are separated by the syllable unit through an unused name list(S202). The space is removed from the program name of EPG and is indexed(S204). The program name is indexed by a bigram unit for robust recognition against a partial spelling error(S206).

Description

Program name and name management method in EP search service {METHOD FOR PROVIDING PROGRAM AND PERSON NAME INFORMATION MANAGEMENT IN ELECTRONIC PROGRAM GUIDE SERVICE}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an EPG (Electronic Program Guide) retrieval service. In particular, the name and name management of a broadcast program in an EPG retrieval service suitable for robustly supporting the retrieval of an EPG service that electronically provides information about a program broadcasted on a TV. It is about a method.

The present invention is derived from the research conducted as part of the IT growth engine technology development project of the Ministry of Information and Communication and the Ministry of Information and Telecommunications Research and Development. [Task Management Number: 2006-S-036-02, Task name: Large-capacity interactive dispersion for new growth engine industries] Development of processing voice interface technology].

There are numerous TV programs in the IPTV (Internet Protocol Television) or DVR (Digital Video Recorder) environment, and EPG is being serviced as an information source for effective search.

The EPG service makes it easy to get the program information you want by using the program name, performer name, and date information. For example, EPG data can be searched for "When do you broadcast Dokgo Youngjae's spot report scandal?", "Find a movie starring Sylvester Stallone", "Find a baseball program to broadcast tomorrow." Most EPG services use program names and performer names to obtain information about desired programs.

However, these names include numbers, foreign words, various symbols, etc., in one or more words, and some are often composed of a single sentence. Most EPG services support searching on the assumption that you know the exact program name or person name in the search using the program name. So if you don't know the name explicitly, you won't find the program any other way than searching with some substring.

For example, in the case of "Doraemon-Season 3 When?", The user uses a variety of search methods if the user does not know the exact name of the program. For example, it is possible to search for "Doraemon", "Doraemon 3", "Doraemon Season 3", "Doraemon", and the like.

Or, you could use a name like "Robert Redford", "Robert Radford", "Robot Radford", "Robot Redford", etc., to find a program like "What is Robert Redford's movie?" .

If you use a program name or person name in the EPG search process, you may not be able to easily find the program you are looking for because of "spelling errors", "use only part of the name", or "errors in English / numeric / symbols". not.

As such, the search using the broadcast program name and the name of the person is the most basic search method in the EPG service. The search program information (start time, plot, performer information, etc.) of the desired program is directly searched by the broadcast program name, or the person or producer who appeared in the program. You can search for a program by the name of a director or a director.

Most program names are in the form of one or more phrases or sentences, and often contain named entities or proper nouns, and simply do not recognize the beginning or end of the program name through morphological or parsing methods. . In addition, in the case of a foreigner's name, it is often composed of one or more words.

In addition, EPG users rarely know long program names or names of people, and even if they use a part of a program name and a part of a name for a search, or use different syllables and words differently, It is not clear whether the program and the name of the person were chosen. In this case, spelling errors, spacing errors, and errors using only a part of the name cannot search the exact name in the EPG DB. Recognition of these names is the most basic step for EPG retrieval, because without knowing the program name or person name, the next step in the EPG service cannot proceed.

Accordingly, the present invention overcomes the error that generally occurs when a user searches for a program by program name or person name in an EPG use environment, and there is a mismatch between the program name or person name of the EPG data DB and the program name or person name used in the search. To provide a solution to the problem.

It is another object of the present invention to provide a solution for effectively indexing and searching for program names and names.

According to a preferred embodiment for solving the problems of the present invention, a broadcast program name and name management method in the EPG search service, multi-step indexing the broadcast program name and person name by word unit and by-gram unit, and the user's search query Broadcast program name and name in the EPG search service comprising the step of recognizing the broadcast program and the name of the user included in the search query of the user by using the indexed broadcast program name and person name and contextual knowledge in the indexing process upon request Provide management methods.

According to the present invention, there is provided an effective and robust method for overcoming common errors that users frequently encounter in name recognition, which is the most important process in EPG search. In order to solve the problem caused by the mismatch between the program name and the name included in the EPG DB, the user always searches while remembering the exact name included in the EPG DB. You can solve the inconvenience.

Users who use the system that provides the EPG information generally want to watch the program by using the program name or performer information, the search using the genre of the program they are looking for, the channel name or the broadcast time zone. Search for. Therefore, in the EPG information retrieval, it is important to accurately recognize the program name or the name of the user desired as a condition for the program retrieval.

'Genre', 'channel name' and 'broadcasting time' are composed of relatively short words, and most of them have a fixed vocabulary, so it is easy to recognize them, while 'program name' or 'person' is spread over many words. However, it is not easy to recognize because it changes frequently and does not consist of a fixed vocabulary.

In addition, most of the problems that users have difficulty in searching for EPGs through 'program name' or 'person name' are due to a mismatch between 'program name' or 'person name' stored in the EPG DB. Occurs. The most common errors are as follows.

First, you want to search using only a fraction of the entire program name or person's name. For example, in order to find a long program of "reading gifted students", the program is often searched by changed names such as "reading students of reading", "field reporters", "scandals of reading" and "scandals". In addition, the actual program name inputs the entire program name including 'optional additional information' such as "live broadcast" or "(1, 2, 3)" in "Morning of Live World (Part 1, 2, 3)". There are few cases. In addition, people searching for programs such as "" CSI Miami: New York Season 3 "in the form of original titles entered into the EPG DB are hard to find. To solve this, the optional additional information is automatically removed and the word index is indexed. And a method of supporting the search, which further requires a method of finding the longest program name by combining partially recognized words, in the above example, a unit of "reading gifted children" and "scandal". After recognizing the program name, a method for extending the recognition to the longest matched "on-site gift scandal" is required.

Second, if a symbol is included in a program name, it is necessary to normalize it in a keyword input or a search through speech recognition. For example, to retrieve the original program "story show-accuse this person", the user may enter "story show accused this person" except the symbol "-". Also, a program called "Animal Strike Back: Puma" will fire or enter keywords except ":". In addition, in the case of "fish carp fishing Q & A with Song Gui-sub", "&" is often inputted as "An", and a method of normalizing it should be applied. To solve this problem, a method such as symbol normalization is required in the index and search query processing steps.

Third, this error occurs because all words are indexed and a list of stopwords cannot be recognized by name. For the names recognized through the above two methods, the misrecognized names should be restored using surrounding contextual information. After recognizing the name of the program, when recognizing the name of a person, when the program is recognized as a program name because "Band" is a part of another program name in "A rising band?" Recognize that it is not a program name and restore it. In addition, in the case of program names such as "Youth World Cup", if "World Cup" is regarded as the program genre and included in the "Name Terminology List", only "Youth" is recognized as the program name, You can determine if it can be an expanded program name and restore it.

Fourth, if the program name is long, it is often not known exactly the name of the program to find. For example, in a query where a user searches for a program with the "Audit column when we live in the world," the actual program title is "Audit column in our world." In the existing Exact matching method, the user cannot search for a desired program. . If the program name or human name is long, the syllable is missing, the spacing is incorrect, or the spelling is changed a bit, it is difficult to recognize it easily. To solve this problem, a robust search method is required even for slight spelling changes. To do this, all program names and names are bigram indexed, and if the program names or names are not clearly recognized through the previous three methods, they should be applied.

Hereinafter, with reference to the accompanying drawings will be described in detail a preferred embodiment of the present invention.

Prior to the description of the embodiments, the technical gist of the present invention prevents a decrease in search performance due to inconsistency between actual DB data and user input when searching for a program using a program name or a human name in an EPG search. In order to improve the EPG retrieval performance by applying a multi-level recognition method and error recovery method, and indexing program names and names in word units and bigram units. It will be easy to achieve.

1 is a system configuration diagram for implementing a broadcast program name and name management method in the EPG search service according to the present invention, the program name and name index section 100, the program name and name recognition section 102, EPG DB 104, program name and person index DB 106.

As shown in FIG. 1, the program name and person indexing unit 100 indexes the program name and the person name included in the EPG DB 104 through the program name indexing process.

The program name and name recognition unit 102 automatically recognizes the program name included in the query in the user's EPG search.

The program name and person index DB 106 stores program name and person name information indexed through the program name and person index unit 100.

Hereinafter, the broadcast program name and the name management method in the EPG search service according to the preferred embodiment of the present invention together with the above-described configuration will be described in detail with reference to FIGS. 2 and 3.

Prior to the description, the broadcast program name and name management method according to the present invention is largely divided into a program name and a person name indexing process (FIG. 2) and a program name and a person name recognition process (FIG. 3).

First, a program name and a person indexing process will be described.

As shown in FIG. 2, in the EPG search service according to the present invention, the broadcast program and the name indexing process use the optional additional information regular pattern 108 represented by a regular expression to select optional additional information from a name. Removing the step (S200), using the name stopword list 110, the word separation step of separating the word from the program name and person by word unit (S202), the space removal and symbols to remove or normalize unnecessary information contained in the name Normalization step (S204) and a bigram extraction step (S206) for extracting a bigram from the name. The specific processing methods and knowledge used for each step are as follows.

Optional additional information removing step (S200)

There are two ways to index names in this step.

First, the space is removed from the program name included in the EPG DB and the original name is indexed.

Second, unnecessary information included in the name is removed and indexed using the optional additional information regular pattern 108 shown in the following [Table 1].

input Build regular expression Print  Weekend Masterpiece <Italian Job>  ^ (Weekend Masterpiece | Youth Drama | Sunday Cinema | Masterpiece Masterpiece | World Masterpiece | Saturday Masterpiece) [] * <(. *)> $  Italian Job  The Unknown World Part 1-3 (. *) [] * ([0-9] + ~ [0-9] + parts) $  the world of unknown  MBC Deadline News  ^ (EBS | EBS TV | MBC | KBS1 | KBS2 | SBS | ESPNMBC-ESPN) (. *) $  Deadline News  Korean Series 3rd game live broadcast  (. *) [] * [0-9] + Charge \ ((Live | Satellite Broadcast | Live Broadcast | Total) \) $  Korea Series  Prison Break Season 3  (. *) (|) \ ((Season [] * [0-9]) \) $ (. *) (|) (Season [] * [0-9]) $ (. *) (|) Series [0-9] $ Prison Break  2007 KLPGA Inter-Burgo Masters <1R Record Relay>  ^ (200 [0-9]) [] * (PGA | LPGA | KPGA | KLPGA | JLPGA | European Tour | European Woman Tour) (. *) <[0-9] R (Recording | Live Broadcast | Rebroadcast | Live Broadcast )>  Inter-Burgo Masters  …  …  …

The program name includes 'channel name', 'year', 'specific contest', 'seasonal display', etc., which solves the problem of the user searching by selectively including or excluding the program name. For example, "07-08 Spanish Primera Liga Highlights", "Mnet Wide Entertainment News Highlights", "2007 KLPGA KB Kookmin Bank Star Tour 4th Tournament", "Lucky Young Ae Season 2" are "Primary Liga", The additional information will be removed so that you can search using "Wide Entertainment News", "Star Tour", "Lucky Young Ae". The regular pattern 105 uses a standardized regular expression and is semi-automatically constructed.

-Word separation step (S202)

In the word separation step, the name stop word list 110, which is hardly regarded as a program name, is used to separate the broadcast program name and the person's name in word units from the original broadcast program name and the person's name. The name stopword list 110 is as shown in the following [Table 2].

Stopword list type Stopword Channel name MBC, KBS1, KBS2, SBS, MBS-ESPN, YTN, tvN, Success TV, Fox Channel Partial channel name Espn, success, fox, channel, style, broadcast, sports, Genre name News, movies, drama, comedy, golf, soccer, baseball, Program Classification Name Health, Fashion Show, Short Drama, Time expression This week, today, tomorrow, next week, weekend, EPG Search Controllable Strings Turn on, turn, volume, record, reserve, … …

"Dongo Gifted's field reporter scandal" is divided into three program names such as 'Dokgo's Gifted', 'On-site reporter' and 'Scandal' for indexing. In this case, "Mnet Wide Entertainment News Highlights" excludes channel names such as "Mnet" or stopwords such as "Highlight" and indexes "wide" and "entertainment news."

-Space Removal and Symbol Normalization Step (S204)

In the space removal and symbol normalization step, spaces are removed and indexed from the program name of the EPG. In text input, general users input without considering spacing, so it is necessary to recognize this as the longest match. Also, foreign words, symbols, etc. are normalized to minimize inconsistency between index and search process. Usually, ':', '-', '!', ',', Etc. are deleted in the indexing process. In programs such as "Challenge! Supermodel", the "!" Is separated from the word "challenging!" Delete it.

-Bygram extraction step (S206)

In the Baigram extraction step, program names are indexed in units of Bygram for robust recognition of partial spelling errors. In particular, a spelling error of a user occurs in a foreign language expression. For example, "The challenge can spread model" to explore the stunning program name using the bayigeuraem information on these spelling errors, there is as much search confuse 'could' and 'shoe' to "challenge the shoe spread model." To be.

Next, a description will be given of the program name and name recognition process.

As shown in FIG. 3, the broadcast program and the name recognition method in the EPG search service according to the present invention are based on the DB indexed through FIG. Recognize people 'effectively.

The program name is recognized by the user's search request using a multi-level method. In the " EPG search &quot; block, if the user's query is a detailed information search using a program name or a program search using a program performer or producer name, the step moves to recognizing the program name or person name appearing in the question. Speech Act Analysis is performed to understand if it is a program search question, but this is omitted because it is not relevant to the present invention and a general statistical or rule-based speech act analyzer may be used.

Symbol normalization step (S300)

If a user's search request question contains symbols or foreign words, normalize them to match the indexed data. Symbols unnecessary for the search are deleted and, in the case of foreign words, normalized according to the promised method.

-Longest match search step (S302)

Find the longest match name from the left using the indexed name DB based on the query with white space removed from the input query. At this time, unnecessary program names are searched together according to the word unit name index.

This misidentified name is restored to the original input using the name context air information 112 shown in [Table 3].

Contextual Information example Survey (this / to / from / to / province / only) What time does she scare her? Who is Yoo Jae-seok? Search Vocabulary (Find, Search,) CSI New York Season 3 Browse Desperate Housewives … …

For example, suppose the program "MB! Meet with him", "Best room, meet there".

Program name Conversion code MBMeet with him Hat Hat 1 MB Hat Hat 2 With him Hat Hat 3 meeting Hat Hat 4 The best room meeting there Hat Hat 5 most Hat 6 Cool Hat 7 room Hat 8 That Hat Hat 9 Where Hat 10

By using this, "Find a meeting with him on MBC on weekends" is recognized as "Look for Hats3 and Hats4 sent by Hat2C on weekends on weekends." 'And' Hat Hat 8 'are restored to their original names, and the result is that "Look for Hat Hat 3 and Hat 4 broadcast on MBC on weekends". Thereafter, through several steps, they are combined and recognized as full names.

Partial Match Name Expansion Recognition Step (S304)

Since the program name is recognized in word units, when the user searches for a part of the program name, the program is partially recognized. In the example of the longest matching search step (S302), in the "Look for Hat3 and Hat4 broadcast on MBC on weekends", "Hat3" and "Hat4" are both partial program names that constitute 'Hat1'. It is integrated into a program name through the intersection of. Thus, it expands to "Look for Hat1 broadcasted by MBC on weekends". If there is no common program name through the intersection, it is recognized as the original name, "Look for Hat 3 and Hat 4 broadcast on MBC on weekends." To this end, the indexed program information has a full-part mapping table 114 as shown in Table 5 below.

Program name type Extension  Hat Hat 1 all  Hat 2, Hat 3, Hat 4  Hat Hat 2 part  Hat Hat 1  Hat Hat 3 part  Hat Hat 1  Hat Hat 4 part  Hat 1, Hat 5  Hat Hat 5 all  Sun 6, Sun 7, Sun 8, Sun 9, Sun 10, Sun 4  Hat Hat 5 part  Hat Hat 5  Hat 7 part  Hat Hat 5  Hat 8 part  Hat Hat 5  Hat Hat 9 part  Hat Hat 5  Hat 10 part  Hat Hat 5

Description of Table 5 is as follows. 'Hathat1' ("MB! Meet with him") consists of hathat2 ("MB"), hathat3 ("he"), hathat4 ("meeting"), and 'hathat1' was originally Express the title. Hat 2 ("MB") is the name of the partial program that makes up Hat 1 ("MB! Meet Him"). Also, Hat 4 ("Meet") indicates that it is part of the programs that make up Hat 1 or Hat 5. Using this information, the relationship between the partially recognized program names is grasped, and the expanded and recognized maximum program names are recognized.

Extended recognition step including stop words (S306)

Include Terminology In the expanded recognition phase, when a user searches for a program titled "discussion sibibi", when a user searches for "when debate sisibi", words representing a program genre such as "discussion" are included as stopwords. It is recognized as "discussion hat 3455 not when". If there are other programs that have "sishibi" as substrings or programs whose original title is "sishibi", several programs are searched. In this step, the context before and after the recognized program name is grasped, and when the front and rear words are stop words, when the stop word and the program name are combined and expandable, there is a combined program name. In the example above, the hat 3454 is expanded to the longest program name, including stopwords.

-Recognition error restoration step (S308)

When recognizing the program name in the preceding step, an error may occur. For example, in a search for "look for a pro with an rising band", the preceding step might recognize the name "band" as a program name, such as "look for a pro with an rising hat 23". At this time, in the recognition error restoration step (S308), such as 'information', 'appear', 'direct', 'produce' or 'twist' which may appear as a channel name, which may appear in a question such as personal information. Restore misrecognized names by using contextual information such as "turn back".

-Bygram recognition step (S310)

Unless the user remembers the exact program name, spelling, spacing, and some word skip errors occur. Since all names with possible errors cannot be indexed beforehand, a back-off method is needed to solve this problem. In the present invention, when none of the program names are recognized in the above-described steps (S300, S302, S304, S306, and S308), in the bygram recognition step (s310), the words other than the 'question_bigram_stopword' in the question aregrammed. Search the index DB by indexing with. Compare the question viagram with the indexed bygram, and search the program information for the index bygram that satisfies a certain weight or more. At this time, the 'question_bigram_stopword' collects and builds vocabularies that can be included in the question for EPG search. For example, EPG search or reservation related commands such as "Find", "Search", "Search", "Alarm", "Schedule", "Record", "Delete", and the channel name, genre name, and date expression are included. . This uses the stop word list [Table 2] used in the word separation step S202 and the stop word inclusion expansion recognition step S306.

As described above, in the EPG search, when searching for a program using a program name or a human name, in order to prevent a decrease in search performance due to inconsistency between actual DB data and user input, It is implemented to improve EPG search performance by applying multi-level recognition method and error recovery method based on indexes of life names in word units and bigram units.

Meanwhile, the embodiments of the present invention have been described in detail, but the present invention is not limited to these embodiments, and various modifications may be made by those skilled in the art within the spirit and scope of the present invention described in the claims below.

1 is a block diagram representing a process of indexing and searching a target name in an EPG DB for recognizing a program name or a human name in the present invention;

2 is a block diagram illustrating a specific method for indexing names in the present invention;

FIG. 3 is a diagram illustrating a detailed sequence of a method of recognizing a name based on an indexed name of the present invention. FIG.

Claims (7)

A broadcast program name and a name management method in the EPG search service, Multi-stage indexing of program names and names by word and by-gram unit; Recognizing the broadcast program and the name included in the search query of the user by using the indexed broadcast program name and the name and the contextual knowledge in the indexing process when the user's search query request Broadcast program name and name management method in the EPG search service comprising a. The method of claim 1, The indexing process, Removing the optional vocabulary from the search input target of the user using an optional additional information regular pattern; Normalizing a key to be indexed by removing spaces and symbol normalization for a broadcast program name and a person name from which the optional vocabularies are removed; Separating the program name and the person's name in word units from the original program name and the person's name by using a name stopword list that does not appear in the user's name search input target; Process of extracting individual viagrams from program name and person's name using name stopword list Broadcast program name and name management method in the EPG search service comprising a. The method of claim 2, The process of removing the optional vocabulary may include: Removing whitespace from the program name included in the EPG DB and indexing the original name; Process of removing and indexing unnecessary information included in a name using the optional additional information regular pattern Broadcast program name and name management method in the EPG search service comprising a. The method of claim 2, The normalization process is to normalize to minimize the discrepancy between the index and the search process for foreign words and symbols, the broadcast program name and name management method in the EPG search service. The method of claim 2, The extracting of the viagram may include indexing the viagram information on the user misspelling in the foreign language expression. The method of claim 1, The recognition process, A symbol normalization process for normalizing symbol / English / numeric representation included in a user's search query to recognize a program name and a person's name, In order to link the program name and person's name indexed by word, and the original program name and person's name, the relationship between the word and person's name and original program name and person's name indexed by word is analyzed and semantically linked. Configuring the partial name mapping table; Expanding and recognizing a partially matched name using the full-part name mapping table for a plurality of broadcast program names and names included in a search query to recognize broadcast program names and names; Searching for a name stopword in a pre-recognized broadcast program name and person's name and surrounding context of the previously recognized broadcast program name and a person's name, and expanding and recognizing the name stopword to a name including the stopword; A recognition error restoration process utilizing the context information of the pre-recognized broadcast program name and the name of the person to restore the mistaken portion of the pre-recognized broadcast program name and the person's name; Removing the stopwords from the search input when the final analysis result does not include the program name and the name of the program; Extracting the bygram after removing the stopwords, searching for the program information of the broadcast program name and the DB name and returning the broadcast program name and person name exceeding a certain weight among them as a search result Broadcast program name and name management method in the EPG search service comprising a. The method of claim 6, The process of recognizing by expanding the name containing the stopwords, Identifying the context before and after the recognized program name, A process of combining and recognizing a case in which the words of the preceding and later contexts are the stopwords and when there are program names that can be expanded by combining the stopwords and the program names. Broadcast program name and name management method in the EPG search service comprising a.

Images