KR101541170B1 - Apparatus and method for summarizing text - Google Patents

Abstract

The present invention relates to an apparatus for summarizing text based on a resource description framework (RDF), which comprises: an RDF triple conversion unit for converting an input text made up of a plurality of sentences into a plurality of RDF triples; a main triple determination unit for extracting one or more main triples from the RDF triples based on at least one among qualities and vocabularies included in a predicate; and a summary generation unit for generating a summary by extracting a sentence corresponding to the main triples from the input text.

Description

[0001] APPARATUS AND METHOD FOR SUMMARIZING TEXT [0002]

The present invention relates to a text summarizing apparatus and method.

Currently, we live in a big data era where the amount of text that is often posted on social media such as news, blog, SNS as well as various documents is increasing. Therefore, excessive time is required to select only the text of interest among many texts. If the computer understands the meaning of the text on behalf of the person, analyzes the emotion of the preference / non-preference, and summarizes the main contents of the text, we can easily improve our everyday life and make useful changes in the lifestyle of mankind. Can be imported. However, computers have limitations that can not sufficiently deal with natural language flexibility and expressive power. Currently, the way the computer performs text summarization is to extract sentences containing vocabulary with the most frequently mentioned words. Therefore, it is difficult to provide practical service to the users with a simple method that ignores this meaning.

Research on automatic processing of natural language on computers has long been attempted, but it was not enough to understand the meaning of the text. Recently, a technique of understanding the meaning of text by converting it into a resource description framework (RDF) triple, which is a format that a computer can understand, is being studied. The RDF triple is an international standard, governed by the World Wide Web Consortium (W3C), which provides knowledge and information to subjects [resource], predicate [property], object [ As shown in Fig. However, the methodology presented so far has limitations in that it can not sufficiently implement emotional analysis, summarization and reporting of text.

International patent application Laid-open Publication No. 1997-7007499 (published on December 01, 1997) Korean Unexamined Patent Publication No. 2003-0039575 (published May 22, 2003) Korean Patent Publication No. 10-2009-0003090 (published on Jan. 09, 2009)

An object of the present invention is to provide an apparatus and method for analyzing emotional contents of a text based on an RDF triple to summarize and report the text.

An apparatus for summarizing text based on a Resource Description Framework (RDF) according to an embodiment of the present invention includes: an RDF triple generator for converting an input statement composed of a plurality of sentences into a plurality of RDF triples; Extracting at least one main triple from among the plurality of RDF triples based on at least one of a quality and a lexicon included in the at least one main triple, and extracting a sentence corresponding to the at least one main triple in the input statement And a summary creating unit for creating a summary.

The main triple determination unit may include a first triple extracting unit for extracting, among the plurality of RDF triples, a triple including the semantic qualities assigned to the predicate kit by the main triple.

The designated semantic feature may be one of a preference feature and a non-preference feature.

The main triple determination unit may further include a second triple extracting unit for extracting, from the plurality of RDF triples, a triple including a vocabulary of a type specified in the predicated kit, as a main triple.

The vocabulary of the specified type may include at least one of "to", "to", "to", "to", "can", and "can not".

Wherein the summary creating unit includes a main sentence extracting unit for extracting at least one main sentence corresponding to the at least one main triple in the input sentence, a non-essential sentence extracting unit for removing the specified phrase from each of the at least one main sentence, And a summary generation unit for collecting the refinement sentences generated from the unnecessary phrase removal unit to generate a summary sentence.

The designated phrase may correspond to any one of a conjunction, an adjective, an idiom, and an exclamation point.

The text summarizing apparatus may further include a summary report reporting unit for reporting the summary prepared by the summary preparing unit to a user.

According to another embodiment of the present invention, there is provided a method for a device to summarize text based on a Resource Description Framework (RDF), comprising: converting an input statement composed of a plurality of sentences into a plurality of RDF triples; Extracting, from the RDF triple, a triple in which at least one of a specified semantic qualification and a designated type of lexicon is included in a predicated kit, and extracting a sentence corresponding to the main triple in the input sentence, .

The step of extracting with the main triple may extract a triple including one of the preference qualities and non-preference qualities of the plurality of RDF triples included in the predicate kit with the main triple.

Wherein the step of extracting with the main triple is a step of selecting one of the plurality of RDF triples as one of the " to ", "to "," to ", "to "," The triple included in the kit can be extracted into the main triple.

Wherein the step of generating the summary comprises extracting at least one main sentence corresponding to each of the major triples in the input statement, removing a specified phrase from the at least one main sentence, And collecting one sentence to form the summary.

The designated phrase may be a phrase corresponding to any one of a conjunction, an adjective, an adjective, and an adjective.

The text summary method may further include reporting the summary to the user.

According to the embodiment of the present invention, the text summarizing device can grasp the meaning of the text on behalf of the user and extract the contents including the emotion of the preference / non-preference to summarize the text. According to the embodiment of the present invention, a computer reads a large amount of text on behalf of the user, extracts and summarizes semantically important sentences, and allows the user to quickly and easily recognize important texts without reading and selecting all the texts. Thus, according to embodiments of the present invention, the user can also reduce the risk of overlooking text of interest or important text.

1 is a block diagram of a text summarizing apparatus according to an embodiment of the present invention.
2 is a flow diagram of a text summarization method in accordance with an embodiment of the present invention.
3 is a flowchart of a method for generating an RDF triple according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly illustrate the present invention, parts not related to the description are omitted, and similar parts are denoted by like reference characters throughout the specification.

Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

A text summarizing apparatus and method according to an embodiment of the present invention will now be described with reference to the drawings.

1 is a block diagram of a text summarizing apparatus according to an embodiment of the present invention.

Referring to FIG. 1, a text summarizer (hereinafter referred to as a "summarizer") 10 converts text composed of a plurality of sentences into RDF (Resource Description Framework) triples, And RDF triples selected with reference to the lexical form.

The abstracting apparatus 10 includes a text input unit 100, an RDF triple generation unit 200, an RDF triple repository 300, a main triple determination unit 400, a summary text generation unit 500, and a summary text reporting unit 600 .

The text input unit 100 receives a text (input statement) composed of at least one sentence. The text input unit 100 can receive various kinds of text such as a web document such as e-mail, Internet news, social media, and a word processor document.

The RDF triple generator 200 converts an input sent to the text input unit 100 into an RDF triple and stores it in the RDF triple repository 300. The RDF triple generating unit 200 includes a morpheme analyzing unit 210, a word generating unit 220, a sentence component analyzing unit 230, and an RDF triple converting unit 240.

The morpheme analysis unit 210 analyzes the input sentence using morpheme analyzer and electronic dictionary as morphemes. A morpheme is the smallest unit of meaning that makes up a sentence. The electronic dictionary uses stemming as a heading and includes grammatical and semantic qualities of each morpheme. Meaning qualities include preference / non-preference qualities. The preference / non-preference qualities are qualities related to emotion.

The phrase generator 220 generates a word phrase based on the morpheme. A phrase is a sentence component that is separated by a space in a spelling sentence. The vernacular is divided into cognition (NN), verb (VV), idiom (MM), adverb (MA), exclamation (IC), and connective (CONJ)

The sentence component analyzing unit 230 analyzes a role in a sentence of a word, that is, a sentence component. Sentence components are divided into subject (SBJ), object (OBJ), predicate (PRD), bore (CMP), modifier (MOD), additional word (AJT), connection word (CNJ) and independent word (INT).

The RDF triple conversion unit 240 generates the RDF triple conversion unit 240 based on the sentence component and sentence segment information analyzed by the morpheme analysis unit 210, the word generating unit 220, and the sentence component analysis unit 230, To an RDF triple. The RDF triple consists of three pairs of subjects, predicate kits, and objects.

The RDF triple repository 300 stores the converted RDF triples in the RDF triple conversion unit 240.

The main triple decision unit 400 selects the main RDF triple among the input query RDF triples stored in the RDF triple repository 300. At this time, the main triple decision unit 400 analyzes the predicate kit of the RDF triple to select the main RDF triple. The main triple determination unit 400 includes an emotional triple extraction unit 410, a main triple addition extraction unit 420, and an output unit 430.

The emotional triple extractor 410 extracts, among RDF triples of the input query, an RDF triple having a specified semantic qualification, for example, preference / non-preference qualities. A Freddie kit in an RDF triple can contain a variety of qualities. If there is a predicate qualification of a Freddie kit, you can extract the RDF triple as an emotional triple and more specifically tag it with a preferred triple . Likewise, if the qualities of the Freddie kit are 'non-favorable', the RDF triple can be extracted as an emotional triple, and more specifically tagged as a non-preferred triple. For example, the sentence "Smartphone A has good image quality, but the reaction is bad" is converted to the RDF triple shown in Table 1. Since the Freddie kit "good" of triple 1 is a favorable quality and the Freddie kit "bad" of triple 1 is a non-favorable quality, emotional triple extractor 410 extracts triple 1 and triple 2 as emotional triples.

Subject Freddie Kit Object Triple 1 Smartphone A good Quality Triple 2 Smartphone A bad Reaction

The main triple adder extractor 420 extracts an RDF triple including a certain type of vocabulary in addition to the emotional triple among the input query RDF triples. The vocabulary of a specific type designated by the main triple addition extractor 420 may vary, and the user can set desired information for each user. For example, the main triple adder extractor 420 may have a form of a form such as " to ", "to ", " A triple can be extracted as a major triple.

The output unit 430 outputs the triples extracted by the emotional triple extraction unit 410 and the main triple addition extraction unit 420 to the main RDF triple.

A method by which the main triple decision unit 400 determines a main RDF triple will be described in the following, for example.

The input is: "The smart watch has a 2-inch curved display, the base display is an analog needle watch, and the watch design with a crisp OLED display is just as flawless. "In the future, we plan to launch a new strap with a new design and materials," said A Electronics president and chief executive officer of the RDF Triple Repository (300). RDF triples as shown in Table 2 are stored.

Subject Freddie Kit Object Triple 1 Screen of Smart Watch be used 2 inch curved display Triple 2 Default Display Wallpaper to be Analog needle clock picture Triple 3 ? x great Clock design Triple 4 ? x Offer Plain wrist strap of rubber material Triple 5 ? x Offer Bracelet-shaped strap Triple 6 ? y You can choose ? x Triple 7 A electron I plan to release it. Strap with new design and material

The main triple decision unit 400 determines whether the predicate kits have a preference / non-preference qualification or a specific type of vocabulary (for example, "to "," to ", " Quot ;, "possibly "," not "). The main triple decision unit 400 includes a triple 3 with a favorable quality in the Freddie kit as shown in Table 3 and a triple 1 with a specific type of vocabulary in the Freddie kit, ), Triple 5 (~), triple 6 (possibly) as the main RDF triple.

Subject Freddie Kit Object Triple 1 Screen of Smart Watch It is being used 2 inch curved display Triple 3 ? x GRT is Clock design Triple 4 ? x And provide the Plain wrist strap of rubber material Triple 5 ? x And provide the Bracelet-shaped strap Triple 6 ? y You can choose ? x

The summary-statement creating unit 500 creates a summary based on the main RDF triple extracted by the main triple determining unit 400. The summary-statement creating unit 500 includes a main-sentence extracting unit 510, an unnecessary-phrase removing unit 520, and a summary-generating unit 530.

The main sentence extracting unit 510 extracts a sentence corresponding to the main RDF triple from the input sentence of the text input unit 100. When the main RDF triple is in Table 3, the main sentence extracting unit 510 extracts a sentence corresponding to the triple 1 (a 2-inch curve display is used for the smart watch screen), a sentence corresponding to the triple 3 The design of the watch through the display is unmistakeable), Triple 4 to Triple 6 (here you can find a custom rubber strap and a bracelet strap to choose from .) Is extracted as a main sentence. The main sentences are as follows.

<Main Sentence> The screen of SmartWatch is a 2 inch curved display. Once a clear OLED display with a watch design is not flawless. Here, both rubber strap and wrist strap are available, so you can choose according to your preference.

The unnecessary phrase removal unit 520 removes unnecessary phrases from the main sentence extracted by the main sentence extraction unit 510 to refine the sentence. Unnecessary phrases can be set in a variety of ways, for example, the unnecessary phrase eliminator 520 removes the conjunctions, adjectives, adjectives, and exclamations in the main sentence.

The summary-statement generating unit 530 collects the sentences refined by the unnecessary-phrase removal unit 520 and outputs them as a summary. For example, the summary-statement generation unit 530 may output the following summary sentences in which the conjunction, adverb, idiomatic, and exclamation have been removed from the main sentence.

<Summary>SmartWatch's screen was a 2-inch curved display. The clock design is amazing. It features a rubber strap and a bracelet strap for your personal preference.

The summary report generator 600 reports the summary to the user. The method of reporting to the user may be various, and the summary report reporting unit 600 may transmit the summary to the terminal designated by the user.

2 is a flow diagram of a text summarization method in accordance with an embodiment of the present invention.

Referring to FIG. 2, the summary apparatus 10 converts an input statement into a plurality of RDF triples (S110).

The summary apparatus 10 extracts, among the plurality of RDF triples, a triple including a semantic feature of interest into a major triple in a Freddie kit (S120). The semantic qualities of interest may be set differently depending on the nature of the input query, the user's interests, and the like. For example, the summary device 10 may set the emotional qualities such as preference / non-preference qualities to the interesting semantic qualities of interest, and the triple with preferred / non-preference expressions such as "good" .

The summary apparatus 10 extracts, among the plurality of RDF triples, a triple including a vocabulary of a designated type as a main triple in a predicated kit (S130). For example, the digesting device 10 may select a triple with a vocabulary of " to ", "to "," do.

The summary apparatus 10 extracts a main sentence corresponding to the main triple in the input statement (S140). The summary device 10 can extract all the sentences corresponding to the main triple as the main sentence, but selects the main sentence according to the weight among the sentences corresponding to the main triple in consideration of the number of sentences, the length, and the number of characters . At this time, the summary apparatus 10 can set a weight of a sentence corresponding to a triple including a semantic feature of interest to a high value.

The summary apparatus 10 removes the unnecessary phrases specified in the main sentence (S150).

The summary apparatus 10 creates a summary with sentences whose unnecessary phrases have been removed (S160).

The summary device 10 reports the summary to the user (S170).

3 is a flowchart of a method for generating an RDF triple according to an embodiment of the present invention.

Referring to FIG. 3, the summary apparatus 10 receives an input statement composed of a plurality of sentences (S210).

The summary apparatus 10 analyzes the input sentence as a morpheme (S220).

The summary apparatus 10 generates a word based on the morpheme of the input query (S230).

The summary apparatus 10 analyzes the sentence components of the input query (S240).

The summary apparatus 10 converts each sentence into an RDF triple based on the sentence analysis information of the input query (S250). The RDF triple is stored in the RDF triple repository 300.

In this manner, the summary device 10 can grasp the meaning of the text on behalf of the user, and can abstract the text by extracting semantically important sentences. Thus, the summary device 10 reads a vast amount of text instead, summarizes and reports semantically significant sentences, so that the user can quickly and easily identify important text without having to read and select all the text. In addition, the user can also reduce the risk of overlooking text of interest or important text.

The embodiments of the present invention described above are not implemented only by the apparatus and method, but may be implemented through a program for realizing the function corresponding to the configuration of the embodiment of the present invention or a recording medium on which the program is recorded.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It belongs to the scope of right.

Claims (14)

An apparatus for summarizing text based on a Resource Description Framework (RDF)
An RDF triple generator for converting an input sentence composed of a plurality of sentences into a plurality of RDF triples,
A main triple decision unit for extracting at least one main triple among the plurality of RDF triples based on at least one of a quality and a lexicon included in the Freddie kit,
Extracting a sentence corresponding to the at least one major triple in the input sentence and creating a summary sentence;
/ RTI &gt; The method of claim 1,
The main triple determination unit
A first triple extracting unit for extracting, from the plurality of RDF triples, a triple including the semantic qualities assigned to the predicated kit,
/ RTI &gt; 3. The method of claim 2,
Wherein the designated semantic feature is one of a preference feature and a non-preference feature. 3. The method of claim 2,
The main triple determination unit
A second triple extracting unit for extracting, from the plurality of RDF triples, a triple including a vocabulary of a type specified in the predicated kit,
Further comprising: 5. The method of claim 4,
The vocabulary of the specified type
A text summarizing device comprising at least one of "to", "to", "to", "to", "can" and "not". The method of claim 1,
The summary-
A main sentence extracting unit for extracting at least one main sentence corresponding to the at least one main triple in the input sentence,
Removing unnecessary phrases from each of said at least one main sentence to generate a respective refinement sentence, and
A summary sentence generation unit for collecting the refined sentences generated in the unnecessary phrase removal unit and generating a summary sentence,
/ RTI &gt; The method of claim 6,
Wherein the designated phrase is a phrase corresponding to one of conjunction, adverb, adjective, and exclamation. The method of claim 1,
A summary report reporting unit for reporting the summary to the user,
Further comprising: A method for a device to summarize text based on a Resource Description Framework (RDF)
Converting an input statement composed of a plurality of sentences into a plurality of RDF triples,
Extracting, among the plurality of RDF triples, a triple including at least one of a specified semantic qualification and a lexicon of a designated type in a predicate kit as a main triple; and
Extracting a sentence corresponding to the main triple from the input sentence and creating a summary sentence
&Lt; / RTI &gt; The method of claim 9,
The step of extracting with the main triple
And extracting, from the plurality of RDF triples, a triple in which one of the preference qualities and the non-preference qualities is included in the predicate kit. The method of claim 9,
The step of extracting with the main triple
Wherein a triple included in the Freddie kit is selected from among the plurality of RDF triples as one of " A text summarization method that extracts into triples. The method of claim 9,
The step of creating the summary
Extracting at least one main sentence corresponding to each of the major triples in the input statement,
Removing the specified phrase from the at least one main sentence, and
Collecting at least one sentence from which the designated phrase has been removed,
&Lt; / RTI &gt; The method of claim 12,
Wherein the designated phrase is a phrase corresponding to one of a conjunction, an adjective, an adjective, and an exclamation. The method of claim 9,
The step of reporting the summary to the user
&Lt; / RTI &gt;

Images