JPH0522940B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a document editing device that can automatically or semi-automatically abstract an input text and further create a summary thereof.

[Technical background of the invention and its problems]

Recently, various text analysis and creation devices using computer systems have been attracting attention. Among these, word processors, machine translation devices, and the like have been developed as typical devices that utilize these technologies.

However, conventional devices of this type simply convert sentences created by humans into another sentence format, and do not perform any intellectual processing. That is,
Japanese word processors simply convert input kana character strings into a sentence format that includes kana and kanji, while machine translation simply converts, for example, Japanese sentences into English sentences. I'm just doing it. For this reason, even if text analysis technology was introduced, it was difficult to immediately decipher the content of a given text. Furthermore, even if input text contents were to be filed, a document storage device with an enormous storage capacity was required.

[Purpose of the invention]

The present invention has been made in consideration of these circumstances, and its purpose is to provide a document editing device that can automatically or semi-automatically generate an abstract or summary of an input text. Our goal is to provide the following.

[Summary of the invention]

The present invention analyzes each sentence in an input sentence syntactically or semantically, and combines the entire or part of the syntactically or semantically analyzed sentence structure with a partial sentence structure registered in advance in a dictionary ( Sentence key structure) is extracted from the input text to extract sentences containing partial sentence structures registered in the dictionary, and the sentences are reconstructed using these extracted sentences. It is. Further, the second invention is such that, for example, a sentence element corresponding to a deletion rule in a partial sentence structure registered in a dictionary is deleted from the sentence extracted from the input sentence to reconstruct the sentence. be.

In other words, whereas in the past, documents were generally edited using words (keywords) in a sentence as clues, in the present invention, the connections between sentence elements that make up a sentence,
That is, document analysis is performed using the partial key structure of a sentence as a clue.

〔Effect of the invention〕

Thus, according to the present invention, it is possible to determine whether the sentence structure obtained by syntactic (semantic) analysis of each sentence in an input sentence includes a key structure having an important meaning that is registered in advance in a dictionary. Therefore, it is possible to determine whether the sentence has an important meaning in the input text, and according to the result of this determination, only the sentences with the important meaning are extracted from the input text, and the abstract is extracted. It becomes possible to compose sentences easily.

In addition, a deletion rule based on the structural (semantic) analysis of the key structure is introduced into the key structure, and unnecessary sentence elements in the sentence are deleted according to this deletion rule to reconstruct the sentence. , it becomes possible to easily create a summary sentence.

Therefore, it becomes possible to easily and effectively abstract or summarize the input text, and a great effect can be achieved in document editing.

[Embodiments of the invention]

Embodiments of the present invention will be described below with reference to the drawings. In this example, the input text is edited to obtain both an "abstract text" and a shorter "summary text," but depending on the purpose, it is possible to obtain only one of the sentences. You can also do this.

This type of device includes a text information input section, an input text analysis/edit processing section, a text information display section, a text information storage section, and the like. FIG. 1 is a schematic configuration diagram of the main part (analysis/editing processing section) in the embodiment apparatus. Further, FIG. 3 is a diagram showing a processing algorithm for abstracting an input text, and FIG. 4 is a diagram showing a processing algorithm for creating a summary sentence.

In FIG. 1, the sentence extraction unit 1 sequentially extracts sentences constituting an input sentence from an input sentence, and extracts one sentence from an input sentence consisting of a plurality of sentences (step 11 in FIG. 3). ). This sentence extraction part 1
The sentence extracted through the . The sentence analysis unit 2 sequentially analyzes the sentence by referring to the grammar rules, part of speech information, etc. registered in the sentence analysis dictionary 3, and obtains the sentence structure of the extracted sentence (Step 12 in Figure 3). , Fig. 4 Step 18). Here, the sentence structure is a representation of the elements (segments and words) that make up a sentence, along with their mutual connections. The sentence analyzed by the sentence analysis section 2 is provided to the key structure matching section 4.

The key structure matching unit 4 matches the whole or part of the sentence structure analyzed by the sentence structure unit 2 with the key structure registered in advance in the key structure dictionary 5, and identifies sentences that match this key structure. It is determined whether the structure exists in the analyzed sentence (Step 13 in Figure 3, Step 19 in Figure 4).

The key structure registered in the key structure dictionary 5 here is one in which the constituent elements of a sentence are described as their connected relationships, as shown in FIG. 2a, for example. It is set according to the text content to be processed. Therefore, the key structure registered in the key structure dictionary 5 is given as a structure of a sentence having an important meaning of the text to be processed. Additionally, deletion rules for sentence constituent elements are also registered in this key structure dictionary 5, as shown in FIG. 2b. This deletion rule means that the connection between the constituent elements of the sentence is "adverb of time",
In the case of "quantity", "sentence adverb", etc., it consists of something that specifies deletion of the sentence constituent element. This deletion rule suggests deletion of unnecessary elements in the sentence, that is, elements that have no important meaning in terms of the content of the sentence. Here, inside [[ ]] are elements that can be deleted,
Also, * indicates an arbitrary element.

The key structure matching unit 4 uses such a key structure dictionary 5.
The sentence extracted from the input sentence with reference to
It is determined whether the sentence contains the above-mentioned key structure that has an important meaning in interpreting the sentence. In this way, when the key structure matching unit 4 detects a sentence component that matches the key structure in the analyzed sentence structure (Step 14 in Figure 3), it extracts the entire sentence (matching sentence) that includes this matching component. and provides it to the sentence reconstruction unit 6 (Step 15 in Figure 3). The key structure matching unit 4 sequentially executes this extraction process on a plurality of sentences making up the input text (step 16 in FIG. 3). The sentence construction unit 6 combines the sentences thus extracted to extract an abstract sentence, and generates an abstract sentence for the input sentence from the extracted sentences (step 17 in FIG. 3).

By the way, the key structure matching unit 4 also determines whether or not there is a sentence component that matches the deletion rule in the key structure dictionary mentioned above in the analyzed sentence. (Step 20 in Figure 4). This is done for all sentence components that can be deleted in the analyzed sentence (Step 21 in Figure 4). For the parsed sentence that has been subjected to this deletion process, the sentence reconstruction unit 6 grammatically combines the remaining sentence constituent elements that were not deleted according to the instructions of the key structure dictionary 5 and outputs it as a summary sentence. be. In other words, in the sentence extracted from the input sentence, a sentence that further summarizes the content of the extracted sentence is generated (see step 4 in Figure 4).
twenty two).

Next, an example of creating an abstract text will be explained with reference to FIG.

FIG. 5a shows an example of an input sentence, and FIG. 5b shows a second sentence extracted from the input sentence. This input text consists of a greeting card that says ``retirement/re-employment,'' and the second sentence is ``Well, I, myself, have retired from Tozai Co., Ltd. as of March 31st. ” It consists of the following sentences. Analyzing this sentence syntactically (semantically), the elements of ``I have retired'' are ``retire; verb, past tense, polite tense,...''
It is semantically analyzed as Similarly, other elements are also subjected to semantic analysis. The sentence analysis unit 2 further structurally analyzes the sentence and obtains the sentence structure as a connection relationship between each element as shown in FIG. 5c. In this case, the element ``now'' connects to ``retire'' with an adverbial meaning,
``I'' has the meaning of the subject, ``Tozai Co., Ltd.'' as the object, ``March 31st'' as an expression of time, and ``this time'' as an adverb of time for ``retiring''. Connect. This connection relationship is the analyzed structure of the sentence.

The key structure collation unit 4 compares and determines whether or not there is a matching key structure in the key structure dictionary 5 for the entire sentence structure shown in FIG. 5c or a partial structure thereof. In this case, since the key structure dictionary 5 includes, for example, "[retire] - (subject) - [I]" as a key structure, this sentence constitutes a sentence with an important meaning in the input sentence. It is determined that the According to this determination result, this second sentence is extracted from the input sentence as an abstract sentence and provided to the sentence reconstruction unit 6.

Similarly, for other sentences in the input sentence,
The sentence structure is analyzed and compared with the key structure, and only sentences including the key structure are extracted from the input sentence and provided to the sentence reconstruction unit 6. Then, the sentence reconstruction unit 6 combines these sentences, for example, as shown in FIG.
An abstract text of the input text is created as shown in FIG. In other words, sentences that have no important meaning in terms of content, such as seasonal greetings and conclusion, and modified sentences in the main text are removed, and the abstract text consists only of sentences that have important meaning in terms of content. will be created.

Also, the sentence summary is performed as follows. 6th
The figure shows an example of this.As shown in figure a, ``Well, one year has already passed since we opened our store here. ” If a sentence is given, this sentence will be analyzed syntactically (semantically) as the structure shown in Figure b. When this sentence structure is compared with the key structure dictionary 5, for example, "[sugar] - (adverb of time) - [[*]]" exists as a deletion rule, and "sugar" ( This shows that the word (*; regardless of its elements) that is combined as an adverb of time (adverb of time) can be omitted from its sentence structure. Similarly, it is shown that the word (element) "year" can also be omitted. When the structure of the sentence is organized in this way according to the key structure deletion rule, the structure is simplified as shown in Figure 6c, and the elements indicating the sentence structure have important meanings in terms of the content of the sentence. Only those with . According to the sentence structure arranged in this way, the sentence structure part 6 may be written as, for example, as shown in Figure 6d, ``One year has passed since we opened our store.'' ” You will get a summary sentence. Note that the reconstruction of this summary sentence is performed by, for example, referring to the flow of a given sentence, but it can also be done by semantically analyzing it from the sentence structure shown in Figure 6c above. good.

As described above, according to this device, the sentence structure is obtained by syntactically (semantically) analyzing the sentence, and the sentence structure is compared with the key structure to perform abstract processing or summary processing. You can intelligently edit written texts. Therefore, great effects can be achieved, such as making it easier to understand the contents of the input text and making it possible to improve the efficiency of the document file. Also, unlike traditional keyword matching,
Since the text is processed by capturing the content of the sentence as the key structure of the sentence, effects such as being able to perform accurate editing work can be achieved.

Note that the present invention is not limited to the embodiments described above. For example, the key structure to be registered in the key structure dictionary 5 may be set in advance according to the type of text to be edited. Furthermore, in addition to sentence deletion rules, replacement rules, addition rules, etc. can also be considered. In this case, the way the rules are written is not based on a single key concept, but may also be in the form of a rewriting rule from key structure 1 to key structure 2, for example. Furthermore, the sentence structure analysis algorithm can be modified in various ways. In short, the present invention can be implemented with various modifications without departing from the gist thereof.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram of the main parts of a device according to an embodiment of the present invention, FIG. 2 is a diagram showing an example of the structure of a key structure dictionary in the device according to the embodiment, and FIGS. 3 and 4 are text abstract processing and summarization. Figures 5 and 6, which are diagrams showing the flow of processing, are diagrams showing examples of abstract processing and summary processing, respectively. 1... Sentence extraction section, 2... Sentence analysis section, 3... Sentence analysis dictionary, 4... Key structure matching section, 5... Key structure dictionary, 6
...Sentence Reconstruction Department.

Claims (1)

[Claims] 1. A dictionary in which a plurality of partial sentence structures are registered in advance according to the content of a text to be processed, and each sentence in the inputted text to be processed can be syntactically or semantically means for analyzing, means for comparing the whole or a part of the sentence structure analyzed by this means with the partial sentence structure registered in the dictionary, and a means for comparing the whole or part of the sentence structure analyzed by this means with the partial sentence structure registered in the dictionary, and Document editing characterized by comprising means for extracting a sentence including a registered partial sentence structure from the text, and means for creating a sentence combining the sentences extracted by this means. Device. 2. A first dictionary in which a plurality of partial sentence structures are registered in advance according to the text content to be processed, and a dictionary created based on the connections between sentence components according to the text content to be processed. a second dictionary in which a plurality of deletion words according to deletion rules are registered in advance; a means for syntactically or semantically analyzing each sentence in the input sentence to be processed; means for comparing the whole or a part of the sentence structure analyzed by the method with the partial sentence structure registered in the first dictionary; a means for extracting a sentence including a partial sentence structure from the sentence, and deleting a deleted word from the extracted sentence according to a deletion rule registered in the second dictionary; and a sentence remaining by this means. 1. A document editing device comprising means for grammatically combining constituent elements to create a sentence.