Solution data edit processing apparatus and method, and automatic summarization processing apparatus and method

A solution-data edit processing method for editing solution data which is used in automatic summarization processing of a digitized text data by a machine learning method, the method comprising:

• obtaining text data and summary data of the text data;
• displaying either the text data or the summary data or both of them;
• obtaining evaluation information which includes measure values chosen by a user from two or more predetermined levels in each of the qualities of the summarization and setting a set of measure values of the evaluation information as a user's evaluation information;
• and
• generating solution data which uses the text data and the summary data as a problem and uses the evaluation information as a solution, and outputting the solution data.

An automatic summarization processing apparatus for generating

• summary data from digitized text data using a supervised machine learning method with solution data which consists of text data and its summary data as a problem and user's evaluation information as its solution, the apparatus comprising:
• means for obtaining text data and summary data of the text data;
• means for displaying either the text data or the summary data or both of them;
• means for, when the summary data is displayed, obtaining evaluation information which includes measure values chosen by a user respectively from two or more predetermined levels in each qualities of the summarization and setting a set of measure values of the obtained evaluation information as a user's evaluation information;
• and
• means for generating solution data which uses the text data and the summary data as a problem and uses the evaluation information as a solution, and storing the solution data in a solution data storage unit;
• means for extracting the solution data from the solution data storage unit and obtaining features of the problem in the extracted solution data by semantically and syntactically analyzing said solution data;
• means for generating a combination with the features of the problem and the user's evaluation information in the solution data;
• means for learning what solution is apt to be produced with what set of features from the combination and storing a result of learning in a learning result-data storage unit;
• means for generating summary candidate data from input text data;
• means for generating a combination with the input text data and each one of the summary candidate data;
• means for obtaining features of each of the combination by semantically and syntactically analyzing said combination;
• means for estimating what solution is apt to be produced with features of the combination based on the result of learning, obtaining a certainty factor as a solution and generating a pair of the summary candidate data and its a certainty factor;
• and
• means for selecting the pair having a largest certainty factor and extracting the summary candidate data of the selected pair as the summary data of the inputted text data.

An automatic summarization processing apparatus for generating summary data from digitized text data using a supervised machine learning method with solution data which consists of text data and its summary data as a problem and user's evaluation information as its solution, the apparatus comprising:

• means for obtaining text data and summary data of the text data;
• means for displaying either the text data or the summary data or both of them;
• means for, when the summary data is displayed, obtaining evaluation information which includes measure values chosen by a user from two or more predetermined levels in each of the qualities of the summarization and setting a set of measure values of the obtained evaluation information as a user's evaluation information;
• and
• means for generating solution data which uses the text data and the summary data as a problem and uses the evaluation information as a solution, and storing the solution data in a solution data storage unit;
• means for extracting the solution data from the solution data storage unit and obtaining features of the problem in the extracted solution data by semantically and syntactically analyzing said solution data;
• means for generating a combination with the features of the problem and one among the solution and solution candidates which are other set of measure value set with all the preset measure values excepting the set of measure values of the user's evaluation information;
• means for learning how probabilities of being a positive example or a negative example are produced at what set of features of the combination and storing a result of learning in a learning-result-data storage unit;
• means for generating summary candidate data from input text data;
• means for generating a combination with the input text data and one each of the summary candidate data;
• means for obtaining features of each the combination by semantically and syntactically analyzing each combination;
• means for estimating what probabilities of being a positive example or a negative example are produced with features of the combination based on the result of learning, obtaining a probability of being a positive example as a solution and generating a pair of the summary candidate data and its probability;
• and
• means for selecting the pair having a largest probability of being a positive example and extracting the summary candidate data of the selected pair as the summary data of the inputted text data.

An automatic summarization processing apparatus for automatically summarizing a digitized text by a machine learning method using solution data which uses a text as a problem and uses the summary of the text as a solution, the processing apparatus comprising:

• solution-data storage means which changes the solution data in accordance with a user's instruction;
• machine learning processing means which extracts a pair of a feature set of the problem and the solution from the solution data, learns what solution is apt to be produced from what feature from the pair, and stores learning result data in learning-result-data storage means;
• summary-candidate generation processing means which generates a summary candidate from input text data;
• generation means of a pair of a summary candidate and an estimated solution, which extracts a pair of a feature set and solution candidate from the text data and the summary candidate, estimates probability of a positive example or probability of a negative example is-produced at what pair of the solution or solution candidate and a feature set based on the learning result data, and generates a pair of the summary candidate and estimated solution;
• and
• solution estimation processing means which extracts a feature set from input text data, and estimates what solution is apt to be produced from the feature set based on the learning result data.

An automatic summarization processing method for generating summary data from digitized text data using a supervised machine learning method with solution data which consists of text data and its summary data as a problem and user's evaluation information as its solution, the method comprising:

• obtaining text data and summary data of the text data;
• displaying either the text data or the summary data or both of them;
• obtaining evaluation information which includes measure values chosen by a user from two or more predetermined levels in each of the qualities of the summarization and setting a set of measure values of the obtained evaluation information as a user's evaluation information;
• and
• generating solution data which uses the text data and the summary data as a problem and uses the evaluation information as a solution, and storing the solution data in a solution data storage unit;
• extracting the solution data from the solution data storage unit and obtaining features of the problem in the extracted solution data by semantically and syntactically analyzing said solution data;
• generating a combination with the features of the problem and the user's evaluation information in the solution data;
• learning what solution is apt to be produced with what set of features from the combination and storing a result of learning in a learning result-data storage unit;
• generating summary candidate data from input text data;
• generating a combination with the input text data and one each of the summary candidate data;
• obtaining features of each the combination by semantically and syntactically analyzing said combination;
• estimating what solution is apt to be produced with features of the combination based on the result of learning, obtaining a certainty factor as a solution and generating a pair of the summary candidate data and its a certainty factor;
• and
• selecting the pair having a largest certainty factor and extracting the summary candidate data of the selected pair as the summary data of the inputted text data.

An automatic summarization processing method for generating summary data from digitized text data using a supervised machine learning method with solution data which consists of text data and its summary data as a problem and user's evaluation information as its solution, the method comprising:

• obtaining text data and summary data of the text data;
• displaying either the text data or the summary data or both of them;
• obtaining evaluation information which includes measure values chosen by a user from two or more predetermined levels in each qualities of the summarization and setting a set of measure values of the obtained evaluation information as a user's evaluation information;
• and
• generating solution data which uses the text data and the summary data as a problem and uses the evaluation information as a solution, and storing the solution data in a solution data storage unit;
• extracting the solution data from the solution data storage unit and obtaining features of the problem in the extracted solution data by semantically and syntactically analyzing said solution data;
• generating a combination with the features of the problem and one among the solution and solution candidates which are other set of measure value set with all the preset measure values excepting the set of measure values of the user's evaluation information;
• learning how probabilities of being a positive example or a negative example are produced at what set of features of the combination and storing a result of learning in a learning-result-data storage unit;
• generating summary candidate data from input text data;
• generating a combination with the input text data and one each of the summary candidate data;
• obtaining features of each the combination by semantically and syntactically analyzing said combination;
• estimating what probabilities of being a positive example or a negative example are produced with features of the combination based on the result of learning, obtaining a probability of being a positive example as a solution and generating a pair of the summary candidate data and its probability;
• and
• selecting the pair having a largest probability of being a positive example and extracting the summary candidate data of the selected pair as the summary data of the inputted text data.

An automatic summarization processing method for automatically summarizing a digitized text by a machine learning method using solution data which uses a text as a problem and uses the summary of the text as a solution, the method comprising:

• changing the solution data in accordance with a user's instruction, and accessing solution-data storage means which stores the changed solution data;
• extracting a pair of a feature set of the problem and a solution from the solution data, learning, at what feature, what solution is apt to be produced from the pair, and storing learning result data in learning-result-data storage means;
• and
• extracting a feature set from input text data and estimating what solution is apt to be produced from the feature set based on the learning result data.