Abstract: A prediction data input unit 91 inputs prediction data that is one or more explanatory variables that are information likely to affect future sales. An exposure pattern generation unit 92 generates an exposure pattern which is an explanatory variable indicating the content of a commercial message scheduled to be performed during a period from predicted time to future prediction target time. A component determination unit 93 determines the component used for predicting the sales, on the basis of a hierarchical latent structure that is a structure in which latent variables are represented by a tree structure and components representing probability models are located at nodes of a lowest level of the tree structure, gating functions for determining a branch direction in the nodes of the hierarchical latent structure, and the prediction data and the exposure pattern.

Abstract: A search system includes a reception unit 601 for receiving specification of a specific sentence in an article, a search unit 602 using a first sentence being the specified sentence as a search query to search an article including a sentence having a meaning similar to the connotations of the first sentence, and a display unit 603 for displaying information of the searched article as a related article to the first sentence, and when there are a plurality of related articles, the display unit 603 preferentially displays the related articles in descending order of difference relative to a first article being the article including the first sentence.

Abstract: The prediction function creation device according to the present invention for creating a prediction function to derive an objective variable by using a set of samples that include explanatory variables and an objective variable, the device includes: a clustering unit that clusters the respective samples by giving labels, and assigns weights to each label in accordance with patterns of missing values for the explanatory variables in labeled samples; a child model creation unit that makes portions of the training data partial training data on the basis of the weights, and determines an explanatory variable that constitutes the prediction function on the basis of patterns of missing values for the explanatory variables in the samples; and a mixture model creation unit that creates the prediction function with respect to each pattern of missing values by using the explanatory variable and the determined partial training data.

Abstract: A method for classifying a new instance including a text document by using training instances with class including labeled data and zero or more training instances with class including unlabeled data, comprising: estimating a word distribution for each class by using the labeled data and the unlabeled data; estimating a background distribution and a degree of interpolation between the background distribution and the word distribution by using the labeled data and the unlabeled data; calculating two probabilities for that the word generated from the word distribution and the word generated from the background distribution; combining the two probabilities by using the interpolation; combining the resulting probabilities of all words to estimate a document probability for the class that indicates the document is generated from the class; and classifying the new instance as a class for which the document probability is the highest.

Abstract: A search system includes a reception unit 601 for receiving specification of a specific sentence in an article, a search unit 602 using a first sentence being the specified sentence as a search query to search an article including a sentence having a meaning similar to the connotations of the first sentence, and a display unit 603 for displaying information of the searched article as a related article to the first sentence, and when there are a plurality of related articles, the display unit 603 preferentially displays the related articles in descending order of difference relative to a first article being the article including the first sentence.

Abstract: The information processing device 1 processes document collections having tags permitting semantic class identification appended to each document and comprises a search unit 2, which creates multiple semantic class units containing one, two, or more semantic classes based on a taxonomy that identifies relationships between semantic classes, and a frequency calculation unit 3 which, for each of the semantic class units, identifies documents that match that semantic class unit in the document collections and, for these matching documents, calculates a first frequency that represents the frequency of occurrence in a designated document collection and a second frequency that represents the frequency of occurrence in non-designated document collections. Once the calculations have been performed, the search unit 2 identifies any of the semantic class units based on the first frequency and the second frequency of the matching documents.

Abstract: A similar sentence set generation unit 81 groups sentences representing a same concept or event from a set of analysis target sentences, to generate a similar sentence set. A similar sentence set extraction unit 82 extracts, using one or more specific sentence extractors each capable of extracting a specific sentence belonging to a specific classification from the set of analysis target sentences, one or more sentences not extracted by any of the specific sentence extractors from among the sentences belonging to the similar sentence set, as an exclusion similar sentence set.

Abstract: A prediction data input unit 91 inputs prediction data that is one or more explanatory variables that are information likely to affect future sales. An exposure pattern generation unit 92 generates an exposure pattern which is an explanatory variable indicating the content of a commercial message scheduled to be performed during a period from predicted time to future prediction target time. A component determination unit 93 determines the component used for predicting the sales, on the basis of a hierarchical latent structure that is a structure in which latent variables are represented by a tree structure and components representing probability models are located at nodes of a lowest level of the tree structure, gating functions for determining a branch direction in the nodes of the hierarchical latent structure, and the prediction data and the exposure pattern.

Abstract: A document classification method includes a first step for calculating smoothing weights for each word and a fixed class, a second step for calculating smoothed second-order word probability, and a third step for classifying document including calculating the probability that the document belongs to the fixed class.

Abstract: Provided is a text processing system which, when an attribute corresponding to one tabulation axis is set, is capable of generating a text group which will produce non-obvious tabulation results when cross-tabulation is performed using that attribute. At the time of input of respective attribute values of an attribute which corresponds to a tabulation axis in cross tabulation and a document associated with any one of the attribute values of the attribute, text extraction means 71 extracts portion not including the attribute value of the attribute from each text obtained by dividing the document into predetermined units. Group generation means 72 performs entailment recognition between texts on the extracted texts and groups texts having an entailment relation.

Abstract: A method for classifying a new instance including a text document by using training instances with class including labeled data and zero or more training instances with class including unlabeled data, comprising: estimating a word distribution for each class by using the labeled data and the unlabeled data; estimating a background distribution and a degree of interpolation between the background distribution and the word distribution by using the labeled data and the unlabeled data; calculating two probabilities for that the word generated from the word distribution and the word generated from the background distribution; combining the two probabilities by using the interpolation; combining the resulting probabilities of all words to estimate a document probability for the class that indicates the document is generated from the class; and classifying the new instance as a class for which the document probability is the highest.

Abstract: A boundary word identification unit (103) identifies a boundary word belonging to a plurality of categories among words gathered in dictionary growth processing. Then, a category membership degree calculation unit (104) calculates, for each category to which the boundary word belongs, a category membership degree indicating a degree to which the boundary word belongs to the category on the basis of information recorded in a gathering process memory unit (108). Next, a category update unit (105) determines the category to which the boundary word belongs on the basis of the category membership degree calculated by the category membership degree calculation unit (104) and updates information stored in a gathered-by-category word memory unit (109) so that the determination result is reflected.

Abstract: In an inverted list of each node in a taxonomy, among each node, an inverted list of the highest node is a list of integer values indicating an identifier of search subject data, and an inverted list of a node other than the highest node, in place of the identifier, is a list of integer values indicating a position in an inverted list corresponding to a node that is higher by one than the node. Furthermore, a list of integer values in an inverted list of each node is divided into two or more blocks, and a differential value between an integer value and an integer value directly before the integer value in the block is converted into a bit string of a variable length integer code.

Abstract: Provided are a word latent topic estimation device and a word latent topic estimation method which are capable of hierarchically performing processing and which are capable of rapidly estimating latent topics of a word while taking into consideration a mixed state of topics.

Abstract: A data classifier system of the present invention selects a plurality of classifications correlated to data groups so as to output classification axes based on hierarchical classifications and data groups. The data classifier system includes a basic category accumulation means, a classification axis candidate creation means and a priority calculation means. The basic category accumulation means accumulates classifications serving as basic categories used for selecting desired classifications in advance. The classification axis candidate creation means creates classification axis candidates based on combinations of classifications each correlated to at least one data among descendant classifications of each basic category. The priority calculation means calculates priorities with respect to the classification axis candidates created by the classification axis candidate creation means based on hierarchical distances of classifications in the classified hierarchy.

Abstract: A data classifier system of the present invention selects a plurality of classifications correlated to data groups so as to output classification axes based on hierarchical classifications and data groups. The data classifier system includes a basic category accumulation means, a classification axis candidate reduction means and a priority calculation means. The basic category accumulation means accumulates classifications serving as basic categories used for desired classifications in advance. The classification axis candidate reduction means selects a plurality of classifications from among classifications descendant from each basic category so as to create classification axis candidates, thus reducing classification axis candidates subjected to calculations based on data quantity of classifications and hierarchical distances of classifications.

Abstract: The prediction function creation device according to the present invention for creating a prediction function to derive an objective variable by using a set of samples that include explanatory variables and an objective variable, the device includes: a clustering unit that clusters the respective samples by giving labels, and assigns weights to each label in accordance with patterns of missing values for the explanatory variables in labeled samples; a child model creation unit that makes portions of the training data partial training data on the basis of the weights, and determines an explanatory variable that constitutes the prediction function on the basis of patterns of missing values for the explanatory variables in the samples; and a mixture model creation unit that creates the prediction function with respect to each pattern of missing values by using the explanatory variable and the determined partial training data.

Abstract: In order to accurately learn a function for evaluating documents, even in the case where sample documents having missing feature values are included as training data, a document evaluation apparatus is provided with a data classification unit (3) that classifies a set of sample documents based on missing patterns of a first feature vector, a first learning unit (4) that uses feature values that are not missing in the first feature vector and evaluation values to learn a first function for calculating a first score which is a weighted evaluation value for each classification, a feature vector generation unit (5) that computes a feature value corresponding to each classification using the first score, and generates a second feature vector having the computed feature values, and a second learning unit (6) that uses the second feature vector and the evaluation values to learn a second function for calculating a second score for evaluating documents targeted for evaluation.

Abstract: A communication assistance device (10) includes a communication level determination unit (11) so as to determine a level of a relationship between users who communicate with each other. The communication level determination unit (11) determines the level (communication level) of the relationship between the users based on similarity between the users obtained from preference information showing preferences of the users, and on user action records showing records of actions taken by a certain user toward a partner user with whom the certain user communicates out of the users.

Abstract: The disclosed apparatus uses a training data generation apparatus 2, which generates training data used for creating characteristic expression extraction rules. The training data generation apparatus 2 includes: a training data candidate clustering unit 21, which clusters a plurality of training data candidates assigned labels indicating annotation classes based on feature values containing respective context information, and a training data generation unit 22 which, by referring to each cluster obtained using the clustering results, obtains the distribution of the labels of the training data candidates within the cluster, identifies training data candidates that meet a preset condition based on the obtained distribution, and generates training data using the identified training data candidates.