Context-based automatic selection of factor for use in estimating characteristics of viewers viewing same content

An apparatus for selecting an optimum factor for use in estimating characteristics of a plurality of viewers who view a same content, the apparatus comprising:

• a factor-list storage device for storing therein a multi-layered hierarchal factor-list of a plurality of pre-selected factors each of which is a candidate of the optimum factor, the factors being a plurality of attribute words which express attributes of the viewers, the factor list having a plurality of layers;
• a learned-text storage device for storing therein a plurality of sets of learned-text information which are relevant to the plurality of attribute words, in sections, on a per-attribute-word basis;
• an extractor configured to extract a plurality of keywords from the same content;
• a first-vector generator configured to generate a first vector indicative of frequencies with which the plurality of keywords occur in the same content, the first vector having a plurality of elements which the plurality of keywords are allocated, respectively, the elements having values indicative of the frequencies of the plurality of keywords, respectively;
• a second-vector generator configured to operate for a successively-selected one of the keywords, to successively generate a second vector indicative of frequencies with which the plurality of keywords occur in ones of the plurality of sets of learned-text information which ones contain an attribute word identical with the successively-selected keyword, the second vector having a plurality of elements which the plurality of keywords are allocated, respectively, the elements having values indicative of the frequencies of the plurality of keywords, respectively;
• a similarity calculator configured to calculate a similarity score between the first and second vectors on a per-keyword basis;
• and
• an optimum-factor selector configured to select, on a per-layer basis, one of the plurality of factors as the optimum factor, wherein the one factor is associated with a sub-plurality of attribute words, which are on a lower layer than that of the one factor, and which are coincident with at least some of the plurality of keywords, so that a dispersion level indicative of how the similarity scores calculated for the sub-plurality of attribute words are dispersed therewithin can be substantially the maximum among the plurality of factors on the same layer, and selecting the sub-plurality of attribute words as optimum attribute words for the optimum factor.

A method of selecting a factor for estimating characteristics of viewers who view a same content, the method comprising:

• storing a multi-layered hierarchal list of pre-selected attribute words expressing viewer attributes, each of which is a candidate of the factor;
• extracting the attribute words from the same content;
• generating a first vector indicative of frequencies with which the attribute words occur in the same content;
• for a successively-selected one of the attribute words, successively generating a second vector indicative of frequencies with which the attribute words occur in ones of sets of learned-text information which contain the successively-selected attribute word;
• calculating a similarity score between the first and second vectors on a per-attribute-word basis;
• and
• selecting one of the attribute words as the factor, which is hierarchically associated with subordinate attribute words on a lower layer than that of the one attribute word, based on a dispersion level indicative of how the similarity scores calculated for the subordinate attribute words are dispersed within the subordinate attribute words.