Low-dimensional information discovery and presentation system, apparatus and method
A search system configured to predict further search intents of a user and to perform exploratory further searches and produce a number of search features and associated relevances and divergence quantifiers for displaying by user equipment at least two-dimensional so as to allow the user to identify relationship of various diverging search terms and to rapidly direct the searching towards information the existence of which may have been previously unknown to the user. Some of the search features can be concealed and shown to the user only if the user magnifies the corresponding area on a display showing the search features returned by the search engine. Files matching to varying degree with the present and predicted further searches are shown to the user with respective lists of search features.
1. A method in a search system, comprising:
receiving a current search request from a user for new or updated search of information, the current search request comprising user input that comprises at least one of: a search query feature; an indication of interest in particular search features; and a request to resume to any preceding search state; a bookmark identifying a file; and a bookmark identifying an earlier search feature; and
in response to the receiving of the current search request automatically:
forming an intent model of a current search intent of the user based on: the user input of the current search request; any earlier received search request from the user that was refined by the current search request; and a database of searchable files;
wherein the intent model comprises or identifies at least one search feature;
wherein the intent model further comprises a measure of the user's interest in the at least one of the search features;
wherein the intent model further comprises a measure of uncertainty of the user's interest in the at least one of the search features;
identifying and ranking a plurality of potential further search features according to the intent model;
determining for each of the search features ability of the search feature in question to change the intent model based on the database of searchable files;
establishing a summarized representation of the intent model and search features predicted as potentially relevant by the model and a summarized representation of the determined ability of the predicted potentially relevant search features to change the intent model;
identifying and ranking a plurality of files according to their estimated relevance according to the intent model and the uncertainty of the estimated relevance; and
sending to the user an output set comprising a representation of the intent model and at least one of: a representation of the search features predicted as potentially relevant by the model; a representation of the predicted relevance of the one or more search features; a representation of predicted relative relevance of the one or more search features; a representation of ability of the predicted potentially relevant search features to change the intent model; an identification of one or more of the identified files; a representation of relevance of the one or more identified files; a representation of relative relevance of the one or more identified files; a representation of the ranking of the one or more identified files.
2. The method of claim 1 , further comprising ranking documents based on a probabilistic multinomial language model.
3. The method of claim 1 , wherein the measure of uncertainty of the user's interest in the at least one feature is determined by predicting relevance of a search feature across further searches.
4. The method of claim 1 , wherein the forming the intent model comprises estimating a search feature weight vector to map search feature features to relevance scores that are produced in the estimating of the relevances of the search features.
5. The method of claim 1 , wherein the summarized representation of the ability of the predicted potentially relevant search features to change the intent model comprises a divergence indication for each of the search features of the output set.
6. The method of claim 1 , further comprising establishing the at least one search feature by the search system if not contained or indicated by the current search request.
7. An apparatus, comprising:
an input configured to receive a current search request from a user for new or updated search of information, the current search request comprising user input that comprises at least one of: a search query feature; an indication of interest in particular search features; and a request to resume to any preceding search state; a bookmark identifying a file; and a bookmark identifying an earlier search feature;
an output configured to send information;
a memory configured to store computer executable program code;
at least one processor configured to execute the program code and accordingly receive the current search request from the input and in response to the receiving of the current search request to automatically:
form an intent model of a current search intent of the user based on: the user input of the current search request; any earlier received search request from the user that was refined by the current search request; and a database of searchable files;
wherein the intent model comprises or identifies at least one search feature;
wherein the intent model further comprises a measure of the user's interest in the at least one of the search features;
wherein the intent model further comprises a measure of uncertainty of the user's interest in the at least one of the search features;
identify and rank a plurality of potential further search features according to the intent model;
determine for each of the search features ability of the search feature in question to change the intent model based on the database of searchable files;
establish a summarized representation of the intent model and search features predicted as potentially relevant by the model and a summarized representation of the determined ability of the predicted potentially relevant search features to change the intent model;
identify and rank a plurality of files according to their estimated relevance according to the intent model and the uncertainty of the estimated relevance; and
send, by controlling the output according to the program code, to the user an output set comprising a representation of the intent model and at least one of: a representation of the search features predicted as potentially relevant by the model; a representation of the predicted relevance of the one or more search features; a representation of predicted relative relevance of the one or more search features; a representation of ability of the predicted potentially relevant search features to change the intent model; an identification of one or more of the identified files; a representation of relevance of the one or more identified files; a representation of relative relevance of the one or more identified files; a representation of the ranking of the one or more identified files.