Ranking and accessing definitions of terms

1. A computer-implemented method of processing information, comprising:

developing a ranking model using training data including a plurality of training definition candidates for a plurality of terms, each training definition candidate having a label that classifies the training definition candidate, wherein the labels include at least two of good, bad and indifferent labels;

collecting text strings of definition candidates from a data source;

identifying features in the text strings of definition candidates;

ranking, using a processor of the computer, the definition candidates for a selected term based on the text strings, wherein ranking the definition candidates for the selected term comprises utilizing the ranking model to generate a definition tuple for each of the associated definition candidates for the selected term, the definition tuple for each of the associated definition candidates comprising:

the selected term, the associated definition candidate, and a score assigned to the associated definition candidate by the ranking model, wherein the score is calculated based on the identified features for the associated definition candidate independent of the selected term, and wherein the score is indicative of a measure of a degree to which the associated definition candidate conveys a meaning of the selected term;

storing the definition tuple for each of the definition candidates for the selected term in a database;

comparing similar definition candidates for the selected term and removing the definition tuples for the similar definition candidates from the database, based on the scores assigned to the similar definition candidates, if the similar definition candidates are substantially duplicates, wherein comparing similar definition candidates comprises comparing a first definition candidate for the selected term and a second definition candidate for the selected term, and wherein removing comprises removing one of the first and second definition candidate from the database having a lower assigned score;

receiving an input query having the selected term; and

accessing the definition tuples for the selected term to present a ranked list of definition candidates for the selected term, wherein the ranked list includes the associated definition candidates obtained from the definition tuples and is ranked in an order based on the scores obtained from the definition tuples, wherein the definition candidates in the ranked list are ranked based on the scores associated with each definition tuple and are ranked independent of the select term and the input query.

2. The method of claim 1 and further comprising identifying terms in the data source having an associated definition candidate.

3. The method of claim 1 and further comprising parsing sentences in the data source.

4. The method of claim 1 wherein collecting definition candidates includes matching patterns in the data source.

5. A computer-implemented method of accessing definition candidates, comprising:

accessing a data source comprising a collection of documents;

identifying terms in the data source having associated definition candidates;

for a term identified in the data source,

collecting text strings of the associated definition candidates for the term from the data source;

identifying features in the collected text strings of the associated definition candidates for the term;

ranking the associated definition candidates for the term based on the text strings, wherein ranking comprises utilizing a ranking model to generate a definition tuple for each of the associated definition candidates for the term, the definition tuple for each of the associated definition candidates comprising:

the term, the associated definition candidate, and a score assigned to the associated definition candidate by the ranking model, wherein the score is calculated based on the identified features for the associated definition candidate independent of the term, and wherein the score is indicative of a measure of a degree to which the associated definition candidate conveys a meaning of the term; and

storing the definition tuple for each of the definition candidates for the term in a database;

receiving an input query having the term; and

accessing the definition tuples for the term to present a ranked list of definition candidates for the term using a processor of the computer, wherein the ranked list includes the associated definition candidates obtained from the definition tuples and is ranked in an order based on the scores obtained from the definition tuples, wherein the definition candidates in the ranked list are ranked based on the scores associated with each definition tuple and are ranked independent of the term and the input query.

6. The method of claim 5 , wherein the score associated with each definition tuple is indicative of a level of goodness of the associated definition candidate associated with the definition tuple.

7. The method of claim 6 wherein the score for a definition candidate is low if text strings are an opinion of the term.

8. The method of claim 6 wherein the score for a definition candidate is high if the text strings are a relevant definition of the term.

9. The method of claim 1 , wherein the scores assigned to the definition candidates are indicative of a level of goodness of the definition candidates.

10. The method of claim 5 , and further comprising:

removing similar definition candidates for the term from the database, based on the scores assigned to the similar definition candidates in the definition tuples, if the similar definition candidates are substantially duplicates by comparing a first definition candidate for the term and a second definition candidate for the term and removing one of the first and second definition candidates from the database having a lower assigned score.