Synonym determination among n-grams

1. A non-transitory, computer-readable storage medium containing code that, when executed by a processor, causes the processor to:

obtain a plurality of n-grams from a plurality of messages;

determine a temporal histogram for each n-gram;

determine synonyms among the n-grams based on a correlation of he histograms and a distance measure between n-grams; and

select from among the synonyms an n-gram for presentation, wherein the selection is based of a difficulty metric.

2. The non-transitory, computer-readable storage medium of claim 1 wherein, when executed, the code causes the processor to compute a similarity measure between n-grams.

3. The non-transitory, computer-readable storage medium of claim 2 wherein, when executed, the code causes the processor to compute the similarity measure by computing a similarity measure based on the correlation of the histograms and the distance measure.

4. The non-transitory, computer-readable storage medium of claim 2 wherein, when executed, the code causes the processor to compute the similarity measure by computing a weighted sum of the correlation of the histograms, the distance measure, and a co-occurrence value.

5. The non-transitory, computer-readable storage medium of claim 1 wherein, when executed, the code causes the processor to compute the distance measure such that the distance measure has a lower value for commonly-used tag creation operations.

6. The non-transitory, computer-readable storage medium of claim 1 wherein, when executed, the code causes the processor to determine synonyms also based on a frequency of occurrence n-grams in a same message.

7. The non-transitory, computer-readable storage medium of claim 1 wherein, when executed, the code causes the processor to determine synonyms by:

computing a correlation between the histograms;

determining a set S 1 of n-grams whose correlations exceed a threshold;

thresholding the set S 1 based on frequency of occurrence of n-grams in the same message to generate a set S 2 of threshoided n-grams;

computing distance measures between the n-grams;

computing similarity measures as weighted sums of the correlations and the distance measures;

thresholding the set S 2 based on the similarity measures to produce a set S 3 ; and

the code, when executed, further causes the processor to compute a difficulty metric of each n-gram in set S 3 and select a synonym from the set S 3 based on the difficulty metrics.

8. A method, comprising:

obtaining, by a processor, a plurality of n-grams from a plurality of messages;

determining, by the processor, a temporal histogram for each n-gram;

determining, by the processor, synonyms among the n-grams based on a combination of a correlation of the histograms and a distance measure between n-grams; and

selecting, by the processor, an n-gram from the synonyms, wherein the selection is based on a difficulty metric computed for each of a plurality of the n-grams.

9. The method of claim 8 further comprising computing the combination of the correlation of the histograms and the distance measure by computing a weighted sum of the correlation of the histograms and the distance measure.

10. The method of claim 8 further comprising generating the plurality of messages from which the n-grams are obtained by performing a search of messages based on a tag, extracting commonly-occurring concepts from the message set, and searching for other messages containing any of the extracted commonly-occurring concepts.

11. The method of claim 8 wherein determining synonyms among the n-grams comprises determining synonyms also based on a frequency of occurrence of n-grams in a same message.

12. The method of claim 8 wherein determining synonyms among the n-grams comprises:

computing a correlation between the histograms;

determining a set S 1 of n-grams whose correlations exceed a threshold;

thresholding the set S 1 based on frequency of occurrence of n-grams in the same message to generate a set S 2 of thresholded n-grams;

computing similarity measures as weighted sums of the correlations and the distance measures; and

thresholding the set S 2 based on the similarity measures to produce a set S 3 ; and

the method further comprises computing a difficulty metric of each n-gram in set S 3 and selecting a synonym from the set S 3 based on the difficulty metrics.

13. A system, comprising:

a temporal histogram engine to determine, via a processor, a temporal histogram for each of a plurality of n-grams from a plurality of messages;

a correlation engine to compute, via the processor, correlations of the temporal histograms;

a distance measurement engine to determine, by the processor, character-based distances between n-grams;

a synonym determination engine to determine, via the processor, a synonym among the n-grams based on the correlations of the histograms and the character-based distances; and

a difficulty metric engine to compute, via the processor, difficulty metrics for at least some of the n-grams.

14. The system of claim 13 further comprising a same message occurrence engine to determine, via the processor, a frequency of occurrence of correlated n-grams in the same message.

15. The system of claim 13 further comprising a similarity measurement engine to compute, via the processor, a similarity between n-grams based on correlation of histograms and distance measurements from the distance measurement engine.

16. The system of claim 15 wherein the similarity measurement engine is to compute the similarity as a weighted sum of the correlation of histograms and the distance measurements.

17. The system of claim 13 further comprising a synonym selection engine to select, via the processor, a synonym based on the difficulty metrics.