# US 2014/0207777 A1 · Computer implemented methods and apparatus for identifying similar labels using collaborative filtering

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2014/0207777 A1
- Document ID: US20140207777A1
- Status: Published application
- Kind: A1
- Application number: 14/161,232
- Filed: 2014-01-22
- Published: 2014-07-24
- Earliest claimed: 2013-01-22
- Assignees: salesforce.com, inc.
- Applicants: salesforce.com, inc.
- Inventors: Joel Palmert; Tony Meng
- Web page: https://patentgrep.com/patent/US20140207777A1

## CPC

- G06F 16/22: Indexing; Data structures therefor; Storage structures
- G06F 16/90335: {Query processing}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

Disclosed are methods, apparatus, systems, and computer-readable storage media for identifying similar labels. In some implementations, one or more servers maintain a plurality of data entries in one or more database tables storing textual data, each data entry of a first portion of the data entries including: a text sequence, a label, and a text-to-label association score, and each data entry of a second portion of the data entries including: a first label, a second label, and a similarity score. The one or more servers analyze the data of the first portion of data entries to generate one or more pairs, each pair including information identifying a first label and a second label. The one or more servers calculate a similarity score for each of the one or more pairs and store the respective similarity scores in the second portion of the data entries.

## Claims

### Claim 1

A computer implemented method for identifying similar labels, the method comprising:
maintaining, by one or more servers, a plurality of data entries in one or more database tables storing textual data, each data entry of a first portion of the data entries including: a text sequence, a label, and a text-to-label association score indicating a number of times that the text sequence appears in a previous incoming text associated with the label, and each data entry of a second portion of the data entries including: a first label, a second label, and a similarity score;
analyzing the data of the first portion of data entries to generate one or more pairs, each pair including information identifying a first label and a second label;
calculating a similarity score for each of the one or more pairs; and
storing each of the one or more pairs and the respective similarity scores in the second portion of the data entries.

### Claim 2 (depends on 1)

The method of claim 1, wherein calculating the similarity score for a pair comprises calculating a cosine similarity score for the first and second labels of the pair using a first vector of text sequences associated with the first label and a second vector of text sequences associated with the second label, wherein a text sequence is associated with a label if it appears in a previous incoming text associated with the label.

### Claim 3 (depends on 1)

The method of claim 1, the method further comprising:
receiving, from a computing device, a request for labels having similar associated text sequences;
identifying, based on the stored one or more pairs and the stored similarity scores, a set of pairs having the same first label; and
selecting a pair of the identified set of pairs, wherein the respective similarity score of the selected pair is higher than that of the other pairs of the identified set of pairs.

### Claim 4 (depends on 3)

The method of claim 3, the method further comprising:
determining that the similarity score of the selected pair is higher than a designated similarity threshold; and
transmitting the first label and the second label of the selected pair to the computing device as a similar pair of labels.

### Claim 5 (depends on 4)

The method of claim 4, wherein the selected pair of similar labels is transmitted to the computing device with one or more other pairs of similar labels.

### Claim 6 (depends on 3)

The method of claim 3, wherein the similar pair of labels is selected in response to receiving from the computing device a request for similar pairs of labels.

### Claim 7 (depends on 1)

The method of claim 1, the method further comprising:
responsive to a request to associate a label with an incoming text, updating the first portion of data entries with one or more text sequences of the incoming text and the requested label.

### Claim 8 (depends on 7)

The method of claim 7, wherein updating the one or more database tables comprises:
for each text sequence of the incoming text:
identifying or creating a first data entry of the first portion of data entries that includes the text sequence and the requested label; and
incrementing the text-to-label association score of the identified or created first data entry by an inflation factor.

### Claim 9 (depends on 8)

The method of claim 8, wherein the inflation factor has a value based on a measure of time.

### Claim 10 (depends on 1)

The method of claim 1, the method further comprising:
periodically identifying one or more incoming texts and associated labels, wherein each incoming text includes one or more text sequences; and
for each incoming text and associated label, updating the first portion of data entries with the one or more text sequences of the incoming text and the associated label.

### Claim 11 (depends on 10)

The method of claim 10, wherein the incoming texts are text documents, wherein a label is a file name of a text document, and wherein the similarity scores identify text documents that are different versions of the same original text document.

### Claim 12 (depends on 1)

The method of claim 1, wherein an incoming text is one of: a social media message, a text document, a customer relationship management (CRM) object including textual data, a feed item, and an article.

### Claim 13 (depends on 1)

The method of claim 1, wherein a text sequence includes one or more words.

### Claim 14 (depends on 1)

The method of claim 1, wherein a label is a user-generated topic assigned to a social media message or a CRM object having textual data.

### Claim 15 (depends on 1)

The method of claim 1, wherein a label is a hashtag.

### Claim 16 (depends on 1)

The method of claim 1, wherein the similarity score of a pair including a first label and a second label is based at least in part on a frequency of one or more text sequences appearing in a first one or more previous incoming texts associated the first label and in a second one or more previous incoming texts associated with the second label.

### Claim 17 (depends on 1)

The method of claim 1, wherein the similarity score of a pair is normalized for a frequency at which the text sequences associated with the labels are used in incoming texts.

### Claim 18 (depends on 1)

The method of claim 1, wherein calculating the similarity score for each of the one or more pairs comprises using a collaborative filtering method to determine cosine similarity scores for the pairs, and wherein the second portion of the data entries is a collaborative filter table having the one or more pairs recorded therein based on the analysis.

### Claim 19

One or more computing devices for identifying similar labels to a user, the one or more computing devices comprising:
one or more processors operable to execute one or more instructions to:
maintain, by one or more servers, a plurality of data entries in one or more database tables storing textual data, each data entry of a first portion of the data entries including: a text sequence, a label, and a text-to-label association score indicating a number of times that the text sequence appears in a previous incoming text associated with the label, and each data entry of a second portion of the data entries including: a first label, a second label, and a similarity score;
analyze the data of the first portion of data entries to generate one or more pairs, each pair including information identifying a first label and a second label;
calculate a similarity score for each of the one or more pairs; and
store each of the one or more pairs and the respective similarity scores in the second portion of the data entries.

### Claim 20

A non-transitory computer-readable storage medium storing instructions executable by a computing device to cause a method to be performed for identifying similar labels to a user, the method comprising:
maintaining, by one or more servers, a plurality of data entries in one or more database tables storing textual data, each data entry of a first portion of the data entries including: a text sequence, a label, and a text-to-label association score indicating a number of times that the text sequence appears in a previous incoming text associated with the label, and each data entry of a second portion of the data entries including: a first label, a second label, and a similarity score;
analyzing the data of the first portion of data entries to generate one or more pairs, each pair including information identifying a first label and a second label;
calculating a similarity score for each of the one or more pairs; and
storing each of the one or more pairs and the respective similarity scores in the second portion of the data entries.
