Identifying topics in a digital work

• US 9,613,003 B1
• Filed: 03/28/2012
• Issued: 04/04/2017
• Est. Priority Date: 09/23/2011
• Status: Active Grant

• Current Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Original Assignee: Amazon Technologies, Inc. (Amazon.com, Inc.)
• Inventors: Goodspeed, Joshua M., Hamaker, Janna S., Iser, Adam J., Killalea, Tom, Patnia, Abhishek, Taborisskaya, Alla
• Primary Examiner(s): ADESANYA, OLUJIMI A
• Application Number:
  US13/433,028
• Time in Patent Office: 1,833 Days
• Field of Search: 704 9
• CPC Class Codes: G06F 16/3328 using graphical result space presentation or visualisation
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/487 using geographical or spatial information, e.g. location
• CPC Class Codes: G06F 16/9038 Presentation of query results
• CPC Class Codes: G06F 40/10 Text processing natural language analysis G06F40/20; semantic analysis G06F40/30; processing or translation of natural language G06F40/40
• CPC Class Codes: G06F 40/103 Formatting, i.e. changing of presentation of documents automatic justification G06F40/189; automatic line break hyphenation G06F40/191
• CPC Class Codes: G06F 40/169 Annotation, e.g. comment data or footnotes
• CPC Class Codes: G06F 40/221 Parsing markup language streams streaming G06F40/149
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising

Abstract

In some implementations, text is extracted from a digital work and a plurality of noun phrases are identified. The noun phrases are checked against a network accessible resource, such as an online encyclopedia, that includes a plurality of interlinked article entries. The noun phrases that have corresponding entries in the network accessible resource are included in a set of candidate topics. The candidate topics are ranked based, at least in part, on the links to and from each of the entries corresponding to the candidate topics. Candidate topics below a ranking threshold are removed from the set of candidate topics. Further, term frequency information for each candidate topic in relation to the digital work is compared against term frequency information for the candidate topic in a large corpus of textual works to remove candidate topics within a frequency difference threshold.

Claims

1. One or more non-transitory computer-readable media maintaining instructions executable by one or more processors to perform operations comprising:

extracting text from a digital work;

identifying a plurality of noun phrases from the text extracted from the digital work;

searching a network accessible resource having a plurality of entries to identify a set of one or more entries in the network accessible resource that contain information related to at least one noun phrase of the plurality of noun phrases, wherein each noun phrase corresponding to an entry in the set of one or more entries is a candidate topic in a set of candidate topics;

ranking the candidate topics based, at least in part, on at least one of a number of incoming links or a number of outgoing links between each of the entries corresponding to the candidate topics;

excluding, from the set of candidate topics, one or more candidate topics ranked below a first threshold;

comparing a first term frequency-inverse document frequency (tf-idf) value with a second tf-idf value, wherein the first tf-idf value is determined with respect to the digital work for each candidate topic remaining in the set of candidate topics, and wherein the second tf-idf value is determined for the candidate topics with respect to a corpus of works;

excluding, from the set of candidate topics, one or more candidate topics for which a difference between the first tf-idf value and the second tf-idf value is less than a second threshold;

generating a digital supplemental information file comprising at least one reference to supplemental information relating to at least one candidate topic remaining in the set of candidate topics;

receiving a request for the digital supplemental information file from an electronic device; and

transmitting the digital supplemental information file to the electronic device, the digital supplemental information file to cause the digital work to include at least one selectable portion that enables display of the at least one reference to supplemental information and a visual representation of at least a location in the digital work of each occurrence of the at least one candidate topic remaining in the set of candidate topics, wherein the visual representation comprises an object with markings corresponding to each occurrence.

2. The one or more computer-readable media as recited in claim 1, wherein the network accessible resource is at least one of:

an online wiki-type site;

an online encyclopedia site;

an online dictionary site;

a reference site;

or

a crowd-sourced information site.

3. The one or more computer-readable media as recited in claim 1, prior to ranking the candidate topics, the operations further comprising, for each entry in the set of one or more entries performing at least one of:

determining if one or more other entries in the set of one or more entries link to the entry; and

determining if the entry links to one or more other entries in the set of one or more entries.

4. The one or more computer-readable media as recited in claim 1, wherein the supplemental information comprises at least a location in the digital work of each occurrence of each candidate topic in the set of candidate topics.

5. The one or more computer-readable media as recited in claim 4, wherein the supplemental information further comprises content related to at least one candidate topic obtained from the set of one or more entries in the network accessible resource.

6. A method comprising:

under control of one or more processors configured with executable instructions,

searching a network accessible resource for at least one entry corresponding to at least one noun phrase obtained from a digital work;

identifying the at least one entry;

generating a set of candidate topics from the at least one noun phrase corresponding to the at least one entry identified;

for at least one candidate topic of the set of candidate topics;

comparing a first indication of a frequency of the at least one candidate topic in the digital work with a second indication of a frequency of the at least one candidate topic in a corpus of digital works, and

removing the at least one candidate topic from the set of candidate topics based, at least partly, on a difference between the first indication and the second indication being less than a threshold amount;

generating a digital supplemental information file comprising at least one reference to supplemental information relating to at least one candidate topic remaining in the set of candidate topics;

receiving a request for the digital supplemental information file from an electronic device; and

transmitting the digital supplemental information file to the electronic device, the digital supplemental information file to cause the digital work to include at least one selectable portion that enables display of the at least one reference to supplemental information and a visual representation of at least a location in the digital work of each occurrence of the at least one candidate topic remaining in the set of candidate topics wherein the visual representation comprises an object with markings corresponding to each occurrence.

7. The method as recited in claim 6, wherein the at least one entry in the network accessible resource contains information related to the at least one noun phrase.

8. The method as recited in claim 6, wherein the network accessible resource is at least one of:

an online wiki-type site;

an online encyclopedia site;

an online dictionary site;

a reference site;

or

a crowd-sourced information site.

9. The method as recited in claim 6, further comprising, for a plurality of noun phrases obtained from the digital work, identifying a plurality of entries in the network accessible resource, each entry corresponding to at least one noun phrase of the plurality of noun phrases.

10. The method as recited in claim 9, wherein the respective noun phrases are each a respective candidate topic in the set of candidate topics, the method further comprising, for a particular entry of the network accessible resource that corresponds to a particular candidate topic performing at least one of:

determining that one or more other entries corresponding to one or more other candidate topics link to the particular entry; and

determining that the particular entry links to one or more other entries that correspond to one or more other candidate topics; and

ranking the candidate topics in the set of candidate topics based, at least in part, on the links to and from the entries corresponding to the candidate topics.

11. The method as recited in claim 10, wherein ranking the candidate topics is based, at least in part, on a page ranking of each of the entries corresponding to the candidate topics, wherein the page ranking is based, at least in part, on a probability of being directed to a particular entry of the entries corresponding to the candidate topics by selecting a random link in another entry corresponding to the candidate topics.

12. The method as recited in claim 10, wherein ranking the candidate topics is based, at least in part, on at least one of:

a number of incoming links or a number of outgoing links between each of the entries corresponding to the candidate topics.

13. The method as recited in claim 10, further comprising, applying a ranking threshold to remove, from the set of candidate topics, one or more candidate topics ranked below the ranking threshold.

14. The method as recited in claim 13, further comprising retaining or rejoining, in the set of candidate topics, regardless of the ranking threshold, a candidate topic that corresponds to at least one of a place or an organization.

15. The method as recited in claim 9, wherein the respective noun phrases are each a respective candidate topic in the set of candidate topics, wherein:

the first indication comprises a first term frequency-inverse document frequency (tf-idf) value and the second indication comprises a second tf-idf value, wherein the first tf-idf value is associated with each respective candidate topic with respect to the digital work, and the second tf-idf value is associated with each respective candidate topic with respect to the corpus of digital works; and

the removing is based, at least partly, on a difference between the first tf-idf value and the second tf-idf value being less than the threshold amount.

16. The method as recited in claim 15, further comprising retaining or rejoining, in the set of candidate topics, regardless of the threshold, a candidate topic that corresponds to at least one of a place or an organization.

17. The method as recited in claim 6, wherein the at least one noun phrase has noun phrase location information associated therewith, the noun phrase location information identifying the location of each occurrence of the at least one noun phrase in the digital work.

18. The method as recited in claim 17, further comprising:

parsing text of the digital work into a plurality of tokens, each token corresponding to at least one of a word or a punctuation mark;

determining token location information for the tokens; and

associating the noun phrase location information with each occurrence of the noun phrase in the digital work based, at least in part, on the token location information.

19. A system comprising:

one or more processors;

one or more computer-readable media; and

one or more modules maintained on the one or more computer-readable media to be executed by the one or more processors to perform operations including;

obtaining a plurality of noun phrases from a digital work;

searching a network accessible resource having a plurality of entries to identify a set of one or more entries that correspond to one or more noun phrases of the plurality of noun phrases;

generating a set of candidate topics from the one or more noun phrases;

removing, from the set of candidate topics, at least one candidate topic based, at least partly, on a difference between a first indication of a frequency of the at least one candidate topic in the digital work and a second indication of a frequency of the at least one candidate topic in a collection of digital works being within a threshold;

generating a digital supplemental information file comprising at least one reference to supplemental information relating to at least one candidate topic remaining in the set of candidate topics;

receiving a request for the digital supplemental information file from an electronic device; and

transmitting the digital supplemental information file to the electronic device, the digital supplemental information file to cause the digital work to include at least one selectable portion that enables display of the at least one reference to supplemental information and a visual representation of at least a location in the digital work of each occurrence of the at least one candidate topic remaining in the set of candidate topics wherein the visual representation comprises an object with markings corresponding to each occurrence.

20. The system as recited in claim 19, the operations further comprising:

determining entries in the set of one or more entries that link to other entries in the set of one or more entries;

creating the set of candidate topics from the one or more noun phrases that correspond to the entries in the set of one or more entries;

ranking the candidate topics based, at least in part, on links between the entries in the set of one or more entries; and

removing from the set of candidate topics one or more candidate topics that rank below a ranking threshold based, at least in part, on the links between the entries in the set of one or more entries.

21. The system as recited in claim 19, the operations further comprising:

creating the set of candidate topics from the plurality of noun phrases that correspond to entries in the set of one or more entries;

parsing content of the entries in the set of one or more entries to identify entries that include, in the content, noun phrases corresponding to other entries in the set of one or more entries;

ranking the candidate topics based on, at least in part, for each entry in the set of one or more entries, a frequency in the content of each entry of noun phrases corresponding to other candidate topics; and

removing from the set of candidate topics one or more candidate topics that rank below a ranking threshold based, at least in part, on the frequency of noun phrases corresponding to other candidate topics.

22. The system as recited in claim 19, wherein the first indication and the second indication comprise term frequency-inverse document frequencies (tf-idfs).

23. The system as recited in claim 19, the operations further comprising, prior to the searching, for each noun phrase containing two or more words, deriving additional smaller noun phrases by employing a backwards walk through the noun phrase.

24. The system as recited in claim 19, wherein the network accessible resource is at least one of:

an online wiki-type site;

an online encyclopedia site;

an online dictionary site;

a reference site;

or

a crowd-sourced information site.

25. The system as recited in claim 19, wherein the supplemental information further comprises information related to the at least one noun phrase obtained from at least one entry in the set of one or more entries.