Character-based attribute value extraction system

Patent metadata

Field	Value
Publication number	US-2016321358-A1
Application number	US-201514700683-A
Country	US
Kind code	A1
Filing date	Apr 30, 2015
Priority date	Apr 30, 2015
Publication date	Nov 3, 2016
Grant date	—

Abstract

A system is provided that extracts attribute values. The system receives data including unstructured text from a data store. The system further tokenizes the unstructured text into tokens, where a token is a character of the unstructured text. The system further annotates the tokens with attribute labels, where an attribute label for a token is determined, in least in part, based on a word that the token originates from within the unstructured text. The system further groups the tokens into text segments based on the attribute labels, where a set of tokens that are annotated with an identical attribute label are grouped into a text segment, and where the text segments define attribute values. The system further stores the attribute labels and the attribute values within the data store.

First claim

We claim:

1 . A computer-readable medium having instructions stored thereon that, when executed by a processor, cause the processor to extract attribute values, the extracting comprising:
receiving data comprising unstructured text from a data store; tokenizing the unstructured text into one or more tokens, wherein a token is a character of the unstructured text; annotating the one or more tokens with one or more attribute labels, wherein an attribute label for a token is determined, at least in part, based on a word that the token originates from within the unstructured text; grouping the one or more tokens into one or more text segments based on the one or more attribute labels, wherein a set of one or more tokens that are annotated with an identical attribute label are grouped into a text segment, and wherein the one or more text segments define one or more attribute values; and storing the one or more attribute labels and the one or more attribute values within the data store.

2 . The computer-readable medium of claim 1 , the extracting further comprising normalizing at least one attribute value of the one or more attribute values.

3 . The computer-readable medium of claim 2 , the normalizing the at least one attribute value further comprising:
pairing an attribute value of the one or more attribute values with one or more target attribute values; selecting a target attribute value that has a highest probability of matching the attribute value; and replacing the attribute value with the selected target attribute value.

4 . The computer-readable medium of claim 3 , the extracting further comprising replacing at least one attribute label that is annotated for at least one token with at least one new attribute label in response to a user interaction.

5 . The computer-readable medium of claim 1 , wherein the unstructured text comprises a product description.

6 . The computer-readable medium of claim 1 , wherein the data store comprises a database.

7 . The computer-readable medium of claim 1 wherein the attribute label for the token is further determined, at least in part, based on a character-based conditional random field.

8 . The computer-readable medium of claim 1 , wherein the one or more tokens are character-based tokens.

9 . The computer-readable medium of claim 1 , wherein the attribute label for the token is further determined, at least in part, based on at least one of: whether the token is a lowercase character; a shape of the token; a punctuation of the token, one or more surrounding tokens; a size of the word that the token originates from within the unstructured text; a position of the token relative to the word that the token originates from within the unstructured text; or a position of the token relative to the unstructured text.

10 . The computer-readable medium of claim 1 , the extracting further comprising:
receiving one or more pre-defined attribute values; annotating one or more characters of the unstructured text with one or more attribute labels by matching the one or more pre-defined attribute values with one or more text segments of the unstructured text; and replacing at least one attribute label that is annotated for at least one character with at least one new attribute label in response to a user interaction.

11 . A computer-implemented method for extracting attribute values, the computer-implemented method comprising:
receiving data comprising unstructured text from a data store; tokenizing the unstructured text into one or more tokens, wherein a token is a character of the unstructured text; annotating the one or more tokens with one or more attribute labels, wherein an attribute label for a token is determined, at least in part, based on a word that the token originates from within the unstructured text; grouping the one or more tokens into one or more text segments based on the one or more attribute labels, wherein a set of one or more tokens that are annotated with an identical attribute label are grouped into a text segment, and wherein the one or more text segments define one or more attribute values; and storing the one or more attribute labels and the one or more attribute values within the data store.

12 . The computer-implemented method of claim 11 , further comprising normalizing at least one attribute value of the one or more attribute values,

13 . The computer-implemented method of claim 12 , the normalizing the at least one attribute value further comprising:
pairing an attribute value of the one or more attribute values with one or more target attribute values; selecting a target attribute value that has a highest probability of matching the attribute value; and replacing the attribute value with the selected target attribute value.

14 . The computer-implemented method of claim 13 , further comprising replacing at least one attribute label that is annotated for at least one token with at least one new attribute label in response to a user interaction.

15 . The computer-implemented method of claim 11 , wherein the attribute label for the token is further determined, at least in part, based on at least one of: whether the token is a lowercase character; a shape of the token; a punctuation of the token, one or more surrounding tokens; a size of the word that the token originates from within the unstructured text; a position of the token relative to the word that the token originates from within the unstructured text; or a position of the token relative to the unstructured text.

16 . A system for extracting attribute values, the system comprising:
a data reception module configured to receive data comprising unstructured text from a data store; a tokenization module configured to tokenize the unstructured text into one or more tokens, wherein a token is a character of the unstructured text; an annotation module configured to annotate the one or more tokens with one or more attribute labels, wherein an attribute label for a token is determined, at least in part, based on a word that the token originates from within the unstructured text; a token grouping module configured to group the one or more tokens into one or more text segments based on the one or more attribute labels, wherein a set of one or more tokens that are annotated with an identical attribute label are grouped into a text segment, and wherein the one or more text segments define one or more attribute values; and an attribute storage module configured to store the one or more attribute labels and the one or more attribute values within the data store.

17 . The system of claim 16 , further comprising a normalization module configured to normalize at least one attribute value of the one or more attribute values,

18 . The system of claim 17 ,
wherein the normalization module is further configured to pair an attribute value of the one or more attribute values with one or more target attribute values; wherein the normalization module is further configured to select a target attribute value that has a highest probability of matching the attribute value; and wherein the normalization module is further configured to replace the attribute value with the selected target attribute value.

19 . The system of claim 18 , further comprising a manual annotation module configured to replace at least one attribute label that is annotated for at least one token with at least one new attribute label in response to a user interaction.

20 . The system of claim 16 ,

Assignees

• Oracle Int Corp

Inventors

• KANANI Pallika Haridas
• WICK Michael Louis
• Pocock Adam Craig

Classifications

• G06Q30/00 Primary

  Commerce · CPC title

• G06F16/88

  Mark-up to mark-up conversion (conversion for visualization in web browsing G06F16/9577) · CPC title

• G06F40/284

  Lexical analysis, e.g. tokenisation or collocates · CPC title

• G06V10/768

  using context analysis, e.g. recognition aided by known co-occurring patterns · CPC title

• G06F17/30699

  Physics · mapped topic